CN102594888B - Method for enhancing real-time performance of network file system - Google Patents
Method for enhancing real-time performance of network file system Download PDFInfo
- Publication number
- CN102594888B CN102594888B CN201210034549.3A CN201210034549A CN102594888B CN 102594888 B CN102594888 B CN 102594888B CN 201210034549 A CN201210034549 A CN 201210034549A CN 102594888 B CN102594888 B CN 102594888B
- Authority
- CN
- China
- Prior art keywords
- request
- real
- time
- client
- data flow
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a method for enhancing the real-time performance of a network file system. The conventional network file system architecture is improved, a client and a server of the network file system are provided with a real-time parameter convention, traffic control and request scheduling mechanism, and real-time file access control capability is provided under a request/response architecture and a remote procedure call (RPC) thread pool mechanism of the conventional network file system. By the method for enhancing the real-time performance of the network file system, the shortcoming about the support of the conventional network file system to the real-time performance can be overcome.
Description
Technical field
The present invention relates to a kind of method that strengthens NFS real-time.
Background technology
NFS (NFS, Network File System) is developed by Sun Microsystems the earliest, for providing under Unix environment the remote access ability of local file system.Because this technology has obtained being widely used, from NFS v.2 after, NFS v.3, NFS v.4 with NFS v.4.1 version all become the agreement (RFC, Request for Comments) that RFC standard specifies, extensively adopted by the operating system such as Linux, Unix.The realization of Follow NFS agreement is called NFS.
NFS agreement is applicable to, in network remote file access application, belong to the presentation layer in OSI network seven layer architectures, can be client the encapsulation to rhosts system access is provided.Wherein, NFS client, for upper layer application provides remote document access interface, provides the file operations such as open, read, write, and these operations are converted to udp data bag send to server end.Server end is monitored after network request analysis that client is sent, calls the bottom document system interface of nfs server end system and carries out actual file access, and return to NFS client after access result is packed by udp protocol.NFS client is receiving after the result of returning, and result is submitted to the upper layer procedures of sending remote document access request.
Accompanying drawing 1 has provided the general principle of existing network file system.In the request/response framework model of the existing NFS shown in figure, upper layer application process is sent file access request to client service task, and client sends to server end by request with the form of UDP message, and wait returns results.Server end is monitored the corresponding network port, and with the order of first serving first, request is distributed to the thread (processing successively by request arrival order) in RPC thread pool.Thread read requests in thread pool is also decoded, and carries out after the operation of corresponding document system, returns to client after result is packaged as to UDP message.Client returns to application process after receiving request results, finishes this request.
From above discussion, be not difficult to find out, existing NFS lacks necessary support to real-time.Be in particular in the following aspects;
The first, existing NFS agreement is not supported the quantitative description to real-time characteristic or QoS, cannot provide necessary time-constrain information and quality of service guarantee for the file access service of different application.
Second, client does not provide flow control mechanism, the available bandwidth of uncontrollable each application, can not solve the mutual isolation of file access data flow separately between different application, cannot ensure that normal real-time stream can not be subject to the impact of the abnormal data stream that fault end node and network node produce.
The 3rd, server end not only lacks the support of real-time characterising parameter, also cannot directly in RPC (remote procedure call, Remote Procedure Call) and thread pool mechanism, directly realize Real-Time Scheduling.
On the other hand, along with continuous increase and the physical system/information system of memory capacity merge improving constantly of degree, in Embedded Application and Stream Media Application, the storage of increasing video, view data and other sensing datas and access have all proposed to explicitly call for to the real-time of NFS.
