CN107948233A

CN107948233A - The method of processing write requests or read request, interchanger, control node

Info

Publication number: CN107948233A
Application number: CN201610896118.6A
Authority: CN
Inventors: 陈灿
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2016-10-13
Filing date: 2016-10-13
Publication date: 2018-04-20
Anticipated expiration: 2036-10-13
Also published as: CN107948233B

Abstract

The embodiment of the present invention provides a kind of method, interchanger and the control node of processing write requests or read request.Forwarding table is preserved in the interchanger, and the interchanger possesses burst and redundant computation ability, after the interchanger receives the first write request, burst and computing redundancy burst are directly carried out to the data carried in first write request based on the forwarding table, then the second write request of corresponding number is generated, second write request is directly then sent respectively to corresponding memory node to complete data preservation.So, data in first write request only need to transmit 1 time on the interchanger with regard to that can complete to preserve, the waste of the live load and Internet resources of interchanger is thus substantially reduced, reduces the time delay of write request processing, greatly improves the treatment effeciency and performance of system.

Description

The method of processing write requests or read request, interchanger, control node

Technical field

The present invention relates to information technology field, and more particularly, to a kind of method of processing write requests or read request, Interchanger, control node.

Background technology

With the explosive growth of data volume, traditional external storage system due to performance and capacity can not linear expansion and It is difficult to meet the requirement of data center, distributed memory system comes into being.In distributed memory system, to ensure data High availability, by realizing the redundancy to the data between each memory node, can such as pass through redundant array of independent disks Modes such as (Redundant Array of Independent Disks, RAID) realizes data redundancy, so may insure During single memory node failure, the data are not lost, and business is not interrupted.For in distributed memory system in the prior art, when When host there are data to need write-in, write request is sent to any memory node, any memory node meeting by interchanger New write request is generated based on each burst after data in the write request are carried out Fragmentation, then again by each The new write request is sent respectively to other memory nodes by the interchanger and preserves again.

Find through inventor's analysis, have the following disadvantages in the prior art：

Interchanger live load weight：The data of write request from the host need to transmit 2 times on the interchanger Data write-in could be completed, causes network resources waste and interchanger live load weight.

The content of the invention

An embodiment of the present invention provides a kind of method of processing write requests or read request, interchanger, control node, Neng Goushi Data to be written only need to transmit on switches 1 time in present distributed memory system, thus reduce the interchanger Live load, save Internet resources, and is also greatly reduced the processing delay of write request, improves the distributed storage system The treatment effeciency and performance of system.

First aspect, there is provided a kind of method of processing write requests, wherein, the method is applied to interchanger, the friendship Be provided with forwarding table in changing planes in advance, the forwarding table includes a plurality of forwarding and records, each forwarding record comprising host address, Service switch address, volume, levels of redundancy numbering and memory node address, the described method includes：

The first write request from host is received, the first metadata and the first data are carried in first write request；

Based on forwarding table described in first metadata query to obtain K bars forwarding record, K is more than or equal to 2；

Fragmentation is carried out to first data based on the levels of redundancy numbering in K bars forwarding record, obtains K Burst；

Based on K the second write requests of the levels of redundancy numbering in K bars forwarding record and the generation of memory node address, institute State each second write request in K the second write requests and carry the second data, second data are in the K burst 1 burst；

Respectively K second write request is sent to K memory node.

With reference to first aspect, in the first possible implementation, first metadata includes the first source address, the One destination address, storage index and payload user data length, wherein first source address is the address of the host, described first Destination address is the address of the interchanger, and the payload user data length is used for the size for indicating first data；

It is then described to be specially based on forwarding table described in first metadata query：

Matched respectively based on first source address, the first destination address and storage index each described in the forwarding table Host address, service switch address and the volume of bar forwarding record are recorded with obtaining the K bars forwarding.

Optionally, first metadata can also include storage class, sequence number.

Optionally, the Fragmentation is multiple to be divided into first data based on the levels of redundancy numbering inquired about Data fragmentation, and at least one redundant slices are calculated based on the multiple data fragmentation, the multiple data fragmentation and described At least one redundant slices form the K burst corresponding to first data.

Optionally, each second write request in the K the second write requests also carries the second metadata, and described second Metadata includes the second source address, the second destination address, storage index, levels of redundancy numbering and payload user data length, wherein institute The address that the second source address is the interchanger is stated, second destination address is 1 forwarding note in K bars forwarding record The memory node address of record, the levels of redundancy numbering are that the levels of redundancy of 1 forwarding record in the K bars forwarding record is compiled Number, the payload user data length is used for the size for indicating 1 burst.

Optionally, second metadata can also include storage class, sequence number.

Optionally, the interchanger can be with creation state table, the shape handled for recording needle first write request State information.

With reference to first aspect, optionally, the method further includes：

The K memory node receives the second write request from the interchanger respectively；

Indexed with the storage in the second metadata entrained by second write request received and described superfluous Remaining Grade numbers store the second data carried in second write request received as key；

Second, which is sent, to the interchanger writes success message.

Optionally, the method further includes：What the interchanger received from the K memory node respectively second is write as Work(message；

First, which is sent, to the host writes success message.

Optionally, the interchanger be also based on it is described each second write as state table described in work(information updating, i.e., The second metadata information according to entrained by described second writes success message inquires about the state table, and it is right in the state table to change The mode of operation that should be gone.

Further, optionally, the interchanger can also delete the state table.

As can be seen that preserving forwarding table in the interchanger, then received when for the interchanger from the master After first write request of machine, directly the data carried in first write request can be divided based on the forwarding table Piece and computing redundancy burst, then generate second write request of corresponding number, then directly by second write request point Corresponding memory node is not sent to to preserve the data of its carrying, can thus complete what is carried in first write request The preservation of data.In embodiments of the present invention, the data in first write request only need to transmit 1 time on the interchanger It can just complete to preserve, avoiding as the prior art that needs transmission to preserve for 2 times, thus substantially reduces the work of interchanger Make the waste of load and Internet resources, decrease the time delay of a write request processing, greatly improve distributed storage system The treatment effeciency and performance of system.

Second aspect, also provides a kind of method for handling read request, and the method is applied to interchanger, in the interchanger Be provided with forwarding table in advance, the forwarding table includes a plurality of forwarding and records, and each forwarding record is handed over comprising host address, service Change planes address, volume, levels of redundancy numbering and memory node address, the described method includes：

The first read request from host is received, trinary data is carried in first read request；

The forwarding table is inquired about based on the trinary data to obtain K articles of forwarding record, K is more than or equal to 2；

Based on K the second read requests of the levels of redundancy numbering in K bars forwarding record and the generation of memory node address；

Respectively K second read request is sent to K memory node.

Optionally, the trinary data includes the 3rd source address, the 3rd destination address, storage index and payload user data length Degree, wherein the 3rd source address is the address of the host, the 3rd destination address is the address of the interchanger, described Payload user data length is used for the size for indicating the data that needs are read；

It is then described to be specially based on the trinary data inquiry forwarding table：

Based on the 3rd source address in the trinary data, the 3rd destination address and the storage index point Not matching each described in the forwarding table forwards the host address recorded, service switch address and volume to obtain the K Bar forwarding record.

Optionally, the interchanger can be with creation state table, the shape handled for recording needle first read request State information.

Optionally, each second read request in the K the second read requests also carries the 4th metadata, and the described 4th Metadata includes the 4th source address, the 4th destination address, storage index, levels of redundancy numbering and payload user data length, wherein institute The address that the 4th source address is the interchanger is stated, the 4th destination address is 1 article of forwarding note in described K articles forwarding record The memory node address of record, the levels of redundancy numbering are that the levels of redundancy of 1 forwarding record in the K bars forwarding record is compiled Number, the payload user data length is used for the size for indicating the data that needs are read, and the method further includes：

Each memory node in the K memory node receives the second read request from the interchanger；

Compiled with the storage index in the 4th metadata entrained by second read request and the levels of redundancy Number as key read data；

Second is sent to the interchanger and runs through message, and described second runs through the data that message carries the reading.

Optionally, the method further includes：

The interchanger receives second of each memory node in the K memory node and runs through respectively to disappear Breath；

The data carried in message are run through to the K second to recombinate；

First is sent to the host and runs through message, and described first runs through the number after the restructuring is carried in message According to.

Optionally, the interchanger be also based on it is described each second run through state table described in information updating, i.e., Metadata information according to entrained by described second runs through message inquires about the state table, changes in the state table and corresponds to row Mode of operation.

Further, optionally, the interchanger can also delete the state table.

By the above embodiment, it is necessary to which the data read are passing through the interchanger to the master from multiple memory nodes Only needed on the direction of machine by the interchanger 1 time, avoiding as the prior art that needs transmission to complete data 2 times Read, thus substantially reduce the waste of the live load and Internet resources of interchanger, decrease read request processing when Prolong, greatly improve the treatment effeciency and performance of the distributed memory system.

The third aspect, also provides a kind of interchanger, the interchanger include receiver, memory, processor, transmitter and RAID engine, wherein：

The receiver, for receiving the first write request from host, first yuan of number is carried in first write request According to this and the first data；

The memory, for forwarding table memory, the forwarding table includes a plurality of forwarding and records, each forwarding record bag Containing host address, service switch address, volume, levels of redundancy numbering and memory node address；

The processor, for parse first write request with obtain first metadata and it is described first number According to based on forwarding table described in first metadata query to obtain K bars forwarding record, K is more than or equal to 2；

The RAID engine, for based on the K bars forwarding record in levels of redundancy numbering to first data into Row Fragmentation, obtains K burst；

The processor, is additionally operable to based on the levels of redundancy numbering in K bars forwarding record and the life of memory node address Into K the second write requests, each second write request in the K the second write requests carries the second data, second number According to for 1 burst in the K burst；

The transmitter, for sending K second write request to K memory node respectively.

With reference to the third aspect, in the first possible implementation, first metadata includes the first source address, the One destination address, storage index and payload user data length, wherein first source address is the address of the host, described first Destination address is the address of the interchanger, and the payload user data length is used for the size for indicating first data；

Then the processor is recorded for described based on forwarding table described in first metadata query with obtaining the forwarding of K bars Specially：

Optionally, the receiver, the receiver, is additionally operable to receive respectively every in the K memory node The second of one memory node writes success message；

The processor, is additionally operable to second based on each memory node in the K memory node Write success message generation first and write success message；

The transmitter, is additionally operable to write success message to host transmission described first.

As can be seen that after the interchanger receives first write request from the host, institute can be based on State forwarding table and burst and computing redundancy burst directly are carried out to the data carried in first write request, then corresponding of generation Several second write requests, is then directly sent respectively to corresponding memory node by second write request and preserves its carrying Data, can thus complete the preservation of data carried in first write request.In embodiments of the present invention, described Data in one write request, which only need to transmit 1 time on the interchanger, just can complete preservation, avoid the picture prior art that Needing to transmit 2 times to preserve, and thus substantially reduce the waste of the live load and Internet resources of interchanger, decrease The time delay of write request processing, greatly improves the treatment effeciency and performance of the distributed memory system.

Fourth aspect, also provides a kind of interchanger, and the interchanger includes receiver, memory, processor and transmitter, Wherein：

The receiver, for receiving the first read request from host, Three-ary Number is carried in first read request According to；

The processor, for parsing first read request to obtain the trinary data, based on the ternary For forwarding table described in data query to obtain K bars forwarding record, K is more than or equal to 2, based on redundancy in K bars forwarding record etc. K the second read requests of level numbering and the generation of memory node address；

The transmitter, for sending K second read request to K memory node respectively.

With reference to fourth aspect, in the first possible implementation, the interchanger further includes RAID engine；

The receiver, the second of each memory node for being additionally operable to receive respectively in the K memory node Run through message；

The processor, second for parsing each memory node in the K memory node reads Completion message described second runs through data entrained by message to obtain each；

The RAID engine, recombinates for running through the data carried in message to the K second；

The processor, is additionally operable to generation first and runs through message, described first, which runs through message, carries the restructuring Data；

The transmitter, message is run through for sending first to the host.

5th aspect, also provides a kind of control node, and the control node includes receiver, processor, memory, transmitting Device, wherein：The transmitter, for sending interchanger ability information query message to interchanger, sends to memory node and stores Node capacity information query message；The receiver, rings for receiving the interchanger ability information from the interchanger and inquiring about Message is answered, the interchanger ability information inquiry response message carries interchanger ability information, and reception comes from the memory node Memory node ability information inquiry response message, the memory node ability information inquiry response message carries memory node energy Force information；The processor, is additionally operable to parse the interchanger ability information inquiry response message to obtain the exchange function Force information, parses the memory node ability information inquiry response message to obtain the memory node ability information；It is described to deposit Reservoir, for preserving the interchanger ability information and the memory node ability information；The receiver, is additionally operable to receive master Machine storage demand；The processor, is additionally operable to based on the host storage demand and the interchanger ability information and described Memory node ability information is the host assignment service switch, distributes multiple memory nodes and configure forwarding table；It is described Transmitter, is additionally operable to the address of the service switch being sent to the host, and the forwarding table is sent to the service Interchanger.

Optionally, the interchanger ability information can include interchanger ID, whether switch address, interchanger possess point Piece and redundant computation ability；The memory node ability information can include：Memory node address, memory node capacity.

Optionally, the control node can be sent to each interchanger and exchange function respectively in a manner of unicast Force information query message, memory node ability information query message is sent in a manner of unicast to each memory node respectively, Can also send above-mentioned message in a broadcast manner.

The service switch should possess burst and redundant computation ability.Optionally, the processor can be random Service switch of 1 interchanger for possessing burst and redundant computation ability as the host is selected, network can also be selected What host described in distance was nearest in route possesses service exchange of the interchanger of burst and redundant computation ability as the host Machine or currently calculating is selected from multiple interchangers for possessing burst and redundant computation ability with reference to load balancing principle Service switch of the less interchanger of task as the host.

Optionally, the processor can be selected memory node at random or consider that load balancing principle is preferential Select the relatively low memory node of currently used rate or select and the service switch is nearest in network topology deposits Store up node.

Optionally, the forwarding table can include following information：Host address, service switch address, volume, levels of redundancy Numbering and memory node address etc..

The control node provided by this embodiment of the present invention, it is possible to achieve to the automatic of the distributed memory system Change centralized configuration, avoiding needs that different interchangers, storage is respectively configured by manually configuring and needing in the prior art The problem of node, the allocative efficiency being greatly improved in distributed memory system.Further, by the control node come Forwarding table is configured, interchanger is forwarded to operation according to the forwarding table that the control node configures, and is achieved in control/forwarding point From it is possible thereby to greatly promote the performance of the distributed memory system.

Brief description of the drawings

In order to illustrate the technical solution of the embodiments of the present invention more clearly, it will make below to required in the embodiment of the present invention Attached drawing is briefly described.

Fig. 1 is a kind of configuration diagram of distributed memory system provided in an embodiment of the present invention.

Fig. 2 is a kind of structure diagram of message provided in an embodiment of the present invention.

Fig. 3 is a kind of indicative flowchart of processing write requests provided in an embodiment of the present invention.

Fig. 4 is a kind of indicative flowchart for handling read request provided in an embodiment of the present invention.

Fig. 5 is a kind of structure diagram of interchanger provided in an embodiment of the present invention.

A kind of Fig. 6 structure diagrams of control node provided in an embodiment of the present invention.

Embodiment

Below in conjunction with the attached drawing in the embodiment of the present invention, the technical solution in the embodiment of the present invention is carried out clear, complete Site preparation describes.

As shown in Figure 1, it is a kind of system architecture of exemplary distributed memory system provided in an embodiment of the present invention. In the distributed memory system, multiple main frames pass through any progress in any interchanger and multiple memory nodes respectively Communication, each interchanger are connected with the other interchangers of at least one, each host can be connected with least one interchanger, often One memory node can also be connected with least one interchanger.Preferably, the distributed memory system can also include control Node, the control node can be communicated with any one or more interchangers, and the control node can also pass through Any one or more interchangers are communicated with any host and any memory node.Those skilled in the art It is understood that in actual networking, there can be one or more hosts, can there is one or more interchangers and at least 2 Memory node more than a.The quantity of host, interchanger and memory node determines that Fig. 1 is only based on practical business demand A kind of exemplary networking structure is provided, actual networking is not formed and is limited.

In distributed memory system provided in an embodiment of the present invention, on one or more interchangers realize to from appoint The data of the write request of one host carry out burst and redundant computation, then will correspond to the write request by the interchanger Each bursts (data fragmentation or redundant slices) of data generate new write request, and each described new is write respectively Request is sent to different memory nodes to complete the preservation to carrying data in the write request.Host is come from for each Write request, be separately sent to preserve on each memory node with regard to each burst after burst processing is done on the interchanger, this Sample, the data of each write request from host are only needed to transmit 1 time on switches and can be saved in memory node, Avoid interchanger live load weight caused by needing to transmit 2 times on switches as the prior art, Internet resources wave Take and the problem of time delay is big.

In the embodiment of the present invention, in the starting stage, control node sends interchanger ability letter respectively to each interchanger Cease query message, it is desirable to which the interchanger reports its ability information.Each described interchanger is receiving the interchanger ability After information inquiry message, interchanger ability information inquiry response message, the exchange function can be returned to the control node Force information inquiry response message can include following information：Whether interchanger ID, switch address, interchanger possess burst and superfluous Remaining computing capability.Explanation is needed exist for, the embodiment of the present invention is not required for each in the distributed memory system and hands over Change planes all and possess burst and redundant computation ability, in actual networking, include one or more conventional switch, it is only necessary to ensure At least 1 interchanger possesses burst in the system of deployment and redundant computation ability is implemented with the present invention.Preferably, from Reliability perspectives consider, it is proposed that the interchanger for possessing burst and redundant computation ability that deployment is more than 2.It is each such as table 1 below The ability information situation of interchanger (table 1 is illustrated exemplified by having 3 interchangers in system)：

Table 1

The control node sends memory node ability information query message to each memory node, it is desirable to described each A memory node reports its ability information.Each memory node after the memory node ability information query message is received, Following information can be returned to the control node by memory node ability information inquiry response message：Memory node address, Memory node capacity.Such as table 2 below, be each memory node ability information (table 2 exemplified by having 6 memory nodes in system into Row explanation)：

Memory node address	Memory node capacity
		200.1.1.110	20T
200.1.1.111	20T
		200.1.1.112	20T
200.1.1.113	20T
		200.1.1.114	20T
200.1.1.115	30T

Table 2

The control node can send above-mentioned inquiry to each interchanger or memory node respectively in the form of unicast Message or in a broadcast manner the broadcast them query message in network；It will be appreciated by persons skilled in the art that Above-mentioned " interchanger ability information query message ", " interchanger ability information inquiry response message ", " memory node ability information Query message " and " memory node ability information inquiry response message " are only a kind of for example, the message name not structure in itself Into being particularly limited to, in different practices, as long as the message for possessing similar effect both falls within the protection domain of the embodiment of the present invention.

The control node is collected into each interchanger and each memory node in the distributed memory system Ability information after, can actively or answer host requirements to host configure storage resource.Such as according to the capacity requirement of host and The requirement of levels of redundancy is the host assignment interchanger, memory node, and to the forwarding table inside the interchanger of the distribution Item is configured.The required memory capacity size of capacity requirement, that is, host, such as 10T or 20T；The levels of redundancy can be with It is RAID level, such as RAID1, RAID 3 or RAID 5 etc.；Preferably, the host can also further will in levels of redundancy The memory node number for forming the RAID level is sought, such as RAID 1/3, that is, represents to be provided with the storage respectively from 3 memory nodes Realize RAID 1 in source.Illustratively, it is as shown in table 3 below：Host have sent 2 storage demands, first need to the control node Seeking Truth requirement is that it distributes the volume that volume identification is LUN A, its capacity is 10T, and levels of redundancy is RAID 1/3, i.e. the LUN A's Levels of redundancy is RAID 1, is distributed on 3 memory nodes；Second demand is that requirement is that its distribution volume identification is LUN B Volume, its capacity is 50T, and levels of redundancy is RAID 5/5, i.e. the levels of redundancy of the LUN B is RAID 5, is distributed in 5 storage sections Point on.

Host address	Volume identification	Capacity requirement	Levels of redundancy
				200.1.1.3	LUN A	10T	RAID 1/3
200.1.1.3	LUN B	50T	RAID 5/5

Table 3

After the control node receives the storage demand from host, based on the foregoing respective switch being collected into and The ability information of memory node configures resource for the storage demand of the host, specifically includes：

1) it is the host specified services interchanger.

The service switch needs to carry out Fragmentation and calculating to the data in the write request from the host Redundant slices, therefore service switch can only be selected from the interchanger for possessing burst and redundant computation ability.If the distribution There is multiple switch all to possess burst and redundant computation ability in formula storage system, then the control node can be selected at random Select service switch of 1 interchanger as the host；It can also select the nearest friendship of host described in the upper distance of network route Change planes, the nearest interchanger of host described in the upper distance of network described here route refers to message that the host is sent on network The interchanger that can be routed to by fewest number of hops；Alternatively, with reference to load balancing principle, possess burst and redundancy from multiple Service switch of the less interchanger of current calculating task as the host is selected in the interchanger of computing capability.For example, With reference to table 1, for above-mentioned first storage demand, the control node can select interchanger 2 to be used as service switch；For Above-mentioned second storage demand, the control node can select interchanger 3 to be used as service switch.

2) it is the host assignment memory node.

The control node selects suitable memory node come to the host according to the ability information of each memory node Storage service is provided.For example, such as above-mentioned first storage demand, the host requirement is that its distribution volume identification is LUN A Volume, its capacity is 10T, and levels of redundancy is RAID 1/3, this means that the control node needs to select 3 storage sections Point, each memory node will at least have the available storage of 3.4T.The control node is receiving the demand of above-mentioned host Afterwards, the ability information of each memory node is checked, selecting 3 in the multiple memory nodes met the requirements from available storage deposits Storage node to provide storage service for the storage demand of the host.The control node can multiple be deposited what is met the requirements 3 memory nodes are selected at random in storage node；Can also consider load balancing principle, it is relatively low preferentially to select currently used rate Memory node, so ensures that the utilization rate of each memory node is more balanced, here, so-called utilization rate refers to that some is deposited as far as possible The capacity being allocated away on storage node accounts for the ratio of all told of the memory node；In addition, the control node can be with Consider the network topology between each memory node and the service switch, can preferentially select and be exchanged with the service Machine memory node nearest in network topology.For example, in distributed memory system as shown in Figure 3, it is assumed that interchanger 2 is The service switch of host 1, then the control node can preferentially select memory node 0 and memory node 1 comes for the master Machine provides service, because this 2 memory nodes are connected directly with interchanger 2, from network topology with the interchanger 2 most Closely, and then from memory node 2 selected into memory node n+1.For example, referring to table 2, for above-mentioned first storage demand, Optional address is that 3 memory nodes of 200.1.1.110,200.1.1.111,200.1.1.112 are the first of the host A storage demand provides storage service；For above-mentioned second storage demand, may be selected address for 200.1.1.111, 200.1.1.112,5 memory nodes of 200.1.1.113,200.1.1.114,200.1.1.115 are the second of the host A storage demand provides storage service.

3) forwarding table is configured based on the service switch specified and for selected memory node.

Based on 2 foregoing storage demands and the selected service switch of the control node and memory node, The control node configures the forwarding table, and the forwarding table, which includes, can include a plurality of forwarding record, each forwarding record It can include following information：Host address, service switch address, volume, levels of redundancy numbering and memory node address；Its In, levels of redundancy numbering can represent that wherein RAID M/N, that is, levels of redundancy, the M represents RAID level number with RAID M/N_X, Its value range is M>=0, it will be appreciated by those skilled in the art that being, the usual values of M are 0,1,3,5,6,10,50 etc.；Wherein The N is represented to form the memory node number of the RAID level number, such as RAID 1/3, is represented to need to be realized by 3 memory nodes RAID 1；Wherein X values are [0 ... N-1], are respectively intended to represent the 0th in the RAID M that are made of N number of memory node ... The N-1 member's memory node, naturally it is also possible to which value is [1 ... N], is respectively intended to represent what is be made of N number of memory node The 1st in RAID ... n-th member's memory node；Here, the forward-path of memory node address, that is, respective record. It is as shown in table 4 below, it is an exemplary forwarding table：

Table 4

As upper table 4 is understood, the clothes that the LUN A for the host for being 200.1.1.3 corresponding to address and address are 200.1.1.101 Business interchanger, has 3 forwarding records, it is respectively RAID 1/3_0, RAID 1/3_ that the levels of redundancy of this 3 forwarding records, which is numbered, 1st, RAID 1/3_2, its corresponding its address of 3 memory nodes be respectively 200.1.1.110,200.1.1.111, 200.1.1.112, namely show that then the data are needed with RAID if the data from the host need to write LUN A 1 mode is respectively written into 3 memory nodes that address is respectively 200.1.1.110,200.1.1.111,200.1.1.112.

The forwarding table actively can be sent respectively to each clothes by the control node after the forwarding table is configured Business interchanger, locally preserves the forwarding-table item by each service switch or comes from a certain friendship receiving After the forwarding table inquiry request changed planes, then the forwarding table issued into the interchanger.Further, the control node can be with Notify the address that the host is its service switch specified, all kinds of business association messages of so follow-up host can be straight The service switch is given in sending and receiving.

It will be apparent to a skilled person that above-mentioned table 1,2,3,4 be only the embodiment of the present invention in be clearly to retouch The exemplary explanation stated and done, but it does not form presentation mode, the preserving type for limiting above- mentioned information, in concrete practice In, preservation and the record of above- mentioned information can be realized with the flexible Application such as various ways such as chained list, file, daily record, these are different real Now it is not considered that beyond the scope of this invention, the embodiment of the present invention is not stated one by one herein.

In embodiments of the present invention, by the above-mentioned means, the automation collection to the distributed memory system can be realized Middle configuration, avoiding needs that different interchangers, memory node is respectively configured by manually configuring and needing in the prior art The problem of, the allocative efficiency that is greatly improved in distributed memory system.Further, configured by the control node Forwarding table, interchanger are forwarded to operation according to the forwarding table that the control node configures, and are achieved in control/forwarding separation, It is possible thereby to greatly promote the performance of the distributed memory system.

As shown in Fig. 2, it is host in the distributed memory system provided in an embodiment of the present invention and interchanger, interchanger The structure diagram of the message to communicate between memory node, communicates between host and interchanger, interchanger and memory node Message can include metadata, payload user data, and wherein metadata is the part that each message has to include, payload user data It is optional part.

The metadata can include 4 message header of source address, destination address and layer, wherein the source address is used to represent The address of the entity of the message is sent, if such as the message is that source address is the address of the host, if this disappears from host transmission Breath is sent from interchanger, then source address is the address of the interchanger；The destination address connects for representing that the message is final Receive main body.

4 message header of layer can include storage head, operational order, levels of redundancy numbering, sequence number and payload user data Length.Wherein described storage head can be that storage class (Storage Type) and storage index (Storage Index), described Storage class is used for the access type for representing storage, such as：Block (Block), file (File), object (Object) etc.；It is described Storage index is corresponding with the storage class, such as：If storage class is block, storage index is volume ID+logical address, If storage class is file, storage index is file system ID (File system ID)+catalogue (Directory)+text Part name (File name)+offset address (Offset), if storage class is object, storage index is key (Key)+version Number (Version).It should be noted that because of the structure for storing index itself corresponding to different storage classes completely not Together, so in embodiments of the present invention, the storage class field can be Optional Field.The operational order is used to represent this Message needs the action type that carries out, in the embodiment of the present invention, the operational order can be " writing ", " reading ", " writing complete ", " writing failure ", " running through ", " reading failure " etc., wherein " writing " indicates that data need to write in memory node, " reading " represents to need Data are read from memory node, " writing complete ", " writing failure " represent 2 kinds of write operation respectively as a result, " running through ", " reading failure " represents 2 kinds of results of read operation respectively.The levels of redundancy is numbered with reference to the above-mentioned description as described in forwarding table of table 4, It is not repeated herein, under normal circumstances, levels of redundancy volume can need not be carried being sent to from host in the message of interchanger Number, need to carry the field in the message for issuing memory node from interchanger.Institute's Serial No. Optional Field, for from host The context identification of the message sent, is obtained by host order is cumulative, such as the sequence number for first message that host is sent Can be Ox0000, the sequence number of second message can be Ox0001, and the sequence number of the 3rd message can be Ox0002, with this Analogize.The payload user data length is also Optional Field, for describing the length of data to be written or needing the data read Length, in the embodiment of the present invention, the value of the payload user data length can be 128 bytes, 256 bytes, 512 bytes, 1024 bytes or 2048 bytes etc..

The payload user data is data to be written or the data of reading.

Message structure and described above shown in Fig. 2 are only that the exemplary of the embodiment of the present invention describes, each field Title etc. do not form the specific of message structure limited, in concrete practice, the title of each field can have change in message Change, but as long as the information of its carrying is realized it is not considered that exceeding this hair without departing from the essence of foregoing description, these differences in itself Bright scope, the embodiment of the present invention are not stated one by one herein.

As shown in figure 3, it is the method provided in an embodiment of the present invention handled the write request from host.

Step 300：The first write request from host is received, the first metadata and the are carried in first write request One data.

Completed in storage demand of the above-mentioned control node based on the host with postponing, be notified that the host is selected for it Service switch, then when host have data need write-in when can send the first write request, carried in first write request First metadata and the first data.With reference to Fig. 2 and foregoing message structure, in the embodiment of the present invention, first yuan of number According to can include：First source address, the first destination address, storage index, operational order and payload user data length, wherein：Institute State the address that the first source address is the host, i.e. 200.1.1.3；First destination address exchanges for the service of the host The address of machine, i.e. 200.1.1.101；The storage index is LUN A+Ox0000FFFF, that is, represents that the write request will be volume A's Initial address is the first entrained data of write-in at Ox0000FFFF；The operational order is " writing "；The payload user data length It can be 512 bytes to spend value.

Optionally, in the embodiment of the present invention, first metadata can also include storage class, such as the storage class For " block ", storage index corresponding thereto is as described above LUN A+Ox0000FFFF；First metadata can also wrap Sequence number is included, the value of the sequence number is a based on the message sent before first write request is sent for the host Data add 1, and the sequence number of such as first write request can be Ox1234.

Step 301：Based on the first metadata query forwarding table.

The service switch is received after first write request, it is necessary to by inquiring about the forwarding table determines such as where Manage the write request.Forwarding table is to obtain K bars as described in can be the first metadata query carried based on first write request Forwarding record, the value of K described here are more than or equal to 2；By taking above-mentioned table 4 as an example, it can be found in the table 4 and host 200.1.1.3, service switch 200.1.1.101 and LUN A is corresponding with 3 forwarding records, its corresponding levels of redundancy numbering It is respectively RAID1/3_0, RAID 1/3_1, RAID 1/3_2, namely the LUN A corresponding to the host 200.1.1.3 respectively Levels of redundancy be RAID 1/3, corresponding forward-path is 3 memory nodes respectively for it, its address is respectively： 200.1.1.110、200.1.1.111、200.1.1.112。

Step 302：Fragmentation is carried out to the first data in first write request.

Levels of redundancy numbering based on inquired K bars forwarding record carries out the first data in first write request Fragmentation, as it was previously stated, the levels of redundancy numbering inquired is respectively RAID 1/3_0, RAID 1/3_1, RAID 1/3_2, I.e. its levels of redundancy is RAID 1/3, and it is RAID 1 also just to say RAID level, is made of 3 memory nodes, because in RAID 1 In the case of be need not to first data carry out striping processing, it is not required that for first data carry out redundancy Calculating, the protections of RAID 1 are that first data are carried out mirror image, in the present embodiment, because levels of redundancy is RAID 1/3, because This in this case, first data directly can be replicated 3 parts again by the Fragmentation, 3 replicated part first Data collectively form the burst corresponding to the first data in first write request, respectively by corresponding 3 storages section Put to preserve.

One of ordinary skill in the art will appreciate that such as table 4, levels of redundancy numbering be respectively RAID5/5_0, Its levels of redundancy is RAID 5/5 in the case of RAID5/5_1, RAID5/5_2, RAID5/5_3, RAID5/5_4, is represented by 5 Memory node forms RAID 5, then needs first data being divided into 4 data fragmentations, and calculate based on first data 1 redundant slices, 5 points corresponding to first data are made of 4 data fragmentations and 1 redundant slices Piece, each of which burst have 1 memory node to preserve respectively.It will be appreciated by persons skilled in the art that for the first number According to how burst and whether should needing computing redundancy burst and need to calculate each and every one more redundant slices, according to RAID technique Basic principle can directly be determined based on the levels of redundancy inquired, herein need not elaborate.Below also It is to be illustrated so that levels of redundancy is RAID 1/3 as an example.

Step 303：Send the second write request.

K the second write requests are generated based on the levels of redundancy numbering in K bars forwarding record and memory node address, point K second write request is not sent to K memory node.As in step 301, because the levels of redundancy inquired is RAID 1/3, be corresponding to 3 levels of redundancy numbering RAID 1/3_0, RAID 1/3_1, RAID 1/3_2, respective stored node address 200.1.1.110,3 memory nodes of 200.1.1.111,200.1.1.112, so the service switch can build 3 Second write request, each second write request include the second metadata and the second data, institute in 3 second write requests It is respectively one of above-mentioned 3 bursts to state the second data, the second source in second metadata in 3 second write requests Address is 200.1.1.101, that is, represents that second write request is sent by the service switch that address is 200.1.1.101；Second Destination address is respectively 200.1.1.110 or 200.1.1.111 or 200.1.1.112, represents that 3 second write requests need It is sent respectively to the memory node that address is 200.1.1.110 or 200.1.1.111 or 200.1.1.112；Described second Operational order in metadata is all " writing "；In second metadata storage index with it is described in first metadata Storage index is identical, is still LUN A+Ox0000FFFF；The payload data length is 512 bytes.

Further, because second write request is to be sent to memory node by the interchanger, therefore described Also need to include the levels of redundancy numbering in second metadata of two write requests, with reference to above-mentioned table 4, writing for described 3 second please Levels of redundancy numbering in the second metadata asked is RAID1/3_0, RAID 1/3_1, RAID 1/3_2 respectively.

Optionally, as previously indicated, the second metadata of second write request can also include storage class, its value It is identical with the storage class value in first metadata, it is such as " block "；Second metadata of second write request may be used also With including sequence number, its value is identical with the sequence number value in first metadata, is such as Ox1234.

Optionally, data check, for example preceding institute of computational methods of data check can also be included in second write request State, details are not described herein.

The service switch by 3 second write requests be sent respectively to address for 200.1.1.110, 200.1.1.111,3 memory nodes of 200.1.1.112.

Optionally, the service switch carried out after first write request is received inquiry forwarding table, burst and , can be with creation state table during sending second write request, the state handled for recording needle the first write request Information, increases status information on the basis of above-mentioned table 4, and the mode of operation in the state table is in first metadata Operational order, is recorded as " writing " here；Sequence number value in the state table is the sequence number in first metadata, example Such as can be Ox1234；Payload length value in the state table is the payload user data length in first metadata, example Such as can be 512 bytes.Shown in table 5 specific as follows：

Table 5

Further, herein it should be noted that above-mentioned state table is created as optional step, and the establishment state Table can be after step 303 or be created that after step 302, naturally it is also possible to be after step 301 Create.

Step 304：Preserve data.

The K memory node receives second write request and preserves the second data of its carrying respectively.With reference to above-mentioned Step, 3 memory nodes of address 200.1.1.110,200.1.1.111,200.1.1.112 are received from the clothes respectively Second write request of business interchanger, then preserves the second data entrained in the second received write request.

Preferably, in embodiments of the present invention, the memory node can preserve described second in a manner of object stores Data, one of ordinary skill in the art will appreciate that object storage is carried out in a manner of key-value (i.e. Key-Value) Preserve, wherein key (Key) is used to index the described value (Value) preserved.Value (i.e. Value) described here is second number According to key (i.e. Key) corresponding with value can be then storage index (the i.e. volume ID in the second metadata in second write request + logical address) and levels of redundancy numbering form, in the embodiment of the present invention, the value of the key can be exactly the volume ID+patrol Collect address+levels of redundancy numbering；Or the value of the key can also be to the volume ID+logical address+levels of redundancy numbering Carry out Hash operation and obtain.In this way, when the follow-up memory node needs to read data, it is possible to directly with the volume ID+ Logical address+levels of redundancy numbering indexes corresponding data as key.

Step 305：Memory node sends second to the service switch respectively and writes success message.

After the K memory node completes the preservation of second data in second write request, to the service Interchanger sends second and writes success message, and as described in abovementioned steps, K is 3 in the present embodiment.Message structure described in reference Fig. 2, Described second, which writes success message, only needs to include meta-data section, it is not necessary to includes payload user data and data check part.Its In, second metadata for writing success message is specially：

Source address：The address of as described memory node, such as respectively 200.1.1.110 or 200.1.1.111 or 200.1.1.112；

Destination address：The address of the service switch, such as 200.1.1.101；

Storage class：With the value of the storage class in second write request, such as " block "；

Storage index：With the value of the storage index in second write request, such as " LUN A+Ox0000FFFF "；

Operational order：Write complete；

Levels of redundancy is numbered：Such as it is respectively " RAID 1/ with the value of the levels of redundancy numbering in second write request 3_0、RAID 1/3_1、RAID 1/3_2”；

Sequence number：With the value of the sequence number in second write request, such as " Ox1234 ".

Because described second, which writes success message, need not carry any data, net also without data in the metadata This field of lotus length.One of ordinary skill in the art will appreciate that reference is previously with regard to the first write request and second The description of write request, described second writes the sequence number of meta-data section in success message, two fields of storage class are also optional 's.

Step 306：The service switch sends first to the host and writes success message.

The service switch receive respectively second from foregoing K memory node write success message after, It may determine that and complete data storage for first write request, therefore send first to the host and write success message, with Described second writes that success message is identical, and described first, which writes success message, only needs to include meta-data section, it is not necessary to includes data Payload and data check part.Wherein, first metadata for writing success message is specially：

Source address：The address of as described service switch, such as 200.1.1.101；

Destination address：The address of the host, such as 200.1.1.3；

Storage class：With the value of the storage class in first write request, such as " block "；

Storage index：With the value of the storage index in first write request, such as " LUN A+Ox0000FFFF "；

Operational order：Write complete；

Sequence number：With the value of the sequence number in first write request, such as " Ox1234 ".

Because described second, which writes success message, need not carry any data, net also without data in the metadata This field of lotus length；As it was previously stated, in the service switch with the message between the host, it is not necessary to carry redundancy etc. Level number field.One of ordinary skill in the art will appreciate that with reference to be previously with regard to the first write request and second write please The description asked, described first writes the sequence number of meta-data section in success message, two fields of storage class are also optional.

Further, if the service switch has establishment state table after first write request is received, at this In link, the service switch often receives 1 and second writes success message from the memory node, just by the state table Mode of operation in (such as table 5) is revised as " being write as work(", specific as shown in table 6, and the service switch is being received from described After the second of each memory node writes success message, state table is inquired about according to the second metadata information of its carrying, by the shape The mode of operation that row is corresponded in state table is revised as " being write as work("：

Table 6

Further, success message is write to trigger the state table when receiving described second from 3 memory nodes Mode of operation be all changed to " being write as work(", then illustrate that first write request is handled successfully, then can to the host send institute State first and write success message, and the state table can be deleted.

The write request processing procedure of above-described embodiment is illustrated so that levels of redundancy is RAID 1/3 as an example, this area Technical staff it is understood that levels of redundancy can be based on RAID 0, RAID 1, RAID 3, RAID 5, RAID 10 or The different RAID level such as RAID 50 and a variety of of corresponding RAID level can be realized by the memory node of varying number Situation, then in the case of different levels of redundancy, should be based on corresponding RAID level in step 302 and form the RAID level Other memory node number to carry out Fragmentation and computing redundancy burst, the Fragmentation and calculating to first data Redundant slices belong to the basic principle of RAID technique, therefore the embodiment of the present invention does not carry out in detail for different levels of redundancy one by one Explanation.

From above-described embodiment as can be seen that preserving forwarding table in the interchanger, then received when for the interchanger To after first write request from the host, the forwarding table can be based on directly to being carried in first write request Data carry out burst and computing redundancy burst, second write request of corresponding number is then generated, then directly described in general Second write request be sent respectively to corresponding memory node preserve its carrying data, can thus complete described first write please Ask the preservation of the data of middle carrying.In embodiments of the present invention, the data in first write request are only needed in the service 1 time is transmitted on interchanger just can complete preservation, and avoiding as the prior art that needs to transmit 2 times and could preserve, thus significantly Reduce the waste of the live load and Internet resources of interchanger, decrease the time delay of a write request processing, carry significantly The treatment effeciency and performance of the distributed memory system are risen.

As shown in figure 4, it is that the read request from host is directed in distributed memory system provided in an embodiment of the present invention Processing procedure.

Step 400：The first read request from host is received, trinary data is carried in first read request.

When the host has the demand for reading data, the first read request can be sent, institute can be carried in first read request State trinary data.The trinary data can include：3rd source address, the 3rd destination address, storage index and data are net Lotus length, wherein：Wherein described 3rd source address be the host address, i.e. 200.1.1.3；3rd destination address is The address of the interchanger, i.e. 200.1.1.101；The storage index is LUN A+Ox0000FFFF, that is, represents the read request Data are read since the initial address of volume A is at Ox0000FFFF；The operational order is " reading "；The payload user data length Degree is used for the size value for indicating the data that needs are read, and can be such as 512 bytes, demonstrates the need for indexing from above-mentioned storage signified The initial address of the volume A shown is the continuous data for reading 512 bytes at Ox0000FFFF.

Optionally, in the embodiment of the present invention, the trinary data can also include storage class, such as the storage class For " block ", storage index corresponding thereto is as described above LUN A+Ox0000FFFF；The trinary data can also wrap Sequence number is included, the value of the sequence number is a based on the message sent before first read request is sent for the host Data add 1, and the sequence number of such as first read request can be Ox5678.

Step 401：Forwarding table is inquired about based on the trinary data.

The service switch is received after first read request, it is necessary to determine how to handle this by inquiring about forwarding table Read request, can be based on the 3rd source address, the 3rd destination address and the storage in the trinary data Index matches each described in the forwarding table and forwards the host address recorded, service switch address and volume to obtain respectively The K bars forwarding record, the value of K is more than or equal to 2.It can be found in the table 4 and host 200.1.1.3, service Interchanger 200.1.1.101 and LUN A are corresponding with 3 forwarding records, its corresponding levels of redundancy numbering is respectively RAID respectively 1/3_0, RAID 1/3_1, RAID 1/3_2, namely be corresponding to the levels of redundancy of the LUN A of the host 200.1.1.3 RAID 1/3, the corresponding forward-path of its difference is 3 memory nodes, its address is respectively：200.1.1.110、 200.1.1.111、200.1.1.112。

Step 402：Send the second read request.

K the second read requests are generated based on the levels of redundancy numbering in K bars forwarding record and memory node address, point K second read request is not sent to K memory node.Such as the embodiment in step 401, because redundancy inquired etc. Level is RAID 1/3, is distinguished corresponding to 3 levels of redundancy numbering RAID1/3_0, RAID 1/3_1, RAID 1/3_2 forward-paths It is 3 memory nodes that address is 200.1.1.110,200.1.1.111,200.1.1.112, so the service switch meeting Build 3 the second read requests, each second read request packet contains the 4th metadata, described in 3 second read requests The 4th source address in 4th metadata is 200.1.1.101, that is, represents second read request by address for 200.1.1.101's Service switch is sent；4th destination address is respectively 200.1.1.110 or 200.1.1.111 or 200.1.1.112, is represented It is 200.1.1.110 or 200.1.1.111 or 200.1.1.112 that 3 second read requests, which need to be sent respectively to address, Memory node；Operational order in 4th metadata is all " reading "；Storage index and institute in 4th metadata The storage index stated in trinary data is identical, is still LUN A+Ox0000FFFF；It is net in 4th metadata Lotus data length is identical with the payload data length in the trinary data, is still 512 bytes.

Further, because second read request is to be sent to memory node by the interchanger, therefore described Also need to include levels of redundancy numbering, with reference to above-mentioned table 4,3 second read requests in 4th metadata of two read requests Levels of redundancy numbering in 4th metadata is RAID 1/3_0, RAID 1/3_1, RAID 1/3_2 respectively.

Optionally, as previously indicated, the 4th metadata of second read request can also include storage class, its value It is identical with the storage class value in the trinary data, it is such as " block "；4th metadata of second read request may be used also With including sequence number, its value is identical with the sequence number value in the trinary data, is such as Ox5678.

The service switch by 3 second read requests be sent respectively to address for 200.1.1.110, 200.1.1.111,3 memory nodes of 200.1.1.112.

Optionally, the service switch carries out inquiry forwarding table after first read request is received and sends institute , can be with creation state table during stating the second read request, the state letter handled for recording needle first read request Breath, increases status information on the basis of above-mentioned table 4, and the mode of operation in the state table is the operation life in trinary data Order, is recorded as " reading " here；Sequence number value in the state table is the sequence number in the trinary data, such as can be with It is Ox5678；Payload length value in the state table is the payload user data length in the trinary data, such as can be with It is 512 bytes.Shown in table 7 specific as follows：

Table 7

Further, herein it should be noted that above-mentioned state table is created as optional step, and the establishment state Table can be after step 401 or be created that after step 402.

Step 403：Memory node sends second and runs through message.

Each memory node in the K memory node receives the second read request from the interchanger, with institute State the storage index in the 4th metadata entrained by the second read request and levels of redundancy numbering is read as key Data.3 memory nodes that address is 200.1.1.110,200.1.1.111,200.1.1.112 receive as described in respectively 1 in 3 second read request of service switch, based on the 4th metadata in second read request from local Read data.

In embodiments of the present invention, as described in Fig. 3 and the description of corresponding embodiment, if the memory node is stored with object Mode preserve data, then in this step, the memory node with the 4th metadata storage index (i.e. volume ID+ Logical address) and levels of redundancy numbering need the data that read as key to index；Or based in the 4th metadata Storage index (i.e. volume ID+logical address) and levels of redundancy numbering progress Hash operation and obtain key, then with resulting The key need the data that read to index.

The K memory node is based respectively on read data generation second and runs through information and sending to the service Interchanger.With reference to the message structure described in Fig. 4, described second, which runs through message, includes metadata and payload user data.Wherein, it is described Second metadata for running through message is specially：

Destination address：The address of the service switch, such as 200.1.1.101；

Storage class：With the value of the storage class in second read request, such as " block "；

Storage index：With the value of the storage index in second read request, such as " LUNA+Ox0000FFFF "；

Operational order：Run through；

Levels of redundancy is numbered：Such as it is respectively " RAID 1/ with the value of the levels of redundancy numbering in second read request 3_0、RAID 1/3_1、RAID 1/3_2”；

Sequence number：With the value of the sequence number in second read request, such as " Ox5678 "；

Payload user data length：With the value of the payload user data length in second read request, such as " 512 byte ".

One of ordinary skill in the art will appreciate that with reference to being previously with regard to the first read request and the second read request Description, described second runs through the sequence number of meta-data section in message, two fields of storage class are also optional.

Second payload user data run through in message is that the memory node is based on second read request from originally The data that ground is read out.

Step 404：IP fragmentation and reassembly.

The service switch is respectively received from described each memory node in the K memory node Second run through message, from it is described each second run through entrained data parsed in message, to the K second The data carried in message are run through to be recombinated.It is to be said so that levels of redundancy is RAID 1/3 as an example in the present embodiment It is bright, thus the service switch by described second from 3 memory nodes run through payload user data in message into After row verification is errorless, the data of 1 512 byte are reassembled as, which is the data that the host needs to read, rear The host is sent the data in afterflow journey.

If one of ordinary skill in the art will appreciate that levels of redundancy be RAID 5/5 in the case of, it is described Service switch can be respectively received second returned from 5 different memory nodes and run through message, then respectively from described next Run through from the second of 5 different memory nodes returns and payload user data is parsed in message, 5 numbers described here parsed 4 data fragmentations and 1 redundant slices are included respectively according to payload, and the service switch is respectively to 5 payload user data Verify it is errorless after 4 data fragmentations and 1 redundant slices are recombinated based on 5 technologies of RAID, described in generation Host needs the data read, and the host is sent the data in follow-up process.It will be appreciated by those skilled in the art that , levels of redundancy can be based on difference RAID such as RAID 0, RAID 1, RAID 3, RAID 5, RAID 10 or RAID 50 Rank and a variety of situations that corresponding RAID level can be realized by the memory node of varying number, then different superfluous In the case of remaining grade, step 604 should based on corresponding RAID level and form the memory node number of the RAID level come The payload user data run through to second from different memory nodes in message carries out IP fragmentation and reassembly, which belongs to RAID The basic principle of technology, therefore the embodiment of the present invention is not described in detail for different levels of redundancy one by one.Below or with Levels of redundancy illustrates exemplified by being RAID 1/3.

Step 405：First, which is sent, to the host runs through message.

The service switch runs through message based on the data generation first after the IP fragmentation and reassembly.With reference to described in Fig. 4 Message structure, described first, which runs through message, includes metadata and payload user data.Wherein, described first run through in message Payload user data is the service switch based on the data after the IP fragmentation and reassembly；Described first runs through the metadata of message Specially：

Destination address：The address of the host, such as 200.1.1.3；

Storage index：With the value of the storage index in second read request, such as " LUN A+Ox0000FFFF "；

Operational order：Run through；

As it was previously stated, in the service switch with the message between the host, it is not necessary to carry levels of redundancy numbering Field.One of ordinary skill in the art will appreciate that with reference to being previously with regard to retouching for the first read request and the second read request State, described first runs through the sequence number of meta-data section in message, two fields of storage class are also optional.

Further, if the service switch has after first read request is received creates state table table 7, In the present embodiment, the service switch often receives 1 and second runs through message from the memory node, just by the shape Mode of operation in state table (such as table 7) is revised as " being read as work(", specific as shown in table 8：

Table 8

Further, message is run through to trigger the state table when receiving described second from 3 memory nodes Mode of operation be all changed to " being read as work(", then illustrate that first read request is handled successfully, then can to the host send institute State first and run through message, and the state table can be deleted.

In the distributed memory system provided in an embodiment of the present invention, when the service configured for the host is handed over Change planes after receiving first read request from the host, the forwarding table can be based on and directly read to ask by described first Ask and split into multiple second read requests, be then sent respectively to corresponding memory node and read data, receive come from respectively , will be the multiple based on RAID technique after multiple the second of the multiple memory node return run through the data carried in message Second, which runs through the data carried in message, is recombinated and runs through message by described first again and return to the host, by This realizes quick read request processing, and the data that host described here needs to read are passing through the clothes from multiple memory nodes Business interchanger avoids that is needed as the prior art to only needing on the direction of the host by the service switch 1 time Digital independent could be completed by transmitting 2 times, thereby reduce the waste of the Internet resources of interchanger, decrease read request processing Time delay, greatly improve the treatment effeciency and performance of the distributed memory system.

Based on same or like inventive concept is described to such as Fig. 1,2,3,4 and above-mentioned corresponding embodiment, such as Fig. 5 institutes Show, be a kind of structure diagram of interchanger provided in an embodiment of the present invention, the interchanger 500 includes receiver 501, processing Device 502, memory 503, RAID engine 504 and transmitter 505.

In the embodiment of the present invention, the receiver 501 is used to receive the interchanger ability information inquiry from control node Message, the processor 502 are used to parse the interchanger ability information query message, are looked into based on the interchanger ability information The ability information that message inquires about the interchanger is ask, as whether interchanger ID, switch address, interchanger possess burst and redundancy Computing capability etc., based on aforementioned capabilities information generate interchanger ability information inquiry response message, the transmitter 505 be used for The control node sends the interchanger ability information inquiry response message.Further, the receiver 501 is additionally operable to connect The forwarding table from the control node is received, the processor 502 is used to transfer to the memory 503 after parsing the forwarding table Preserve the forwarding table.

Further, the receiver 501 is additionally operable to receive the first write request from host, in first write request Carry the first metadata and the first data；The processor 502, is additionally operable to parse first write request to obtain described One metadata and first data, are recorded, K based on forwarding table described in first metadata query with obtaining the forwarding of K bars More than or equal to 2；The RAID engine 504, for for based on the levels of redundancy numbering in described K articles forwarding record to described the One data carry out Fragmentation, obtain K burst；The processor 502, is additionally operable to based on superfluous in K bars forwarding record K the second write requests of remaining Grade numbers and memory node address generation, second write please for each in the K the second write requests Ask and carry the second data, second data are 1 burst in the K burst；The transmitter 505, is additionally operable to be used for Respectively K second write request is sent to K memory node.The receiver 501, is additionally operable to receive respectively from the K Second of each memory node in a memory node writes success message；The processor 502, is additionally operable to come from based on described Second of each memory node in the K memory node writes success message generation first and writes success message；The transmitting Device 505, is additionally operable to write success message to host transmission described first.Entrained by the forwarding table and the first write request Referring to Fig. 1,2,3 and accordingly, the description of embodiment, present apparatus embodiment repeat no more the composition of one metadata；Equally, the place Manage how device 502 is based on the first metadata query forwarding table, Fragmentation, generates K the second write requests and based on institute State second of each memory node in the K memory node and write the reality that success message is write in success message generation first The description that details also refers to Fig. 1,2,3 and corresponding embodiment is applied, present apparatus embodiment repeats no more；Further, with reference to Fig. 3 And the description of corresponding embodiment, in the processing procedure of write request, the processor 502 can be also used for creation state table, and And after receive each memory node in the multiple memory node second writes success message described in renewal State table and the deletion state table.

The interchanger provided in an embodiment of the present invention, after first write request from the host is received, Burst and computing redundancy burst directly can be carried out to the data carried in first write request based on the forwarding table, then Second write request of corresponding number is generated, second write request is directly then sent respectively to storage section accordingly Point, then the memory node preserve the data carried in the second write request that it is received respectively, can thus complete institute State the preservation of the data carried in the first write request.In embodiments of the present invention, the data in first write request only need 1 time is transmitted on the service switch just can complete preservation, and avoiding as the prior art that needs to transmit 2 times and could protect Deposit, thereby reduce the waste of the Internet resources of interchanger, decrease the time delay of a write request processing, greatly improve institute State the treatment effeciency and performance of distributed memory system.

Further, the receiver 501, is additionally operable to receive the first read request from host, first read request Middle carrying trinary data；The processor 502, is additionally operable to parse first read request to obtain the trinary data, The forwarding table is inquired about based on the trinary data to obtain K articles of forwarding record, K is more than or equal to 2, is forwarded based on the K bars K the second read requests of levels of redundancy numbering and the generation of memory node address in record；The transmitter 505, is additionally operable to distinguish K second read request is sent to K memory node.The receiver 501, is additionally operable to receive respectively and is deposited from the K Store up each memory node in node second runs through message；The processor 502, it is described from described to be additionally operable to parsing Second of each memory node in K memory node runs through message and described second runs through message institute to obtain each The data of carrying；The RAID engine 504, is additionally operable to run through the data carried in message to the K second and carries out weight Group；；The processor 502, is additionally operable to generation first and runs through message, described first runs through the number that message carries the restructuring According to；The transmitter 705, message is run through for sending first to the host.Modules/component is every in the interchanger The implementation detail of one action refers to Fig. 4 and the description of corresponding embodiment, and details are not described herein for the present embodiment.

, can be with after the interchanger provided in an embodiment of the present invention receives first read request from the host First read request is directly split into by multiple second read requests based on the forwarding table, is then sent respectively to corresponding Memory node read data, receive multiple second returned from the multiple memory node respectively and run through in message and carry Data after, the multiple second is run through based on RAID technique the data carried in message and recombinate and again by described First, which runs through message, returns to the host, it is thus achieved that quickly read request processing, host described here need to read Data from multiple memory nodes by the service switch to only needing by the service on the direction of the host Interchanger 1 time, avoiding as the prior art that needs transmission to complete digital independent 2 times, thereby reduces interchanger The waste of Internet resources, decreases the time delay of read request processing, greatly improves the processing effect of the distributed memory system Rate and performance.

Based on same or like inventive concept is described to such as Fig. 1 and above-mentioned corresponding embodiment, as shown in fig. 6, being this A kind of structure diagram for control node that inventive embodiments provide, the control node 800 include receiver 601, processor 602nd, memory 603, transmitter 604.The transmitter 604, for sending interchanger ability information query message to interchanger, Memory node ability information query message is sent to memory node；The receiver 601, for receiving from the interchanger Interchanger ability information inquiry response message, the interchanger ability information inquiry response message carry interchanger ability information, The ability can include following information：Whether interchanger ID, switch address, interchanger possess burst and redundant computation energy Power；It is additionally operable to receive the memory node ability information inquiry response message from the memory node, the memory node ability Information inquiry response message carries memory node ability information, and the ability can include following information：Include following information：Deposit Store up node address, memory node capacity.The processor 602, for parsing the interchanger ability information inquiry response message To obtain the interchanger ability information, it is described to obtain to be additionally operable to parse the memory node ability information inquiry response message Memory node ability information；The memory 603, for preserving the interchanger ability information and memory node ability letter Breath.The receiver 601, is additionally operable to receiving host storage demand；The processor 602, is additionally operable to store based on the host Demand and the interchanger ability information and the memory node ability information are the host assignment service switch, distribution Multiple memory nodes and configuration forwarding table.The transmitter 604, is additionally operable to the address of the service switch being sent to institute Host is stated, the forwarding table is sent to the service switch.The interchanger ability information, memory node ability information Presentation mode, preserving type such as Fig. 1 and the description of corresponding embodiment；The processor 602 is servicing exchange for the host assignment The implementation detail of machine, the multiple memory nodes of distribution and configuration forwarding table refer to Fig. 3 such as and the description of corresponding embodiment, the present invention Details are not described herein for embodiment.

It should be noted that " first ", " second ", " the 3rd ", " the 4th " mentioned in all embodiments of the present invention Used each other with describing convenient Deng just to distinguish, it is impossible to which these sequence numbers are read as the entity scope to being connected below Restriction.

Those of ordinary skill in the art may realize that each exemplary list described with reference to the embodiments described herein Member and algorithm steps, can be realized with electronic hardware, computer software or the combination of the two, in order to clearly demonstrate hardware With the interchangeability of software, each exemplary composition and step are generally described according to function in the above description.This A little functions are performed with hardware or software mode actually, application-specific and design constraint depending on technical solution.Specially Industry technical staff can realize described function to each specific application using distinct methods, but this realization is not It is considered as beyond the scope of this invention.

It is apparent to those skilled in the art that for convenience of description and succinctly, foregoing description is The specific work process of system, device and unit, may be referred to the corresponding process in preceding method embodiment, details are not described herein.

In several embodiments provided herein, it should be understood that disclosed systems, devices and methods, can be with Realize by another way.For example, device embodiment described above is only schematical, for example, the unit Division, is only a kind of division of logic function, can there is other dividing mode, such as multiple units or component when actually realizing Another system can be combined or be desirably integrated into, or some features can be ignored, or do not perform.In addition, shown or beg for The mutual coupling, direct-coupling or communication connection of opinion can be the INDIRECT COUPLING by some interfaces, device or unit Or communication connection or electricity, the connection of mechanical or other forms.

The unit illustrated as separating component may or may not be physically separate, be shown as unit The component shown may or may not be physical location, you can with positioned at a place, or can also be distributed to multiple In network unit.Some or all of unit therein can be selected to realize the embodiment of the present invention according to the actual needs Purpose.

In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, can also It is that unit is individually physically present or two or more units integrate in a unit.It is above-mentioned integrated Unit can both be realized in the form of hardware, can also be realized in the form of SFU software functional unit.

If the integrated unit is realized in the form of SFU software functional unit and is used as independent production marketing or use When, it can be stored in a computer read/write memory medium.Based on such understanding, technical scheme is substantially The part to contribute in other words to the prior art, or all or part of the technical solution can be in the form of software product Embody, which is stored in a storage medium, including some instructions are used so that a computer Equipment (can be personal computer, server, or network equipment etc.) performs the complete of each embodiment the method for the present invention Portion or part steps.And foregoing storage medium includes：USB flash disk, mobile hard disk, read-only storage (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disc or CD etc. are various can store journey The medium of sequence code.

The above description is merely a specific embodiment, but protection scope of the present invention is not limited thereto, any Those familiar with the art the invention discloses technical scope in, various equivalent modifications can be readily occurred in or replaced Change, these modifications or substitutions should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with right It is required that protection domain subject to.

Claims

A kind of 1. method of processing write requests, it is characterised in that the method is applied to interchanger, matches somebody with somebody in advance in the interchanger Forwarding table is equipped with, the forwarding table is recorded comprising a plurality of forwarding, and each forwarding record is with including host address, service switch Location, volume, levels of redundancy numbering and memory node address, the described method includes：

The first write request from host is received, the first metadata and the first data are carried in first write request；

Based on forwarding table described in first metadata query to obtain K bars forwarding record, K is more than or equal to 2；

Fragmentation is carried out to first data based on the levels of redundancy numbering in K bars forwarding record, obtains K points Piece；

Based on K the second write requests of the levels of redundancy numbering in K bars forwarding record and the generation of memory node address, the K Each second write request in a second write request carries the second data, and second data are 1 in the K burst Burst；

Respectively K second write request is sent to K memory node.
2. according to the method described in claim 1, it is characterized in that, first metadata includes the first source address, the first mesh Address, storage index and payload user data length, wherein first source address be the host address, first purpose Address is the address of the interchanger, and the payload user data length is used for the size for indicating first data；

It is then described to be specially based on forwarding table described in first metadata query：

Each described in the forwarding table is matched respectively based on first source address, the first destination address and storage index to turn Host address, service switch address and the volume for sending out record are recorded with obtaining the K bars forwarding.
3. according to the method described in claim 1, it is characterized in that, each in the K the second write requests second write please Ask and also carry the second metadata, second metadata includes the second source address, the second destination address, storage index, redundancy etc. Level numbering and payload user data length, wherein second source address is the address of the interchanger, second destination address is The memory node address of 1 forwarding record in the K bars forwarding record, the levels of redundancy numbering are the K bars forwarding note The levels of redundancy numbering of 1 forwarding record in record, the payload user data length are used for the size for indicating 1 burst.
4. according to the method described in claim 3, it is characterized in that, the method further includes：

The K memory node receives the second write request from the interchanger respectively；

With the storage index in the second metadata entrained by second write request received and described redundancy etc. Level numbering stores the second data carried in second write request received as key；

Second, which is sent, to the interchanger writes success message.
5. according to the method described in claim 4, it is characterized in that, the method further includes：

What the interchanger received from the K memory node respectively second writes success message；

First, which is sent, to the host writes success message.
A kind of 6. method for handling read request, it is characterised in that the method is applied to interchanger, matches somebody with somebody in advance in the interchanger Forwarding table is equipped with, the forwarding table is recorded comprising a plurality of forwarding, and each forwarding record is with including host address, service switch Location, volume, levels of redundancy numbering and memory node address, the described method includes：

The first read request from host is received, trinary data is carried in first read request；

The forwarding table is inquired about based on the trinary data to obtain K articles of forwarding record, K is more than or equal to 2；

Based on K the second read requests of the levels of redundancy numbering in K bars forwarding record and the generation of memory node address；

Respectively K second read request is sent to K memory node.
7. according to the method described in claim 6, it is characterized in that, the trinary data includes the 3rd source address, the 3rd mesh Address, storage index and payload user data length, wherein the 3rd source address be the host address, the 3rd purpose Address is the address of the interchanger, and the payload user data length is used for the size for indicating the data that needs are read；

It is then described to be specially based on the trinary data inquiry forwarding table：

Based on the 3rd source address in the trinary data, the 3rd destination address and the storage index difference Host address, service switch address and volume with the forwarding record of each described in the forwarding table are turned with obtaining the K bars Hair record.
8. the method according to the description of claim 7 is characterized in that each second reading in the K the second read requests please Also the 4th metadata of carrying is sought, the quaternary data include the 4th source address, the 4th destination address, storage index, redundancy etc. Level numbering and payload user data length, wherein the 4th source address is the address of the interchanger, the 4th destination address is The memory node address of 1 forwarding record in the K bars forwarding record, the levels of redundancy numbering are the K bars forwarding note The levels of redundancy numbering of 1 forwarding record in record, the size for the data that the payload user data length needs to read for instruction, The method further includes：

Each memory node in the K memory node receives the second read request from the interchanger；

Made with the storage index in the 4th metadata entrained by second read request and levels of redundancy numbering Data are read for key；

Second is sent to the interchanger and runs through message, and described second runs through the data that message carries the reading.
9. according to the method described in claim 8, it is characterized in that, the method further includes：

What the interchanger received each memory node in the K memory node respectively second runs through message；

The data carried in message are run through to the K second to recombinate；

First is sent to the host and runs through message, and described first runs through the data after the restructuring is carried in message.
10. a kind of interchanger, it is characterised in that the interchanger includes receiver, memory, processor, transmitter and RAID Engine, wherein：

The receiver, carried for receiving the first write request from host, in first write request the first metadata with And first data；

The memory, for forwarding table memory, the forwarding table includes a plurality of forwarding and records, and each forwarding record includes master Machine address, service switch address, volume, levels of redundancy numbering and memory node address；

The processor, for parsing first write request to obtain first metadata and first data, base In forwarding table described in first metadata query to obtain K bars forwarding record, K is more than or equal to 2；

The RAID engine, for being divided based on the levels of redundancy numbering in K bars forwarding record first data Piece operates, and obtains K burst；

The processor, is additionally operable to based on the levels of redundancy numbering in K bars forwarding record and memory node address generation K Second write request, each second write request in the K the second write requests carry the second data, and second data are institute State 1 burst in K burst；

The transmitter, for sending K second write request to K memory node respectively.
11. interchanger as claimed in claim 10, it is characterised in that first metadata includes the first source address, first Destination address, storage index and payload user data length, wherein first source address is the address of the host, first mesh Address be the interchanger address, the payload user data length is used to indicate the sizes of first data；

Then the processor is used for described specific to obtain K bars forwarding record based on forwarding table described in first metadata query For：

Each described in the forwarding table is matched respectively based on first source address, the first destination address and storage index to turn Host address, service switch address and the volume for sending out record are recorded with obtaining the K bars forwarding.
12. interchanger as claimed in claim 10, it is characterised in that the receiver, is additionally operable to receive respectively from the K Second of each memory node in a memory node writes success message；

The processor, is additionally operable to second based on each memory node in the K memory node and is write as Success message is write in work(message generation first；

The transmitter, is additionally operable to write success message to host transmission described first.
A kind of 13. interchanger, it is characterised in that the interchanger includes receiver, memory, processor and transmitter, wherein：

The receiver, for receiving the first read request from host, trinary data is carried in first read request；

The memory, for forwarding table memory, the forwarding table includes a plurality of forwarding and records, and each forwarding record includes master Machine address, service switch address, volume, levels of redundancy numbering and memory node address；

The processor, for parsing first read request to obtain the trinary data, based on the trinary data The forwarding table is inquired about to obtain K bars forwarding record, K is more than or equal to 2, is compiled based on the levels of redundancy in K bars forwarding record Number and memory node address generation K the second read requests；

The transmitter, for sending K second read request to K memory node respectively.
14. interchanger as claimed in claim 13, it is characterised in that the interchanger further includes RAID engine；

The receiver, the second of each memory node for being additionally operable to receive respectively in the K memory node runs through Into message；

The processor, second for parsing each memory node in the K memory node runs through Message described second runs through data entrained by message to obtain each；

The RAID engine, recombinates for running through the data carried in message to the K second；

The processor, is additionally operable to generation first and runs through message, described first runs through the data that message carries the restructuring；

The transmitter, message is run through for sending first to the host.
A kind of 15. control node, it is characterised in that the control node includes receiver, processor, memory, transmitter, its In：

The transmitter, for sending interchanger ability information query message to interchanger, memory node is sent to memory node Ability information query message；

The receiver, for receiving the interchanger ability information inquiry response message from the interchanger, the interchanger Ability information inquiry response message carries interchanger ability information, receives the memory node ability information from the memory node Inquiry response message, the memory node ability information inquiry response message carry memory node ability information；

The processor, is additionally operable to parse the interchanger ability information inquiry response message to obtain the interchanger ability letter Breath, parses the memory node ability information inquiry response message to obtain the memory node ability information；

The memory, for preserving the interchanger ability information and the memory node ability information；

The receiver, is additionally operable to receiving host storage demand；

The processor, is additionally operable to be based on the host storage demand and the interchanger ability information and the memory node Ability information is the host assignment service switch, distributes multiple memory nodes and configure forwarding table, the forwarding table bag Containing it is a plurality of forwarding record, each forwarding record comprising host address, service switch address, volume, levels of redundancy numbering and Memory node address；

The transmitter, is additionally operable to the address of the service switch being sent to the host, the forwarding table is sent to The service switch.