CN103546556A - Virtual machine on-line migration method in future network XIA - Google Patents
Virtual machine on-line migration method in future network XIA Download PDFInfo
- Publication number
- CN103546556A CN103546556A CN201310496753.1A CN201310496753A CN103546556A CN 103546556 A CN103546556 A CN 103546556A CN 201310496753 A CN201310496753 A CN 201310496753A CN 103546556 A CN103546556 A CN 103546556A
- Authority
- CN
- China
- Prior art keywords
- virtual machine
- host
- migration
- main frame
- xia
- 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.)
- Granted
Links
Images
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention relates to a virtual machine on-line migration method in future network XIA. The virtual machine on-line migration method in the future network XIA comprises the following steps of (1) a virtual machine migration sending process, (2) a virtual machine migration receiving process and (3) a performance testing process. In the virtual machine migration sending process, migration data of a virtual machine are written into a standard I/O flow by a host computer of a sender, and data flow and control flow are sent at the same time. In the virtual machine migration receiving process, a host computer of a receiver receives the data flow sent by the host computer of the sender and writes the data flow into the standard I/O flow, and the virtual machine is loaded and run. In the performance testing process, under the condition of single autonomous domain and cross autonomous domains, the host computer of a client side has access to services in the virtual machine through the XIA networks, whether the conducted migration process is on line or not is verified, meanwhile, other services provided by the host computer of the sender are also accessed to in the migration process, and whether the service performance of the host computer of the sender is influenced or not is verified. Compared with the prior art, the virtual machine on-line migration method in the future network XIA has the advantages of achieving on-line seamless migration of the virtual machine, guaranteeing the expedite dynamic self-adapting operation mechanism of the networks after the completion of the migration and the like.
Description
Technical field
The present invention relates to cloud computing technology field and network communications technology field, especially relate to a kind of in future network XIA the online moving method of virtual machine.
Background technology
Along with the development of network interconnection net, the problem that IP network embodies is paid close attention to by society day by day.Academia and industrial quarters are all that the architectural framework of research future network has carried out a large amount of trials, and main thought comprises: (1) is modified and be perfect on original TCP/IP network; (2) cast off the existing TCP/IP network architecture once and for all, set up the brand-new network system.And the viewpoint of the second clean slate becomes the most important thing of research especially.
In FIA (the Future Internet Architecture) project of subsidizing at NSF (National Science Foundation), XIA (eXpressive Internet Architecture) is comparatively typical and at present comparatively perfect one.On the one hand, in XIA system, in network, addressing does not re-use IP, but for dissimilar principal, corresponding XID (as the HID for main frame, for the CID of content, for the SID of service etc.) is set, and user can work accordingly by accessing corresponding XID.The addressing mechanism of XIA network is also more flexible, and the address using is that the mode with DAG represents, when destination node is unreachable, can complete by the fallback path in access DAG.On the other hand, because the address (being respectively CID and HID) of content and service etc. is that mode by Hash obtains, such mechanism has realized security performance inherently.
Intel Virtualization Technology is one of key technology in cloud computing, in Er data center and cluster, by the migrating technology of virtual machine, can effectively complete load balancing, host maintenance and managing electric quantity etc., thus more and more obvious in the practical value of cloud computing and large data age virtual machine (vm) migration technology.
Virtual machine (vm) migration is mainly divided into off-line migration and online migration, wherein off-line migration be exactly by virtual machine under off-mode from a host migration to another main frame, the service that subscriber's main station cannot accesses virtual machine in downtime section; Online migration refers in virtual machine (vm) migration, also will guarantee that the service moving on virtual machine can not interrupt, and the service simultaneously moving in host is not also because virtual machine (vm) migration produces considerable influence.Therefore the online migration of virtual machine is more suitable in the occasion that virtual machine service is had relatively high expectations.
Online migration mainly can be divided into again migration and the piece migration of sharing storage.Share the mode of storage, the source host of migration and the VM on destination host share one and can, by the memory of Network Capture, as SAN or NAS, only need move CPU state and the memory information of virtual machine during migration; Piece migration pattern needs whole virtual machine to carry out complete migration, comprises the content on virtual disk.
At present, to the research of future network often to the design of system and the application and development based on certain framework.Yet the migration that realizes virtual machine based on future network framework is but a blank in research.
Summary of the invention
Object of the present invention be exactly in order to overcome the defect that above-mentioned prior art exists, provide a kind of in future network XIA the online moving method of virtual machine.