The NFS document relevant to patent application is as follows: domestic invention " towards the NFS flow control method of udp protocol " (patent application numbers 200910312057.4, application publication number CN 101741747A) in proposed to use the method that threshold value and packet loss retransmit of giving out a contract for a project to adjust request transmission rate, prevent from causing network congestion or server overload.The client Flow Control that the present patent application proposes is the quantitative Flow Control for data flow, in method, has significantly different from foregoing invention.In addition, the present invention also transforms at server end, has added request scheduling device at nfs server end, and the request of the different data streams that arrives nfs server is carried out to Real-Time Scheduling.Foregoing invention does not adopt similar techniques.Aspect purposes, the present invention to be to ensure that each data flow real-time is as target, and can be used for avoiding affecting other normal use real-time files access because of the excess data flow that node or its upper layer application fault cause.Therefore, this invention is not identical from the present invention different.In addition the paper " CA-NFS:A Congestion-Aware Network File System " that, Alexandros Batsakis of Johns Hopkins university in 2009 etc. delivers in FAST international conference has also been introduced the congestion avoidance mechanism in a kind of NFS.But this article is (online competitive tender method for the exchanging load information between distributed NFS client, nfs server) of realizing QoS by the Buffer management policy of adjustment NFS client, server and online competitive tender method.Technology that this article is carried can not provide quantitative QoS or real-time ability of regulation and control to NFS, therefore the memory access can only be in macroscopic view requiring for different business provides resource guarantee, does not solve several subject matters that the real time implementation that proposes faces herein above.Therefore the method also cannot meet the accurate control requirement of real-time to performance, resource parameters.
Be not difficult to find out according to above-mentioned discussion, existing research and invention all do not propose to realize the effective way of enhancing real-time performance of network file system.The present invention is intended to propose a kind of system architecture addressing this problem, by client, server end and the interaction mechanism thereof of transformation NFS, under NFS, provide soft real-time or hard real-time support (support soft in real time or when strong, the operating system of moving with system be soft when real-time or strong and bottom document system whether support that the storage medium access control when strong is relevant).And dispatch and strengthen the ability of regulation and control of local file system being accessed to real-time by I/O.
Summary of the invention
The deficiency of real-time being supported in order to overcome existing network file system, the invention provides a kind of method of transforming NFS framework, by client traffic control, server end request scheduling/distribution, ensures network file access real-time.Specifically reach above-mentioned target by dispose corresponding real-time control device at client, server end.The present invention has designed a kind of real-time reconstruction structure that is applicable to NFS structure, specifically comprises the following steps.
Step 1: arrange real-time parameter between NFS client and nfs server, for the real-time traffic control on NFS client and nfs server and Real-Time Scheduling device provide necessary parameter.The present invention requires real-time constraint parameter at least to comprise following content:
The first, data flow identifiers.System designer or keeper are that each real-time stream is specified a unique data flow identifiers.This method regulation, data flow identifiers both can have been distributed to each process, thread or task; Also can distribute to one group of continuous file access request sequence into each process, thread, task creation; Also can be one or several continuous read-write operation sequence and point to the data flow specified identifier of same file, file set, catalogue or volume (logical volume, physical volume).
The second, data stream property parameter set.Specifically comprise data flow bandwidth requirement (unit is bytes per second or I/O number of request per second), cycle or largest interval, buffer size, maximum delay or off period, transmission direction (unidirectional or two-way), request execution sequence (keep or allow out of order), whether allow to return results in batches, flow control method, dispatching method etc.These parameters provide design parameter index, the real-time control that system can be quantized for each data flow for Flow Control and scheduling.
Pay special attention to, for transmission and storage efficiency are further provided, a data flow of this method regulation can be sent one or more read/write requests within the single cycle, but all requests in one-period must all complete within the off period in this cycle.Typical example is that a two field picture is write to the request sequence that storage system is sent while comprising image transmitting.Therefore, the cycle of setting or largest interval can comprise multiple data access request.
Step 2: determine under current configuration whether to meet the real-time of all data flow.If real-time can meet, perform step the real time access operation after 3.Otherwise the software and hardware scheme of redesign or amendment system.This step can adopt the mode of static off-line analysis or dynamic accommodation control meter side to complete.This method does not limit concrete analysis or the side's of counting method.
Step 3: the upper layer application of client is initiated network file reading and writing request.Require, in the client of NFS, volume control device is set, each remote document access request of sending is carried out to flow restriction (as accompanying drawing 2).The main purpose of this step has two, and one is that protection and isolation are provided between client different data streams; Another is to prevent from producing a large amount of abnormal data stream when upper layer application from breaking down, and affects the real-time of other normal flows.
In this step, the parameter such as cycle, bandwidth that client specifies according to real-time parameter, is the pre-side of utilizable flow in each stream setting cycle.In the time that system just starts to carry out, set the pre-side Li of spendable initial flow (i is data flow identifiers) and be:
The bandwidth of Li (0)=(allow the maximum number of byte sending in interval or ask number)/data flow i
In the time not receiving file read-write request, if current time t is apart from being greater than Li (0) remaining time of this measurement period T finish time, i.e. (T-t) >=Li (0), keeps Li constant.If (T-t) < Li (0), even without receiving read-write requests, the also side of counting Li=T-t again.When entering next cycle while starting, Li is initialized as to Li (0).
In the time receiving that length is the file read-write request of len, if Li-len >=0 allows to send request, and Li=Li-len is set.If condition does not meet, this request is waited in buffering area, until meet in certain follow-up moment Li-len >=0 (pre-side is supplemented).
For the operation of performing step 3, must transform NFS client.Concrete mode has two kinds (two select one).
(A) the code porch that receives upper access request in NFS client code increases flow tagging mechanism, and the described data flow of each request is carried out to mark.The pre-assigned bandwidth of scheduler or flow controller and selected dispatching method are ranked and flow control to file access request." NFS file request data flow token and Flow Control " in accompanying drawing 2 has been responsible for data flow token and Flow Control function, do not use " for the Flow Control of Internet Transmission " in accompanying drawing 2.
(B) in the network layer of NFS client, increase control of network flow quantity and scheduling mechanism.In the time that sending real network request, NFS client classifies according to data flow label, by more network scheduling, the flow-control mechanism of bottom are processed according to label." NFS file request data flow token and Flow Control " in accompanying drawing 2 only completes data flow token function, do not carry out Flow Control.Actual flow is processed according to the mark in data flow by " for the Flow Control of Internet Transmission ".
Concrete flow control can adopt leaky bucket method, token bucket submethod or reaction type flow control method etc., and dispatching method can adopt time poll, EDF etc.This method is not done requirement to that method of concrete selection that may take, the method that user can be appropriate according to actual selection.
Step 4: the server end of existing network file system receives the request from client, puts into network packet queue.Current execution thread in RPC thread pool arrives sequencing by request, and the request that taking-up arrives at first from queue returns to client by result after processing.
Transform this step as multiple sub-steps.Its general principle is by increasing independently request scheduling device, controls processing sequence and the processing speed (according to the parameter such as priority, cycle) of network file access request, thereby realizes the control to Data Stream Processing bandwidth, speed, delay etc.Except the regulation and control of request order, request scheduling device can also be decomposed into large data access request multiple little requests, and in the time returning, multiple little request orders is returned to (need to increase flow control and scheduling).
Sub-step comprises the following steps:
Step 4.1: network packet receipts/hair device in accompanying drawing 3 (this method require increase), according to stream identification to future automatic network remote document access requests classification deposit corresponding request queue in.
Step 4.2: the NFS request scheduling device in accompanying drawing 3 is according to the current practice condition of the parameters such as available thread and priority thereof in Request Priority, cycle, off period, maximum delay, thread pool and all data flow requests, request in request queue is assigned to concrete thread by the reef knot fruit of real-time scheduling method and processes (and the thread of giving in RPC thread pool process, i.e. task T1～Tn in accompanying drawing 3).Scheduler carries out real-time files access scheduling according to the flow that completes in each data flow current period with flow to be completed.The request packet sequence that is specially each data flow distributes execution thread and the method on opportunity, can adoption rate method of rate allocation, EDF method etc. realizes.
Pay special attention to, this step master is if it were not for utilizing the different priorities of thread pool thread to realize real-time, but utilize can speed control queue scheduling reach this target.But this method allows to increase a small amount of high priority thread, to meet the preferential special circumstances of carrying out of accidental request; Or distribute to part thread and carry out faster frequency.This method does not limit the concrete establishing method mode of thread priority.
Step 4.3: the each request processing thread (the task T1～Tn in accompanying drawing 3) in NFS RPC thread pool reads next NFS access request to be processed from queue, and these requests are decomposed into a series of asynchronous file read-write request that can have by local file system processing.Thereafter the RPC thread that receives request can start to process next request.
Step 4.4: local file system returns to result after the RPC thread of the request of sending after handling small documents request, and result data is returned to client by the network packet receipts/hair device in rear accompanying drawing 3.
Note, if step 1 carry out real-time parameter approximately timing allow to return results in batches, RPC thread returns results little request to give successively network packet receipts/hair device, then returns to the client of the request of sending by it.Client also returns to upper layer application by result successively.After all returning results and receiving, client finishes this remote document access request.Otherwise RPC thread is only being received disposable being returned after all results.And client is also only receiving that after whole request results, final result is returned to upper layer application by.
Can be by reference to the accompanying drawings step 4.3 and 4.4 be described in 3: if current pending task is T1, T1 obtains rqi and processes.T1 according to rqi produce a series of bottom document access request frq1, frq2 ...Local file system, I/O scheduler and memory device return corresponding result reti1, reti2 ... give T1.Result is returned to network packet transmitter by NFS scheduler, and then result is packaged as ack1～acku and returns to NFS client and pass to the upper strata task of sending rqi.While returning to ack1～acku in said process, can dispatch and flow control according to timing, the real-time parameter etc. of data flow corresponding to rqi.If the real-time parameter that step 1 is arranged does not allow to return in batches, be packaged in together by T1 corresponding to all fret1～fretu of request rqi, return to client with the form of single result bag.Now T1 only need return to ack1.
Step 5: client, according to real-time parameter configuring condition, receives returning results of each data flow by set rate from network.After whole results of a request are all returned, client is deleted this request from pending queue, and finishes to this request of upper strata application report.Whole result return course can be according to user's configuration requirement, to submit to successively or the disposable mode of returning of packing is given upper layer application.If this packet is last packet in this cycle, this data flow is set to finish (stop in this cycle the scheduling to this data flow and the meter side of residual flow).
The invention has the beneficial effects as follows: transformation existing network file system framework, by being that the client and server end of NFS increases real-time parameter agreement, flow control and request scheduling mechanism, under existing network file system requests/response architecture and RPC thread pool mechanism, provide real-time files access control ability.
Brief description of the drawings
Fig. 1 is the request/response framework model schematic diagram of existing NFS.
Fig. 2 is that the real-time network file system client system that the present invention proposes forms schematic diagram.
Fig. 3 is that the real-time network file system server end system that the present invention proposes forms schematic diagram.
Embodiment
Below in conjunction with the accompanying drawing in the invention process case, technical scheme of the present invention is clearly and completely described.
Specific embodiments as shown in Figure 3.First system is formed with expression mode and carries out brief description (with the quantity of lower node, task, volume etc. only for convenience of discussing and supposing, the present invention is not to specifying that concrete quantity carries out)
System comprises three client node C1, C2, C3.On each node, there are T1, T2, tri-tasks of T3.Taking the T3 task on C1 node as example, can be expressed as C1.T3.
System comprises a nfs server S (note, the present invention is equally applicable to the situation of multiple nfs servers).On this server carry V1, V2, tri-volumes of V3.On each volume, there are Dir1, Dir2, tri-catalogues of Dir3.Under each catalogue, there are F1, F2, tri-files of F3.S.V1.Dir1 represents the catalogue Dir1 in the V1 volume on S, and S.V2.Dir3.F3 represents the file F3 under catalogue Dir3 in the V1 volume on S.
File access request is sent by the task in client, sends to server end through client.Once concrete required list is shown rq1=(stream1, write, length, buffer), and its implication is the write request that data flow stream1 sends, and it is the data that length, content are buffer that this request writes length to server.
Step 1: definition data flow and real-time constraint thereof.
For simplicity, present case only specifies 2 kinds of data flow.
The data flow of [data stream type 1] oriented mission and node
Here suppose that each task only has a data flow.Each data flow can be to read or write, but difference is read and write simultaneously.According to this supposition, each data flow is unique corresponding to a task in a client.Therefore in discussion below, represent the data flow identifiers of the upper Tj data flow that task is sent of Ci with Ci.Tj.
Also can be for certain end node definition data flow identifiers, for example Ck dispatches as an overall data stream and resource is distributed.
The data flow of [data stream type 2] modular object storage
S.V1 is defined as a data flow identifiers, can carry out scheduling and controlling to the data access request of the V1 volume of all access services device S.S.V2.Dir3 can be used for limiting the access situation of Dir3 catalogue on V2 volume.S.V3.Dir1.F2 represents the data flow identifiers of all access request sequence definitions to the F2 in volume V3, Dir1 catalogue.
Provide the real-time constraint specification of data flow in the present embodiment below.For the sake of simplicity, be only each stream data definition cycle, largest request length, maximum delay, four parameters of priority here.These four parameters are expressed as: Period, MaxLen, MaxDelay and Priority.
As the cycle of data flow C1.T1 can be expressed as C1.T1.Period, the maximum Bao Changwei C2.MaxLen of data flow C2, the maximum delay of data flow S.V1 is S.V1.MaxDelay.
Step 2: the present invention does not specify concrete technology, user can use any technology to verify.Therefore supposition can requirement of real time here.
Step 3: client is initiated the process of request.
Sending write request taking C1.T1 is described as follows as example:.
Step 3.1, C1.T1 sends write request rq1=(C1.T1, write, length, buffer).
Step 3.2, the NFS client on C1 is received this request and is put into client queue.Client according to the real-time parameter meter side of C1.T1 stream is current can transmission rate, and (present case supposition adopts leaky bucket method according to leaky bucket method.The present invention does not limit adoptable concrete grammar.But any NFS client of transforming in a similar manner, to make it have the technical scheme of data stream being carried out to the ability of flow control and scheduling, all belongs to the scope of the claims in the present invention.)。After being sent or be decomposed into multiple son requests, sends by whole rq1 bag.Concrete meter side mode is as false code below:
Even if this step guarantees that by leaky bucket method each data flow produces the excess data request that exceedes the requirement of real-time constraint specification, can not affect the real-time of other data flow on this node or other node yet.
Step 3.3, rq1 deposits in the request queue to be completed of NFS client, until receive returning results from server end.If request queue to be completed is full, in present case, abandon the request of follow-up arrival.
Step 4: server end receives and process rq1.
Step 4.1, the network packet receipts/hair device in server S receives packet from network, and the packet of receiving is put into the afterbody of the corresponding queue of this data flow according to data flow identifiers corresponding to packet.
Step 4.2, the NFS request scheduling device use on this programme supposition S adds right queuing method and from queue, obtains request, and scheduler is arranged in independently periodic duty (this duty cycle is less than all RPC tasks).While obtaining request, by the byte number having completed in current period according to this data flow and also achievable from the byte number side of counting.If also have enough byte number processing in epicycle scheduling, from this traffic queue, take out a request, otherwise this data flow epicycle is not taken out new data and is processed.Moreover, request processing queue is put in pending request by NFS request scheduling device.
Step 4.3, in this programme, the RPC task of all NFS has identical priority and timeslice, and do not use independently local file system access task (in whole real-time network file system framework, to use independently local file system access task, also be a kind of mode of framework that the present invention carries, belong to prescription scope of the present invention)., in the time that a RPC task frees out, obtain next request bag to be processed according to the meter reef knot fruit of NFS request scheduling device.This request is put into the queue of processing request simultaneously, and start timing and record the data read/write amount that it has completed.This task, according to the type and size of request bag, sets it as entirety or is divided into multiple subdivisions and process.When processing, carry out actual file read or write by the interface of this RPC task call bottom local file system.
In step 4.3, while dispatching with flow control to rq1, will carry out concrete management according to the type of rq1 institute corresponding data stream.If rq1 is corresponding to S.V1, bottom scheduler can be optimized according to the special parameter of setting for S.V1.For example S.V1 corresponding data stream only produces sequential write operation, when S.V1 being write, can not use complicated I/O Buffer management policy.If rq1 is corresponding to S.V2.Dir3, RPC thread uses identical real-time management strategy when All Files access request under processing this catalogue, sets up and manages tactful expense independently thereby be reduced to each file.If there is node C4, and for client node C4 definition data flow, the scheduling to this data flow and flow control are by effective to all application on C4.
Note, for whether can, by the problem of different RPC tasks processing and out of order execution, being realized by the mode specifying in the time that step 1 is set real-time parameter from same application and the request of the N continuous FS in one-period.If do not allow out of order execution, this request sequence can only be carried out in same RPC task.
Step 4.4, RPC task returns results entirety packing or gives in order NFS request scheduling device local file system.Request scheduling device according to the real-time parameter value of this data flow and certainly request receive the time of having expended to current time, return to from result the result data bag of choosing proper data stream queue and give network packet receipts/hair device and return to the application in client.After the result data bag of a request is all returned, server S is deleted the project of this request correspondence from queue to be completed.
Step 5: client, according to real-time parameter configuring condition, receives returning results of each data flow by set rate from network.After whole results of a request are all returned, client is deleted this request from pending queue, and finishes to this request of upper strata application report.Whole result return course can be according to user's configuration requirement, to submit to successively or the disposable mode of returning of packing is given upper layer application.If this packet is last packet in this cycle, this data flow is set to finish.
Real-time network file system requests/response architecture model that the present invention proposes, basically identical with the request/response model of existing NFS.But client has increased by one group for realizing flow restriction and the data flow token mechanism of real-time QoS constraint therein; Increase independently request queue/assignment and scheduling and result at server end and returned to the container of thread as real-time scheduling method, and allowed to adopt different real-time network file access request scheduling methods.The framework model that the present invention proposes has not only retained the advantage of existing network file system to greatest extent, has avoided again developing the problem that New Deal may be brought, and provides effective framework model for realizing real-time network file system.
Claims (2)
1. a method for enhancing real-time performance of network file system, is characterized in that step is as follows:
Step 1: arrange real-time parameter between NFS client and network file system server, described real-time parameter, for the real-time traffic control on NFS client and nfs server and Real-Time Scheduling device provide necessary parameter, comprising:
Provide stream identification definition: specifically for the data flow of task, process or thread, node, volume, catalogue, file set, file mark;
Every data flow to following attribute are set less: bandwidth requirement, cycle or largest interval, buffer size, maximum delay or the off period of data flow, transmission direction be unidirectional or two-way, request execution sequence for keep or allow out of order, whether allow to return results in batches, flow control algorithm, dispatching algorithm; Described bandwidth requirement is: unit is bytes per second or I/O number of request per second;
For further improving transmission and storage efficiency, a data flow can be sent one or more read/write requests within the single cycle, but all requests in one-period must all complete within the off period in this cycle; Cycle or the largest interval set comprise multiple data access request;
Step 2: adopt static state or dynamic authentication algorithm, according to the current real-time parameter setting of system and current practice condition, whether checking real-time can meet; If requirement of real time, execution step 3; Otherwise re-start flow planning and real-time parameter setting or stop carrying out this access;
Step 3: NFS client is carried out telefile read-write operation, and the read/write requests of utilizing Real-Time Scheduling and current-limiting apparatus to send each stream carries out current limliting and scheduling, specific as follows:
The first, flow control and dispatching algorithm, taking stream identification as foundation, are budget flow and residual flow in each data-flow computation current period;
The second, NFS client is transformed, adopt by the marking of NFS client, and in this layer, carry out flow control; Or to adopt NFS client be not only each data flow marking, and stamp unique network Flow Control mark for network layer data bag corresponding to each data flow, and carry out flow control by the network layer soft and hardware system of client place node;
Step 4: after network file system server end is ranked to the request receiving, residual flow in the real-time parameter being arranged according to step 1 by file access request scheduling device and each data flow current period, distributes to RPC processing threads with different rates by the request in each data flow request queue and processes; RPC processing threads is received/is sent out scheduler by result through the network packet of network file system server end and returns to client; Concrete enforcement is divided into following four sub-steps:
Step 4.1, arranges special network file access request and receives and send scheduler, and this scheduler is queued up to requests classification according to stream identification;
Step 4.2, arranges special NFS request scheduling device, the real-time parameter that this scheduler is arranged according to step 1, and the thread of the request in queue being given in RPC thread pool with different rates is processed;
Step 4.3, if request bag is excessive, RPC thread is decomposed into a series of less local file access request according to system present load situation, transfers to local file system to carry out with synchronous or asynchronous system; If employing asynchronous system, RPC thread can start to process next request; Described isolation has completed flow, flow to be completed and priority according to this data flow or scheduling strategy determines jointly;
Step 4.4, local file system returns to result after the RPC thread of the request of sending after handling small documents request, the real-time parameter that RPC thread is arranged according to step 1, the mode of returning or returning in batches with unification after packing returns to result the client of the request of sending;
Step 5: client, according to real-time parameter configuring condition, receives returning results of each data flow by set rate from network;
After whole results of a request are all returned, client is deleted this request from pending queue, and finishes to this request of upper strata application report;
Whole result return course is according to user's configuration requirement, to submit to or the disposable mode of returning of packing is given upper layer application in batches;
If this packet is last packet in this cycle, this data flow is set to finish, and stops calculating and queue scheduling to this data flow residual flow in this cycle.
2. the method for enhancing real-time performance of network file system according to claim 1, is characterized in that: in step 4.2, right
In processing aperiodic request, accidental hard Real time request and other special circumstances, by RPC thread pool
Middle introducing different priorities, or distribute the mode of different time of implementation frequencies to be solved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210034549.3A CN102594888B (en) | 2012-02-16 | 2012-02-16 | Method for enhancing real-time performance of network file system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210034549.3A CN102594888B (en) | 2012-02-16 | 2012-02-16 | Method for enhancing real-time performance of network file system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102594888A CN102594888A (en) | 2012-07-18 |
| CN102594888B true CN102594888B (en) | 2014-07-02 |
Family
ID=46483083
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210034549.3A Expired - Fee Related CN102594888B (en) | 2012-02-16 | 2012-02-16 | Method for enhancing real-time performance of network file system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102594888B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104038314B (en) * | 2014-05-09 | 2018-02-02 | 中煤电气有限公司 | A kind of new safety supervision networking dynamic data RTTS and method |
| CN107239328B (en) * | 2016-03-29 | 2020-06-09 | 上海普兰金融服务有限公司 | Task allocation method and device |
| ES2939659T3 (en) * | 2016-04-26 | 2023-04-25 | Umbra Tech Ltd | Sling routing logic and load balancing |
| CN107872395A (en) * | 2016-09-22 | 2018-04-03 | 阿里巴巴集团控股有限公司 | Flow control methods and equipment |
| CN106790645A (en) * | 2017-01-13 | 2017-05-31 | 郑州云海信息技术有限公司 | Protocol streams control method for designing in a kind of distributed storage |
| CN107682280A (en) * | 2017-09-22 | 2018-02-09 | 郑州云海信息技术有限公司 | The method, apparatus and equipment of QOS flows control based on NFS |
| CN109067905B (en) * | 2018-09-05 | 2021-03-16 | 黄秋琼 | Method for realizing decentralized network application |
| CN109309720B (en) * | 2018-10-11 | 2021-10-15 | 郑州云海信息技术有限公司 | Method and system for processing file access request based on nfs protocol |
| CN114640579A (en) * | 2020-11-30 | 2022-06-17 | 华为技术有限公司 | Method, device and system for managing network device |
| CN112751773B (en) * | 2020-12-23 | 2022-08-16 | 信而泰(无锡)信息技术有限公司 | Network traffic scheduling method and device |
| CN116700913B (en) * | 2022-09-13 | 2024-05-31 | 荣耀终端有限公司 | Scheduling method, equipment and storage medium of embedded file system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7356598B1 (en) * | 1999-04-12 | 2008-04-08 | International Business Machines Corporation | System using IP transmission priority for improving overall network response time in telnet 3270 sessions |
| CN101741747A (en) * | 2009-12-23 | 2010-06-16 | 上海交通大学 | NFS (Network File System) flow control method for UDP (User Datagram Protocol) protocol |
| CN101917396A (en) * | 2010-06-25 | 2010-12-15 | 清华大学 | Real-time repetition removal and transmission method for data in network file system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030145230A1 (en) * | 2002-01-31 | 2003-07-31 | Huimin Chiu | System for exchanging data utilizing remote direct memory access |
-
2012
- 2012-02-16 CN CN201210034549.3A patent/CN102594888B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7356598B1 (en) * | 1999-04-12 | 2008-04-08 | International Business Machines Corporation | System using IP transmission priority for improving overall network response time in telnet 3270 sessions |
| CN101741747A (en) * | 2009-12-23 | 2010-06-16 | 上海交通大学 | NFS (Network File System) flow control method for UDP (User Datagram Protocol) protocol |
| CN101917396A (en) * | 2010-06-25 | 2010-12-15 | 清华大学 | Real-time repetition removal and transmission method for data in network file system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102594888A (en) | 2012-07-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102594888B (en) | Method for enhancing real-time performance of network file system | |
| Rajkumar et al. | Resource kernels: A resource-centric approach to real-time and multimedia systems | |
| US7823154B2 (en) | System and method for providing, by a plurality of schedulers, differentiated service to consumers of distributed resources | |
| CN102170396B (en) | QoS control method of cloud storage system based on differentiated service | |
| Molano et al. | Real-time filesystems. Guaranteeing timing constraints for disk accesses in RT-Mach | |
| CN106817317A (en) | Traffic management with in-let dimple | |
| CN109768939A (en) | A kind of labeling network stack method and system for supporting priority | |
| CN109246031A (en) | A kind of switch port queues traffic method and apparatus | |
| CN1606301A (en) | A resource access shared scheduling and controlling method and apparatus | |
| CN110069341A (en) | What binding function configured on demand has the dispatching method of dependence task in edge calculations | |
| Nandagopal et al. | Lowering inter-datacenter bandwidth costs via bulk data scheduling | |
| CN103294548A (en) | Distributed file system based IO (input output) request dispatching method and system | |
| US7986705B2 (en) | Determining a transmission order for frames based on bit reversals of sequence numbers | |
| Barbieru et al. | Soft real-time hadoop scheduler for big data processing in smart cities | |
| Singh et al. | A comparative study of various scheduling algorithms in cloud computing | |
| CN100588174C (en) | Method and system for scheduling synchronous and asynchronous data packets over same network | |
| Rashtian et al. | Balancing message criticality and timeliness in IoT networks | |
| CN109324886A (en) | cluster resource scheduling method and device | |
| Gopalan | Real-time support in general purpose operating systems | |
| Sha et al. | Distributed real-time system design: Theoretical concepts and applications | |
| CN107329819A (en) | A kind of job management method and device | |
| Coulson et al. | Supporting continuous media applications in a micro-kernel environment | |
| Tsai et al. | Wrr-scan: a rate-based real-time disk-scheduling algorithm | |
| Sanati et al. | LBBA: an efficient online benefit-aware multiprocessor scheduling for QoS via online choice of approximation algorithms | |
| AU746999B2 (en) | Device for shared management of a resource among several users |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20140702 Termination date: 20210216 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |