JP2018032936A - Routing control system and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To optimize routing of a virtual machine in a data center and a client terminal under external network, even under circumstances of connecting data centers by a wide area network.SOLUTION: Data centers 10a, 10b perform routing control by using a set of an IP address of a virtual machine 40, as an ID, and Locator in a wide area L2 network 60, and each of gateways 35a, 35b advertises the IP address of the virtual machine 40, as routing information, for external networks 50a, 50b, respectively. The external network 50 transfers a packet, from a client terminal 1 to the advertised IP address, to a gateway located in the neighborhood of the client terminal 1.SELECTED DRAWING: Figure 4

Description

  The present invention relates to a route control technique in a wide area network connecting a plurality of data centers.

  In a data center, a plurality of logical networks are included in the data center, so that a plurality of services and networks of businesses are mixed and superimposed on one data center network.

  Further, a plurality of data centers may be connected using a dedicated line or a tunneling protocol to form a wide area L2 network, thereby forming a huge virtual network of services and operators. By doing so, it becomes possible to allocate necessary resources to places where resources are required. For example, for a business continuity plan (BCP), important resources in the company are moved to an unaffected area at the time of a disaster and the business is continued.

  In many use cases of data centers such as XaaS, represented by IaaS (Infrastructure as a Service) and SaaS (Software as a Service), resources are allocated in the data center. Exists outside the data center. FIG. 1 illustrates a usage form of XaaS. As shown in FIG. 1, the data center 10 includes a server 20 that provides XaaS and a gateway 30, and one or more virtual machines (VMs) 40 are constructed in the server 20. The data center 10 at the base in the A area can communicate with the client terminal 1 via the network 50 in the A area.

  A case where the client terminal 1 is used from a distance from the data center 10 will be described with reference to FIG. FIG. 2 is a diagram showing a form of using XaaS from a long distance. In FIG. 2, symbols such as “a” and “b” are added after the reference numbers when distinguishing a plurality of existing elements such as the network 50, for example. The same applies to the following description and other drawings.

  When used from a distance, it is used from a location far away from a physical server or storage. In the example of FIG. 2, the client terminal 1 receives data via the network 50b in the B region, a transmission path (for example, a submarine cable) between the network 50a in the A region and the network 50b in the B region, and the network 50a in the A region. Communicate with the center 10. For this reason, there is a problem that QoE (Quality of Experience) is remarkably lowered or cannot be used due to delay or congestion in the route. In order to solve this problem, it is conceivable to move the VM 40 to the data center in the vicinity of the user. Note that moving the VM 40 to another server in another data center or the like while maintaining service provision by the VM 40 is called live migration.

D. Farinacci, 3 others, "The Locator / ID Separation Protocol (LISP)", RFC6830, January 2013, [online], [searched July 21, 2016], Internet <https: // tools. ietf.org/html/rfc6830> C. Partridge, 2 others, "Host Anycasting Service", RFC1546, November 1993, [online], [searched July 21, 2016], Internet <https://tools.ietf.org/html/ rfc1546>

  An example in which a VM is moved across data centers is illustrated in FIG. As shown in FIG. 3, the data center 10a and the data center 10b are communicably connected by the inter-base connection 65. In other words, a logical wide area L2 network 60 including the data center 10a and the data center 10b is formed. Is formed. The wide area L2 network 60 may be virtualized by overlay technology such as VLAN (Virtual Local Area Network) or VxLAN (Virtual eXtensible Local Area Network), or a dedicated line or IP-VPN as the inter-base connection 65. It may be configured almost physically using (Internet Protocol-Virtual Private Network). Before and after the movement of the VM 40, the gateway 30a of the data center 10a in the A area is directed toward the network 50a side of the A area, which is a network outside the data center 10a, in other words, toward the outside of the wide area L2 network 60. The global IP (Internet Protocol) address of the VM 40 is advertised as route information. For this reason, since the advertisement source of the global IP address of the VM 40 does not change when viewed from outside the wide area L2 network 60, the packet is routed to the data center 10a of the original base, and so-called “trombone” which becomes a redundant route. "Traffic Trombone" occurs. In the example of FIG. 3, communication between the client terminal 1 and the VM 40 takes a route of the network 50b in the B area, the Internet 50c, the network 50a in the A area, the data center 10a, the inter-base connection 65, and the data center 10b.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a route between a virtual machine in a data center and a client terminal under an external network even in an environment in which the data centers are connected by a wide area network. It is to provide a routing system and method that can be optimized.

  To achieve the above object, the present invention provides a communication system comprising a plurality of data centers connected to an external network by a gateway and provided with virtual machines, and connected to the plurality of data centers by a logical wide area network. Is a path control system for controlling a path between a client terminal and a virtual machine via a gateway, and a path between the gateway and the virtual machine using a combination of the ID and Locator with the IP address of the virtual machine as an ID in the wide area network The plurality of data centers are configured such that each of the plurality of gateways advertises the IP address of the virtual machine to the external network as route information, and is addressed to the advertised IP address from the client terminal. The pa Tsu bets is characterized by being configured the external network to be forwarded to the gateway located in the vicinity of the client terminal among the plurality of gateways.

  According to the present invention, when a client terminal sends a packet addressed to an advertised IP address, the packet reaches a data center located in the vicinity of the client terminal. On the other hand, in the wide area network, the packet of the advertisement IP address reaches the virtual machine by the ID / Locator separation technique. That is, the route between the virtual machine in the data center and the client terminal under the external network is optimized to the shortest without causing the trombone phenomenon.

The figure explaining the usage form of XaaS The figure explaining the case where XaaS is used from a long distance Diagram explaining the trombone phenomenon The figure explaining the outline | summary of this invention Basic configuration diagram of the route control system of the present invention functional diagram of exMS / MR The figure explaining an example of a mapping table The figure explaining movement of VM VM migration sequence Packet transfer control sequence Packet transfer control sequence The figure explaining the communication path between a terminal and VM

  A route control system according to an embodiment of the present invention will be described with reference to the drawings. First, the outline of the present invention will be described with reference to FIG. FIG. 4 is a diagram for explaining the outline of the present invention. In addition, the same code | symbol is attached | subjected about the conventional same element here.

  The present invention advertises IP addresses that may move from each site in advance, and IP mobility that enables route optimization only by modifying the routing inside the wide area L2 network 60 whenever the VM 40 moves. It is a route control system. The timing of this advertisement is a time sufficient for the advertisement to be propagated to the external network 50 and to be route-controlled based on the advertisement in the external network 50 by the time when the movement of the VM 40 to the other base is completed. It is preferable to ensure.

  One of the points of the present invention is that the route control system establishes a connection route from the nearest gateway 30 to the VM 40 following the moved VM 40. In addition, one of the points of the present invention is that the network 50 to which the client terminal 1 belongs, autonomously selects a nearby one of a plurality of bases advertising the global IP of the VM 40, and sends packets from the client terminal 1 to the base Therefore, the trombone phenomenon does not occur. In addition, one of the points of the present invention is that, within a base, a packet that arrives from the outside in accordance with the movement of the VM 40 is transferred to the VM 40 via an inter-base connection through an appropriate route within the base and between the bases. Is.

  As means for realizing such a point, in the present invention, Locator / ID separation is performed on the IP address of the VM 40 in the wide area L2 network 60, and the IP address of the VM 40 is used as the ID. The Locator uses some address information that can identify the location of the VM 40 in the wide area L2 network 60. What should be noted here is that the Locator may change depending on the movement of the VM 40 between the data centers 10, but the ID is fixed.

  The gateway 35 of the data center 10 at each base advertises the IP address as the ID as route information to the VM 40 to the outside. That is, there are a plurality of gateways 35 that advertise the same IP address. On the other hand, in the network 50 outside the wide area L2 network 60, the route control is performed so that the packet addressed to the IP address is transferred to the gateway 35 of the data center 10 in the vicinity of the plurality of gateways 35 that advertise the IP address. I do.

  As a result, as shown in FIG. 4, even if the VM 40 moves, the client terminal 1 connects to the VM 40 with the shortest path to the nearest gateway 35. Further, even if the network configuration changes in the wide area L2 network 60, it is transferred to an appropriate node by the gateway 35.

  Next, a basic configuration of the route control system according to the present embodiment will be described with reference to FIG. In this embodiment, LISP (The Locator / ID Separation Protocol) is used as the Locator / ID separation means. In the LISP, the ITR (Ingress Tunnel Router) performs route advertisement, forwards the received packet to the ETR (Egress Tunnel Router) under which the node having the destination IP address exists, according to the contents of the table, and the ETR This is a technique for transferring the packet to a node having a destination IP address. See Non-Patent Document 1 for details of LISP.

  On the external network side, IP Anycast technology is used. IP Anycast is an act of performing route advertisement from a plurality of AS (Autonomous System). Refer to Non-Patent Document 2 for details of IP Anycast.

  The data center 10 is a base capable of route advertisement and has a function of accommodating the virtual network 15. As shown in FIG. 5, each data center 10 includes an exITR (extended ITR) 36, an exMS / MR (extended Mapping Server / Mapping Resolver) 70, a virtual network 15, An ETR 16 arranged at the boundary with the virtual network 15, a VM 40 arranged in the virtual network 15, and a data center controller 17 are provided. The plurality of data centers 10 are connected to each other through a tunnel 65, thereby forming a wide area L2 network 60 connecting the plurality of data centers 10. The wide area L2 network 60 may be virtualized by an overlay technique such as VLAN or VxLAN), or may be almost physically configured as a tunnel 65 using a dedicated line or IP-VPN. Also good.

  In the data center 10, the inside of the virtual network 15 is an EID (Endpoint ID) name space, and the outside of the virtual network 15 is an RLOC (Routing Locator) name space. Here, EID is information for identifying a terminal in the LISP. RLOC is an IP address assigned to the tunnel router in the LISP. In the LISP, a UDP (User Datagram Protocol) tunnel is generated between the RLOC of the transmission side tunnel router and the RLOC of the destination side tunnel router, and the actual communication passes therethrough. The inter-base connection tunnel 65 is characterized by high speed because a dedicated line is generally used.

  The exITR 36 extends the function of the LISP and has a function equivalent to the ITR defined by the LISP, that is, a function as a tunnel router that performs encapsulation at the transmission side site. Furthermore, the exITR 36 has a function of advertising a route to the external network 50 side, and an NAT (Network Address Translation) function as an optional function. The exITR 36 has a function of accepting a request to delete a route information cache and deleting the cache.

  The exMS / MR 70 extends the functions of the LISP. As shown in FIG. 6, the MS function unit 71 having a function equivalent to the MS defined by the LISP, that is, a function of accepting a registration request from the ETR, And an MR function unit 72 having a function equivalent to the MR, that is, a function of receiving an inquiry from the ITR. The exMS / MR 70 has a function of mapping EID and RLOC, and includes a mapping table 73 for providing the function. Further, the exMS / MR 70 includes a synchronization function unit 74 having a function of synchronizing the mapping table 73 owned by the other exMS / MR 70.

  FIG. 7 shows an example of the mapping table 73. FIG. 7A shows an example when the IP address of the VM 40 is a global IP address. As shown in FIG. 7A, the mapping table 73 includes the IP address of the VM 40 as EID, the IP address of the ETR 16 of the network to which the VM 40 belongs as RLOC, and Priority.

  FIG. 7B shows an example when the IP address of the VM 40 is a private address. When exITR 36 also performs NAT processing of private IP and global IP at the same time, it is necessary to replace the global IP address of VM 40 with the private IP address and then check the mapping to identify the RLOC. ) Table configuration is required. As shown in FIG. 7B, the mapping table 73 is assigned to the VM 40 as the IP address or IP prefix of the VM 40 as the EID, the IP address of the ETR 16 of the network to which the VM 40 belongs as the RLOC, and the global IP address of the VM. It includes a global IP address or IP prefix and Priority.

  Priority in FIG. 7 is used to give priority to a higher Priority value when a plurality of RLOCs are set for one EID.

  The ETR 16 is defined by LISP, and is a tunnel router that releases the encapsulation at the destination site.

  The data center controller 17 has a function of migrating the VM 40 to another data center 10 and a function of notifying the ETR 16 that the VM 40 has migrated.

  The VM 40 is a virtualized environment such as a virtual machine in hypervisor type virtualization or a container in container type virtualization.

  The network 50 allows a plurality of nodes (exITR 36 in the present embodiment) to advertise the same IP address as route information corresponding to IP Anycast. A packet destined for the IP address from the client terminal 1 is routed to a node near the client terminal 1. In the present embodiment, BGP (Border Gateway Protocol) is used as an example of means for performing such route control.

  Next, a sequence at the time of movement of a VM in the route control system according to the present embodiment will be described with reference to FIGS. FIG. 8 shows a configuration diagram of the path control system after the VM movement. Here, a sequence when the VM 40 in the data center 1 (10a) moves into the data center 2 (10b) will be described with reference to FIG. In this sequence, the mapping table is updated and the cache is deleted during VM migration. Hereinafter, the sequence will be described in detail.

  First, as preparations prior to the VM migration process, in each data center 10, between the ETR 16 and the MR function unit 72 of the exMS / MR 70, and between the exITR 36 and the MR function unit 72 of the exMS / MR 70, the LISP Establish a session. This LISP session establishment process is performed when the virtual network 15 is constructed.

  Next, the data center controller 1 (17a) of the data center 1 (10a) and the data center controller 2 (17b) of the data center 2 (10b) work together, and the data center 1 (10a) to the data center 2 The migration process of the VM 40 to (10b), that is, the VM migration is started.

  Next, the data center controller 1 (17a) of the data center 1 (10a) transmits a VM migration notification to the ETR1 (16a) (step S1). This notification notifies that the EID of the VM 40 is not under the control of ETR1 (16a). The ETR1 (16a) deletes the entry related to the EID of the VM 40 (step S2). On the other hand, the data center controller 2 (17b) of the data center 2 (10b) transmits a VM migration notification to the ETR2 (16b) (step S3). This notification notifies that the EID of the VM 40 is under ETR2 (16b). The ETR2 (16b) adds an entry related to the EID of the VM 40 (step S4). Here, the above processing is performed asynchronously between the data center 1 (10a) and the data center 2 (10b).

  Next, the ETR2 (16b) of the destination data center 2 (10b) transmits the map registration to the MS function unit 2 (71b) of the exMS / MR2 (70b) (step S5). This map registration transmission requests registration of a pair of EID and its own RLOC. The exMS / MR2 (70b) updates the mapping table 2 (73b) (step S6). The synchronization function unit 2 (74b) of the exMS / MR2 (70b) differs from the synchronization function unit 1 (74a) of the exMS / MR1 (70a) of the data center 1 (10a) with the difference (updated) of the mapping table 2 (73b). Or all are transmitted and both are synchronized (step S7).

  Next, the exMS / MR1 (70a) of the data center 1 (10a) transmits a cache deletion request to the exITR1 (36a) (step S8). exITR1 (36a) deletes the old cache of EID (step S9). Similarly, the exMS / MR2 (70b) of the data center 2 (10b) transmits a cache deletion request to the exITR2 (36b) (step S10). exITR2 (36b) deletes the old cache of EID (step S11). Here, the above processing is performed asynchronously between the data center 1 (10a) and the data center 2 (10b).

  With the above processing, the VM 40 moves from the data center 1 (10a) to the data center 2 (10b). Here, prior to the VM migration between the data center 1 (10a) and the data center 2 (10b), the exITR1 (36a) of the data center 1 (10a) is assigned to the IP address of the VM 40 in the network 1 (50a) in advance. Is advertised as route information, and the exITR2 (36b) of the data center 2 (10b) also advertises the IP address of the VM 40 to the network 2 (50b) as route information. With the above processing, the packet addressed to the advertised IP address may arrive at the exITR1 (36a) of the data center 1 (10a) or the exITR2 (36b) of the data center 2 (10b). The route control is performed in the wide area L2 network 60 so as to be transferred.

  Next, a packet transfer sequence from the client terminal to the VM will be described with reference to FIGS. Here, the packet transfer sequence after the VM movement described with reference to FIG. 9 will be described. Here, at the time of packet transfer processing, route information advertised in advance by exITR1 (36a) and exITR2 (36b) is propagated to at least the network 1 (50a) and the network 2 (50b). First, a packet transfer sequence from the client terminal 1 (1a) under the network 1 (50a) to the VM 40 will be described with reference to FIG.

  First, as advance preparations prior to packet transfer processing, the EID of the VM 40 is assigned to the MS function unit (71a) of the exMS / MR1 (70a) of the data center 1 (10a) and the exMS / MR2 (70b) of the data center 2 (10b). The map is registered in the MS function part (71b) of ETR2 (16b). This map registration process is performed, for example, at a timing when the VM 40 is generated and an IP address is assigned to the VM 40.

  The client terminal 1 (1a) transmits a packet addressed to the VM to the network 1 (50a) (step S11). Here, the destination IP address / source IP address of the packet is the global IP address of the VM 40 / the IP address of the client terminal 1 (1a). The network 1 (50a) transfers the packet to the data center 1 (10a) advertising the global IP address of the VM 40 (step S12).

  Next, as an optional process, when NAT to a private address is required, exITR1 (36a) performs a NAT process (step S13).

  As an optional process, when the EID is not registered in the exITR1 (36a) of the data center 1 (10a), a map request is transmitted to the exMS / MR1 (70a) with the destination address as EID (step S14). The exMS / MR1 (70a) acquires the ELOC RLOC from the mapping table 1 (73a), and transmits a map request to the corresponding RLOC ETR2 (16b) (step S15). The ETR2 (16b) transmits a map response to the exITR1 (36a) (step S16). exITR1 (36a) caches EID / RLOC (step S17). Through the above optional processing, a communication path between the client terminal 1 (1a) and the VM 40 is established.

  Next, exITR1 (36a) encapsulates the packet received from the network 1 (50a) with LISP (step S18), and transfers the packet to ETR2 (16b) (step S19). The ETR2 (16b) decapsulates the packet (step S20) and transfers it to the VM 40 (step S21).

  Next, a packet transfer sequence from the client terminal 2 (1a) under the network 2 (50b) to the VM 40 will be described with reference to FIG.

  First, as an advance preparation prior to packet transfer processing, the EID of the VM 40 is transferred to the exMS / MR1 (70a) of the data center 1 (10a), the exMS / MR2 (70b) and the ETR2 (16b) of the data center 2 (10b). Register the map.

  The client terminal 2 (1b) transmits a packet addressed to the VM to the network 2 (50b) (step S31). Here, the destination IP address / source IP address of the packet is the global IP address of the VM 40 / the IP address of the client terminal 2 (1b). The network 2 (50b) transfers the packet to the data center 2 (10b) advertising the global IP address of the VM 40 (step S32).

  Next, as an optional process, if NAT to a private address is required, exITR2 (36b) performs a NAT process (step S33).

  As an optional process, when the EID is not registered in the exITR2 (36b) of the data center 2 (10b), a map request is transmitted to the exMS / MR2 (70b) with the destination address as EID (step S34). The exMS / MR2 (70b) acquires the ELOC RLOC from the mapping table 2 (73b), and transmits a map request to the corresponding RLOC ETR2 (16b) (step S35). The ETR2 (16b) transmits a map response to the exITR2 (36b) (step S36). exITR2 (36b) caches EID / RLOC (step S37). Through the above optional processing, a communication path between the client terminal 2 (1b) and the VM 40 is established.

  Next, exITR2 (36b) encapsulates the packet received from the network 2 (50b) with LISP (step S38), and transfers the packet to ETR2 (16b) (step S39). The ETR2 (16b) decapsulates the packet (step S40) and transfers it to the VM 40 (step S41).

  FIG. 12 illustrates a packet transfer path 81 from the client terminal 1 (1a) to the VM 40 and a packet transfer path 82 from the client terminal 2 (1b) to the VM 40.

  As described above, according to the routing control system of the present invention, when the client terminal 1 sends a packet addressed to the advertised IP address, the packet reaches the data center 10 located in the vicinity of the client terminal 1. On the other hand, in the wide area network 60, the packet of the advertisement IP address reaches the virtual machine 40 by the ID / Locator separation technology linked with the live migration of the VM. That is, the route between the virtual machine 40 in the data center 10 and the client terminal 1 under the external network 50 is optimized to the shortest without causing the trombone phenomenon.

  Although one embodiment of the present invention has been described in detail above, the present invention is not limited to this. For example, the protocols such as LISP, IP Anycast, and BGP mentioned in the above embodiment are merely examples, and the present invention can be implemented with other protocols as long as they have equivalent functions.

DESCRIPTION OF SYMBOLS 1 ... Client terminal 10 ... Data center 15 ... Virtual network 16 ... ETR
17 ... Data center controller 20 ... Server 30, 35 ... Gateway 36 ... exITR
40 ... Virtual machine (VM)
50 ... Network 60 ... Wide area L2 network 65 ... Tunnel 70 ... exMS / MR
71 ... MS function part 72 ... MR function part 73 ... Mapping table 74 ... Synchronization function part

Claims (4)

In a communication system that has multiple data centers that connect to an external network through gateways and deploy virtual machines, and that connects multiple data centers with a logical wide-area network, the route between client terminals and virtual machines via the external network A routing system for controlling,
In the wide area network, the IP address of the virtual machine is used as an ID to control the route between the gateway and the virtual machine using a set of the ID and the Locator, and the plurality of gateways respectively provide the IP of the virtual machine to the external network. Configuring the plurality of data centers to advertise addresses as routing information;
The routing system according to claim 1, wherein the external network is configured such that a packet addressed to the advertised IP address from a client terminal is forwarded to a gateway located in the vicinity of the client terminal among the plurality of gateways. The gateway of the first data center advertises the IP address of the deployed virtual machine in the first data center to the external network;
The gateway of the second data center advertises the IP address of the virtual machine to an external network before the virtual machine moves from the first data center to the second data center. The described routing system. The path control according to claim 1 or 2, wherein the data center performs path control using LISP (The Locator / ID Separation Protocol) and the gateway advertises the IP address of the virtual machine by IP Anycast. system. In a communication system that has multiple data centers that connect to an external network through gateways and deploy virtual machines, and that connects multiple data centers with a logical wide-area network, the route between client terminals and virtual machines via the external network A path method to control,
In the wide area network, the plurality of data centers use the virtual machine IP address as an ID to control a route between the gateway and the virtual machine using a set of the ID and Locator, and each of the plurality of gateways is connected to an external network. Advertise the IP address of the virtual machine as route information,
The route control method, wherein the external network performs route control so that a packet addressed to the advertised IP address from a client terminal is transferred to a gateway located in the vicinity of the client terminal among the plurality of gateways. .

Images