Object of the present invention can be achieved through the following technical solutions:
In future network XIA, the online moving method of virtual machine, is characterized in that, comprises the following steps:
(1) virtual machine (vm) migration process of transmitting, transmit leg main frame is written to the migration data of virtual machine in standard I/O stream, sends data flow simultaneously and controls stream;
(2) virtual machine (vm) migration receiving course, recipient's main frame receives the data flow of transmit leg main frame transmission and is written in standard I/O stream, with this, loads and move virtual machine;
(3) performance test process, under single Autonomous Domain and the condition across Autonomous Domain, client host is by the service in XIA access to netwoks virtual machine, and whether the transition process that checking is carried out is online; In transition process, also access other service that transmit leg main frame provides, whether the service performance of checking transmit leg main frame is affected simultaneously.
By the migration data of virtual machine, the exec mode by QEMU-KVM is written in standard I/O stream described transmit leg main frame, sends data flow in CHUNK mode simultaneously, in STREAM mode, sends and controls stream; Recipient's main frame is by reading data flow from CHUNK piece.
Due to the special process that CHUNK mode is transmitted, the CID that recipient's main frame need to provide according to transmit leg main frame (Content ID) proposes request of data to transmit leg main frame;
Every, take turns in request, recipient's main frame,, is not sent request, and directly reads local data running into while having identical CID at local cache by the data that first request arrived.Can effectively solve identical CID cannot requested problem; Meanwhile, can also effectively reduce request number of times, improve data transmission efficiency.
The operating mechanism that also comprises the local dynamic self-adapting of migration virtual machine afterwards, is specially:
In same AD territory, when migration completes, carry out route modification;
In a plurality of different AD territory, while having moved, carry out route modification.
Virtual machine moves to another recipient's main frame and (is referred to as below HOST_B) from transmit leg main frame (being referred to as below HOST_A) on the one hand, running status and the network state of virtual machine the inside all can be held, and therefore the original route table items to HOST_A is also retained; The upper route table items to virtual machine VM of HOST_A is not held because receiving the message of any modification yet on the other hand, and after migration finishes like this, between HOST_A and VM, xping is logical mutually.In same AD, when migration completes, need to delete in the upper route entry to virtual machine VM of HOST_A and VM to the route entry of HOST_A simultaneously, just can guarantee that migration finishes the also normally service in accesses virtual machine VM of recipient's main frame afterwards; And in a plurality of AD, except above-mentioned two, AD_DEFAULT in virtual machine and HID_DEFAULT item also need to revise, the down hop of AD_DEFAULT originally and HID_DEFAULT item is the HID (Host ID) of router in former AD, now needs to be revised as the HID of the router in new AD.In addition, also need to revise the route entry that on first router being connected with migration transmit leg, destination address is VM, because the HID that was VM to the next hop address of the list item of VM on this router originally, router can directly be issued the virtual machine VM on HOST_A by packet like this, if not correct after having moved, can make packet forwarding when through this router of mailing to VM make mistakes and be dropped.
In same AD territory, when migration completes, carry out route and revise specific as follows:
1), after the migration data on transmit leg main frame HOST_A is sent completely, sends order and delete the route entry that in HOST_A, destination host HID is virtual machine;
2) first HOST_A sends to delete command recipient's main frame HOST_B, then by HOST_B, is transmitted to virtual machine VM;
3) HOST_B is when forwarding, timer is set, after forward command, waits for and receive replying of self virtualizing machine VM at every turn, if received, return, illustrate that VM has received delete command and successfully deleted the route entry that arrives HOST_A, otherwise after overtime, again forward this order;
4) virtual machine VM, when receiving delete command, sends a response packet, informs that the route entry that in this virtual machine of HOST_B, destination address is HOST_A is deleted.
In a plurality of different AD territory, while having moved, carry out route modification and be specially:
1), after the migration data on transmit leg main frame HOST_A is sent completely, sends order and delete the route entry that in HOST_A, destination host HID is virtual machine;
2) first HOST_A sends to delete command recipient's main frame HOST_B, then by HOST_B, is transmitted to virtual machine VM;
3) HOST_B is when forwarding, timer is set, after forward command, waits for and receive replying of self virtualizing machine VM at every turn, if received, return, illustrate that VM has received delete command and successfully deleted the route entry that arrives HOST_A, otherwise after overtime, again forward this order; The number of times restriction of setting if surpass is also failed to receive and is replied, stops forwarding this order, and make error handling processing;
4) virtual machine on HOST_B is after continuing operation through downtime, on the one hand when receiving delete command, carry out corresponding deletion action, and send a response packet to HOST_B, inform that the route entry that in this virtual machine, destination address is HOST_A is deleted; On the other hand, due to the special mechanism of XIA, can receive the broadcast packet from the router of new AD, the information that has comprised new AD in this packet, VM, by relatively, is revised as the down hop of the AD_DEFAULT of self and HID_DEFAULT item the HID of the router in new AD;
5) HOST_A identifies by obtaining the AD at HOST_B place, and compare with the AD of self, if find the two difference, send one and revise order to the router of this AD, after this router is received this order, read the list item that arrives HOST_B place AD in routing table, with its next hop address and interface, upgrade the route table items that destination address is VM.
The name Service that also comprises XIA network is upgraded adaptively, is specially:
The name Service of XIA network operates on the transmit leg main frame of migration, when virtual machine is when transmit leg main frame moves, in routing table, can dynamically add the route table items that a destination address is HID_NAMESERVER, yet virtual machine (vm) migration is to after on recipient's main frame, because the state of routing table is held, this is not modified yet, when virtual machine is after recipient's host-initiated, to name server, register new DAG information, wait for replying of name server, and start a timer, when timer expiry, allow program continue operation.
Yet in routing table now, to the route entry of HID_NAMESERVER, owing to being modified, do not become wrong, cause this log-on message to send to, so xhcp_client process can be blocked because receiving to reply.Meanwhile, the order to source host route entry in described deletion VM also cannot be performed because of the obstruction of virtual machine, has so just formed deadlock.A kind of feasible method is that virtual machine, after sending log-on message, starts a timer, rather than blocks wait name server always and reply, and when timer expiry, allows program continue operation, to avoid the generation of deadlock like this.
Described step (3) performance test process also comprises that client host passes through the mode Sampling network performance of xping and measures the downtime time, and the every 100ms of client host sends an xping and asks virtual machine.
Because virtual machine is that mode by bridge joint is connected in network, in XIA prototype, used software router click to carry out processing and the forwarding of packet simultaneously, this mechanism can cause the packet that mails to virtual machine VM also can be processed by the click of host when through bridger, thereby has formed redirected bag.Redirected direct result is each xping bag that mails to VM, all can receive 4 and reply.In the present invention, propose by the redirected function of blocking-up, the packet that can make to mail to virtual machine VM can not produced repetition by correct reception and processing.
Compared with prior art, the present invention has the following advantages:
Under future network framework XIA, realize the online seamless migration of virtual machine, and the dynamic self-adapting operating mechanism of guaranteeing a network connections after migration completes, the development of future network and cloud computing technology is had to certain impetus simultaneously.
Accompanying drawing explanation
Fig. 1 be the present invention based on experiment porch and the corresponding hardware configuration schematic diagram of virtual machine (vm) migration under same AD;
Fig. 2 be the present invention based on experiment porch and the corresponding hardware configuration schematic diagram of virtual machine (vm) migration under a plurality of AD;
Fig. 3 is the flow graph of virtual-machine data while using CHUNK mode to move under XIA.
Embodiment
Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.
Embodiment 1
The present invention is the online moving method of virtual machine in future network XIA, and it at least comprises:
(1) virtual machine (vm) migration sending module.By the migration data of virtual machine, the exec mode by QEMU-KVM is written in standard I/O stream transmit leg main frame, while migrator reading out data from stream, and in the mode of CHUNK data block, virtual machine is intactly moved to recipient's main frame, in transition process with STREAM mode transmitting control information;
(2) virtual machine (vm) migration receiver module.Recipient's main frame is by reading out data from CHUNK piece and be written in standard I/O stream, with this, loads and move virtual machine;
(3) user evaluates and tests module.Under single Autonomous Domain and the condition across Autonomous Domain, client host is by the service in XIA access to netwoks virtual machine, and whether the migration that checking is carried out is online (Live Migration); The echo service providing on host's main frame in migration is also provided simultaneously, and whether the service performance of checking host is affected.In addition, user's mode Sampling network performance by xping measuring the downtime time also.
This moving method comprises the steps:
Step 1: build the required migration platform of experiment, and configure corresponding XIA network (with AD or a plurality of AD);
Step 2: start KVM virtual machine on sourcesink main frame;
Step 3: the order by KVM on object host main frame starts virtual machine image, waits migration data to be loaded;
Step 4: input migration order in the terminal of QEMU, start the piece migration of exec mode, specific instructions is: migrate-b exec:/xia_client_chunk_10, wherein xia_client_chunk_10 is the transmission program of migration;
Step 5: start calculation services (for example application of CPU intensity---the checking to Goldbach's Conjecture is about to each even number and is decomposed into two prime number sums) in VM, and output on user rs host HOST_C;
Step 6: start echo_server process on sourcesink main frame, the random word string of sending from client is carried out to simple process and returned to client;
Step 7: be spaced apart 100 milliseconds by adjusting giving out a contract for a project of xping (program of ping in the similar IP network in XIA) on HOST_C, constantly xping virtual machine VM;
Step 8: migration finishes, and migrating processes triggers and complete the operative mechanism of Network aware automatically, guarantees that virtual machine normally carries out alternately with extraneous main frame after migration;
Step 9: whether the program that checks calculation services in the upper request of HOST_C VM stops, and checks the packet drop of xping simultaneously, estimation downtime;
Step 10: with same method, the virtual machine moving on HOST_B is moved back to HOST_A again.
The detailed implementation procedure of step 1 recited above also comprises:
1) configuration/etc/network/interfaces file on the sourcesink master host moving and object host main frame respectively adds br0, and carries out corresponding bridge configuration in this document;
2) start the XIA network of main frame.HOST_A and HOST_C start XIA network as host pattern, specific instructions is xianet-t-i hosta/hostc start, HOST_B is used as host on the one hand, and nameservice service is provided on the other hand, orders the start into xianet-t-n-i hostb;
3) start the XIA network of router.Because existing router is not supported XIA network, so the router under XIA need to be realized by the main frame operating software router click of many network interface cards.The order that starts router is xianet-r-n-i router start.
State in realization in the process of method, the present invention has also solved some relevant technical problems, mainly comprises:
1) be redirected the problem of bag.After configuration network, from HOST_C or HOST_B, xping, to VM, can find the request of an xping of every transmission, can receive four response packets; And during HOST_A xping VM, can receive two at every turn and reply.By the br0 at HOST_A, eth0, tap0 network interface card, open respectively wireshark packet capturing simultaneously, analyze the flow direction of each packet, contrast the source code of XIA simultaneously, discovery is owing to having used click software router in XIA framework, make host main frame also there is the function of the redirected packet of router, and in fact, redirected function is generally provided by router, as end host, should not there is this function.Each mails to the packet of VM, when by bridge joint br0, bridger has been transmitted to tap0 network interface card on the one hand, simultaneously because XIA starts host pattern based on interface br0, so click also can receive this packet, now this packet of HOST_A detecting host mails to VM, just produce and be redirected, but simultaneously also can be by package forward to VM, so VM has just received two xping bags that sequence number is identical, VM processes and will send after request the replying of this two bags, and has two response packets.Final receiver but these two response packets are being received by HOST_A again when the br0, and produced and has been redirected, so can be received replying of four xping.The scheme solving is: disposing on the main frame of bridge joint, producing the place blocking-up that is redirected bag in the click program of XIA, at every turn when run to can generation packet redirected time, no longer carry out follow-up program and directly return.So just can effectively block redirectedly, make xping and other packet can normally arrive virtual machine VM.
2) problem of identical CID.In migrator design, the data that at every turn read from stream are divided into 10 CHUNK and transmit, the size of each CHUNK is 10K.When carrying out migration; transition process often can just stop after having sent several CHUNK; by printing the CID of each CHUNK that recipient receives; find in the CID of 10 CHUNK that when each migration stops, HOST_B receives; always have identical CID; tracing it to its cause is to have identical CHUNK (as full 0) in 10 CHUNK being divided into of the data block that need to send, and CID obtains by Hash, will cause so identical CID to produce.XIA does not have to consider to contain identical CID in each information of asking when design, therefore its processing method is after each recipient (HOST_B) asks a CID, transmit leg (HOST_A) is just deleted content corresponding to this CID from Hash Table, will cause like this, when running into identical CID request, will makeing mistakes because can not find corresponding CHUNK.A kind of simple effective method is after the each request of recipient (HOST_B) receives data, when being written to I/O stream, CID and corresponding CHUNK content are first buffered in to this locality, before sending next CID request, first check local cache, if can find identical CID, just only from this locality, obtain CHUNK content, and without sending again request to data receiver.
3) network state dynamic self-adapting mechanism problem after migration.A key character of virtual machine (vm) migration is exactly, and the final state that the virtual machine state after having moved should be with on sourcesink main frame time is consistent.Like this, virtual machine moves to another main frame HOST_B above from host's main frame HOST_A, and all states of virtual machine the inside all can be held, and comprise that in original VM, the route table items to HOST_A is also retained; Meanwhile, the upper route table items to virtual machine VM of HOST_A is not retained because receiving the message of any modification yet, and after migration finishes like this, between HOST_A and VM, xping is logical mutually.When migration finishes, need to delete in the upper route entry to virtual machine VM of HOST_A and VM to the route entry of HOST_A simultaneously, just can guarantee that migration finishes the also normally service in accesses virtual machine VM of sourcesink master host afterwards.
4) how effectively to delete 3) described in the problem of route table items.After migrator transmission data on HOST_A complete, the route entry that in HOST_A, destination host HID is virtual machine is deleted in transmitting system order.Yet, on HOST_B, because being accomplished to startup completely from migration, virtual machine can need the regular hour, therefore after recipient's reception migrator executes, can not directly send the order of deletion routing table to virtual machine VM, otherwise this packet can be lost because virtual machine is not activated.In this method, propose first delete command to be sent to Xinsu master host HOST_B (because the finger daemon xhcp_client of HOST_B is monitoring always and processing the packet of receiving), then by HOST_B, be transmitted to virtual machine VM; HOST_B is when forwarding, timer need to be set, if receive response packet after each forward command, illustrate that VM has received delete command and successfully deleted the route entry that arrives HOST_A, otherwise again forward this order after overtime (0.5s), when expired times are greater than 50 times, no matter whether receive and reply, all stop forwarding delete command.
5) the adaptive updates mechanism problem of the name Service of XIA network.In virtual machine (vm) migration process, if the name Service of XIA network (nameservice) operates on the sourcesink master host of migration, when virtual machine is when sourcesink main frame moves, in routing table, can dynamically add the route table items that a destination address is HID_NAMESERVER, yet virtual machine (vm) migration to destination host after, because the state of routing table is held, this is not modified yet, when virtual machine is after destination host starts, its xhcp_client process can be registered new DAG information to name Service, and waits for replying of name server in the mode of blocking.Yet in routing table now, to the route table items of HID_NAMESERVER, owing to being modified, do not become wrong, cause this log-on message to send to, so xhcp_client process can be blocked because receiving to reply.Meanwhile, the deletion of the above also cannot be performed because of the obstruction of virtual machine to the order of source host route entry, has so just formed deadlock.A kind of feasible method that the present invention proposes is that virtual machine, after sending log-on message, starts a timer, rather than blocks wait name server always and reply, and when timer expiry, allows program continue operation, to avoid the generation of deadlock like this.
Embodiment 2
Present case is chosen in the single AD of XIA network, virtual machine adopts from sourcesink master host HOST_A and object host main frame HOST_B, after completing, move back to the process of HOST_A, the while, HOST_C was as user rs host, the service providing in accesses virtual machine and host again.Concrete steps are as follows:
1) according to the topology shown in Fig. 1, build experiment porch, wherein HOST_A, HOST_B are two notebooks, configuration i3 processor, 8GB internal memory, and to distribute the swap subregion of 8GB, ROUTER be to be inserted with three PCs of throwing the net card, configuration i3 processor, the swap subregion of 4GB internal memory and 8GB, HOST_C is used as user rs host, configures equally i3 processor, the swap subregion of 4GB internal memory and 8GB;
2) bridge joint on configuration HOST_A and HOST_B, makes virtual machine can be linked in network before and after migration;
3) on HOST_A, move order and the relevant parameter startup virtual machine of KVM, design parameter is:
／usr／local／kvm／bin／qemu-system-x86_64
-m640
-smp1
-drive?file=／var／lib／libvirt／images／ubuntu.img，if=virtio，format=raw
-net?nic，vlan=0，macaddr=52：54：00：6d：82：f9
-net?tap，vlan=0，ifname=tap0，script=no
4) configuration XIA network, be specially HOST_A, HOST_VM, HOST_C moves host pattern (xianet-t-i hosta/hostvm/hostc start), HOST_B also starts nameservice service (xianet-t-n-i hostb start) in operation host, and all main frames can be registered to the host name of oneself and address (representing with DAG) on nameserver; ROUTER is used as router, when starting, moves router pattern (xianet-r-n-i router start);
5) transmit leg that in configuration file/NAMEFILE, input is this time moved respectively on source host HOST_A and destination host HOST_B and recipient's host name (being hosta, hostb here);
6) mirror image that the order that moves KVM on HOST_B starts virtual machine with etc. migration data to be loaded, design parameter is 3) in parameter add
-incoming exec:/xia-core/applications/example/xia_server_chunk_10_ 3, wherein xia_server_chunk_10_3 is the sender terminal of the virtual machine (vm) migration of realization;
7) at virtual machine window interface, press ctrl+alt+2, open the control desk panel of QEMU, input starts migration order migrate-b exec:/xia-core/applications/example/xia_client_chunk_10_ 3 to start transition process, the transmission program that wherein xia_client_chunk_10_3 moves under XIA for the virtual machine of realizing;
8) in virtual machine VM, start the process (as verified Goldbach's Conjecture by calculating) of a CPU intensity, and on user rs host HOST_C, print the Output rusults of calculating;
9) on HOST_A, start echo_server process, start echo_client on HOST_C, its function is that server returns to client end by the character string of random 512 bytes that generate of client after simple process;
10) on HOST_C, start xping; with the time intervals of every 100 milliseconds, send the request bag (realizing by add timer mechanism in the source code of xping) of an xping; the unimpeded situation of Sampling network; can estimate at the downtime downtime of virtual machine (vm) migration simultaneously; the packet loss number that can calculate is altogether 48, so the downtime time is between 4.7s-4.9s.
11) end to be migrated such as, xping connection status after inspection migration in client computer and the access situation of calculation services, and check whether routing table is revised, comprise the route entry that in route entry that in HOST_VM, object HID is HOST_A and HOST_A, object HID is HOST_VM.Owing to having added the program of at once revising route table items while having moved in migrator, after having moved, can xping logical between (comprising HOST_A, HOST_VM, HOST_B, HOST_C) between all main frames.
12) check complete after, once source host and the destination host name of migration in input in configuration file, here owing to being to move back to HOST_A, so input source main frame and destination host are respectively: hostb, hosta.
13) close the virtual machine being suspended on HOST_A, again according to 6) in method open VM mirror image and wait for and receive migration;
14) for the current virtual machine VM having operated on HOST_B, according to 7) in mode input migration order, this virtual machine is moved back to HOST_A main frame again, to simulate virtual machine in actual applications, need the process of moving to and fro between a plurality of hosts.
15) transition process finishes, and can keep unimpeded between All hosts, and virtual machine also can continue operation on HOST_A, and normally externally communicates.
Claims (8)
1. the online moving method of virtual machine in future network XIA, is characterized in that, comprises the following steps:
(1) virtual machine (vm) migration process of transmitting, transmit leg main frame is written to the migration data of virtual machine in standard I/O stream, sends data flow simultaneously and controls stream;
(2) virtual machine (vm) migration receiving course, recipient's main frame receives the data flow of transmit leg main frame transmission and is written in standard I/O stream, with this, loads and move virtual machine;
(3) performance test process, under single Autonomous Domain and the condition across Autonomous Domain, client host is by the service in XIA access to netwoks virtual machine, and whether the transition process that checking is carried out is online; In transition process, also access other service that transmit leg main frame provides, whether the service performance of checking transmit leg main frame is affected simultaneously.
According to claim 1 a kind of in future network XIA the online moving method of virtual machine, it is characterized in that, by the migration data of virtual machine, the exec mode by QEMU-KVM is written in standard I/O stream described transmit leg main frame, in CHUNK mode, send data flow, in STREAM mode, send and control stream simultaneously; Recipient's main frame is by reading data flow from CHUNK piece.
According to claim 2 a kind of in future network XIA the online moving method of virtual machine, it is characterized in that, due to the special process that CHUNK mode is transmitted, the CID that recipient's main frame need to provide according to transmit leg main frame proposes request of data to transmit leg main frame;
Every, take turns in request, recipient's main frame,, is not sent request, and directly reads local data running into while having identical CID at local cache by the data that first request arrived.
According to claim 2 a kind of in future network XIA the online moving method of virtual machine, it is characterized in that, also comprise the operating mechanism of the local dynamic self-adapting of virtual machine after migration, be specially:
In same AD territory, when migration completes, carry out route modification;
In a plurality of different AD territory, while having moved, carry out route modification.
According to claim 4 a kind of in future network XIA the online moving method of virtual machine, it is characterized in that, in same AD territory, when migration completes, carry out route and revise specific as follows:
1), after the migration data on transmit leg main frame HOST_A is sent completely, sends order and delete the route entry that in HOST_A, destination host HID is virtual machine;
2) first HOST_A sends to delete command recipient's main frame HOST_B, then by HOST_B, is transmitted to virtual machine VM;
3) HOST_B is when forwarding, timer is set, after forward command, waits for and receive replying of self virtualizing machine VM at every turn, if received, return, illustrate that VM has received delete command and successfully deleted the route entry that arrives HOST_A, otherwise after overtime, again forward this order;
4) virtual machine VM, when receiving delete command, sends a response packet, informs that the route entry that in this virtual machine of HOST_B, destination address is HOST_A is deleted.
According to claim 4 a kind of in future network XIA the online moving method of virtual machine, it is characterized in that, in a plurality of different AD territory, while having moved, carry out route and revise and to be specially:
1), after the migration data on transmit leg main frame HOST_A is sent completely, sends order and delete the route entry that in HOST_A, destination host HID is virtual machine;
2) first HOST_A sends to delete command recipient's main frame HOST_B, then by HOST_B, is transmitted to virtual machine VM;
3) HOST_B is when forwarding, timer is set, after forward command, waits for and receive replying of self virtualizing machine VM at every turn, if received, return, illustrate that VM has received delete command and successfully deleted the route entry that arrives HOST_A, otherwise after overtime, again forward this order; The number of times restriction of setting if surpass is also failed to receive and is replied, stops forwarding this order, and make error handling processing;
4) virtual machine on HOST_B is after continuing operation through downtime, on the one hand when receiving delete command, carry out corresponding deletion action, and send a response packet to HOST_B, inform that the route entry that in this virtual machine, destination address is HOST_A is deleted; On the other hand, due to the special mechanism of XIA, can receive the broadcast packet from the router of new AD, the information that has comprised new AD in this packet, VM, by relatively, is revised as the down hop of the AD_DEFAULT of self and HID_DEFAULT item the HID of the router in new AD;
5) HOST_A identifies by obtaining the AD at HOST_B place, and compare with the AD of self, if find the two difference, send one and revise order to the router of this AD, after this router is received this order, read the list item that arrives HOST_B place AD in routing table, with its next hop address and interface, upgrade the route table items that destination address is VM.
According to claim 4 a kind of in future network XIA the online moving method of virtual machine, it is characterized in that, also comprise that the name Service of XIA network is upgraded adaptively, be specially:
The name Service of XIA network operates on the transmit leg main frame of migration, when virtual machine is when transmit leg main frame moves, in routing table, can dynamically add the route table items that a destination address is HID_NAMESERVER, yet virtual machine (vm) migration is to after on recipient's main frame, because the state of routing table is held, this is not modified yet, when virtual machine is after recipient's host-initiated, to name server, register new DAG information, wait for replying of name server, and start a timer, when timer expiry, allow program continue operation.
According to claim 1 a kind of in future network XIA the online moving method of virtual machine, it is characterized in that, described step (3) performance test process also comprises that client host passes through the mode Sampling network performance of xping and measures the downtime time, and the every 100ms of client host sends an xping and asks virtual machine.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310496753.1A CN103546556B (en) | 2013-10-22 | 2013-10-22 | One kind online moving method of virtual machine in future network XIA |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310496753.1A CN103546556B (en) | 2013-10-22 | 2013-10-22 | One kind online moving method of virtual machine in future network XIA |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103546556A true CN103546556A (en) | 2014-01-29 |
| CN103546556B CN103546556B (en) | 2017-04-05 |
Family
ID=49969595
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310496753.1A Expired - Fee Related CN103546556B (en) | 2013-10-22 | 2013-10-22 | One kind online moving method of virtual machine in future network XIA |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103546556B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104184803A (en) * | 2014-08-08 | 2014-12-03 | 同济大学 | DAG based method for adaptively completing online migration of virtual machine in XIA network |
| CN104754033A (en) * | 2015-03-02 | 2015-07-01 | 同济大学 | Method for realizing online migration based on bidirectional pipeline manner under NDN (Named Data Networking) |
| CN106973014A (en) * | 2017-01-22 | 2017-07-21 | 阿里巴巴集团控股有限公司 | Anycast route delivery system, method and electronic equipment |
| CN108139974A (en) * | 2015-10-21 | 2018-06-08 | 华为技术有限公司 | distributed caching dynamic migration |
| CN109639705A (en) * | 2018-12-27 | 2019-04-16 | 成都国信安信息产业基地有限公司 | Cloud platform safety detection method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101465812A (en) * | 2009-01-06 | 2009-06-24 | 北京航空航天大学 | Redirection method for virtual machine network connection when on-line migrating striding sub network |
| CN101609419A (en) * | 2009-06-29 | 2009-12-23 | 北京航空航天大学 | The data back up method and the device of the migration of virtual machine consistently online |
| US20120185856A1 (en) * | 2009-09-28 | 2012-07-19 | Koji Ashihara | Computer system and migration method of virtual machine |
| CN102609361A (en) * | 2012-01-16 | 2012-07-25 | 北京邮电大学 | Method and device for transferring storage data of virtual machine |
-
2013
- 2013-10-22 CN CN201310496753.1A patent/CN103546556B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101465812A (en) * | 2009-01-06 | 2009-06-24 | 北京航空航天大学 | Redirection method for virtual machine network connection when on-line migrating striding sub network |
| CN101609419A (en) * | 2009-06-29 | 2009-12-23 | 北京航空航天大学 | The data back up method and the device of the migration of virtual machine consistently online |
| US20120185856A1 (en) * | 2009-09-28 | 2012-07-19 | Koji Ashihara | Computer system and migration method of virtual machine |
| CN102609361A (en) * | 2012-01-16 | 2012-07-25 | 北京邮电大学 | Method and device for transferring storage data of virtual machine |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104184803A (en) * | 2014-08-08 | 2014-12-03 | 同济大学 | DAG based method for adaptively completing online migration of virtual machine in XIA network |
| CN104184803B (en) * | 2014-08-08 | 2017-12-15 | 同济大学 | The method that virtual machine migrates online in adaptive completion XIA networks based on DAG |
| CN104754033A (en) * | 2015-03-02 | 2015-07-01 | 同济大学 | Method for realizing online migration based on bidirectional pipeline manner under NDN (Named Data Networking) |
| CN104754033B (en) * | 2015-03-02 | 2018-04-17 | 同济大学 | The method that virtual machine migrates online is realized based on two-way pipe method under NDN |
| CN108139974A (en) * | 2015-10-21 | 2018-06-08 | 华为技术有限公司 | distributed caching dynamic migration |
| CN106973014A (en) * | 2017-01-22 | 2017-07-21 | 阿里巴巴集团控股有限公司 | Anycast route delivery system, method and electronic equipment |
| CN106973014B (en) * | 2017-01-22 | 2021-07-30 | 创新先进技术有限公司 | Anycast route publishing system and method and electronic equipment |
| CN109639705A (en) * | 2018-12-27 | 2019-04-16 | 成都国信安信息产业基地有限公司 | Cloud platform safety detection method |
| CN109639705B (en) * | 2018-12-27 | 2021-08-31 | 成都国信安信息产业基地有限公司 | Cloud platform security detection method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103546556B (en) | 2017-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20220061059A1 (en) | Distributed network connectivity monitoring of provider network edge location resources from cellular networks | |
| US10887276B1 (en) | DNS-based endpoint discovery of resources in cloud edge locations embedded in telecommunications networks | |
| US10909066B2 (en) | Virtual RDMA switching for containerized applications | |
| US11095534B1 (en) | API-based endpoint discovery of resources in cloud edge locations embedded in telecommunications networks | |
| US9535871B2 (en) | Dynamic routing through virtual appliances | |
| US10255147B2 (en) | Fault tolerance for containers in a virtualized computing environment | |
| US9454392B2 (en) | Routing data packets between virtual machines using shared memory without copying the data packet | |
| US9363172B2 (en) | Managing a configurable routing scheme for virtual appliances | |
| US9548959B2 (en) | Address resolution suppression for data center interconnect | |
| US9110703B2 (en) | Virtual machine packet processing | |
| CN114946166A (en) | Dynamic resource movement in heterogeneous computing environments including cloud edge locations | |
| US9537798B1 (en) | Ethernet link aggregation with shared physical ports | |
| US9910687B2 (en) | Data flow affinity for heterogenous virtual machines | |
| CN101924693A (en) | Be used for method and system in migrating processes between virtual machines | |
| CN103546556A (en) | Virtual machine on-line migration method in future network XIA | |
| JP6604336B2 (en) | Information processing apparatus, information processing method, and program | |
| US11219034B1 (en) | Distributed network connectivity monitoring of provider network edge location resources from cellular networks | |
| US11743325B1 (en) | Centralized load balancing of resources in cloud edge locations embedded in telecommunications networks | |
| US10628198B2 (en) | Hypervisor management of migration notification and response messages for virtual machines | |
| US8533320B2 (en) | Coalescing network notifications for live migration | |
| US11403319B2 (en) | High-availability network device database synchronization | |
| CN115225634B (en) | Data forwarding method, device and computer program product under virtual network | |
| US11537422B2 (en) | Virtual machine migration downtime reduction using a multicast address | |
| CN113127144A (en) | Processing method, processing device and storage medium | |
| US11748131B2 (en) | Network updates for virtual machine migration |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170405 Termination date: 20191022 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |