The present invention relates to a network system including virtual network functions, and in particular, to control techniques of the network system.
In current communication systems, various network functions (NFs) such as broadband remote access server (BRAS), network address translation (NAT), router, firewall (FW), and deep packet inspection (DPI) are implemented by dedicated hardware (appliances). As such, when launching a new network service, a network operator is forced to introduce new dedicated hardware appliances. This requires significant costs for purchasing appliances, installation spaces, and the like. In view of such a situation, consideration is given to a technology of virtually implementing network functions implemented by hardware, by software (network function virtualization), recently (Non-Patent Literature 1). As an example of network service virtualization, Patent Literature 1 discloses a method in which a plurality of virtual routers are constructed on communication node devices, and resources of the virtual routers are dynamically distributed according to the communication quality.
Further, a technology of providing various network services by transferring a communication flow to a communication path in which a plurality of virtual network functions (VNFs) are combined is also considered (See Non-Patent Literature 2, for example).
As illustrated in
The virtual network functions VNF-1 to VNF-5 in the forwarding graph operate on general-purpose servers SV1 to SV4 in the NFV infrastructure (NFVI). By operating carrier grade functions on general-purpose servers rather than dedicated servers, it is possible to achieve cost reduction and easy operation.
However, when attempting to construct NFV by general-purpose servers, there is a case where a bottleneck occurs in CPU (central processing unit) processing of a server, communication between servers, and the like. In order to prevent such a bottleneck, it is indispensable to achieve high-speed processing of the servers. As a technology of accelerating CPU processing, in addition to an increase of the number of CPU cores, an accelerator technology of connecting a field-programmable gate array (FPGA) to a CPU has been known (for example, “Xeon+FPGA Platform for the Data Center” ISCA/CARL 2015 <http://www.ece.cmu.edu/˜calcm/carl/lib/exe/fetch.php?media=carl15-gupta.pdf>).
However, in the case of constructing NFV with use of such a server to which an FPGA is added, a VM/VNF operates not only on the CPU but also on the FPGA. Accordingly, it is necessary to manage a correspondence between the FPGA and the VM in the network. For example, it is necessary to perform update of a FPGA program depending on a VM, and perform addition/update of an FPGA according to addition/update of a VM due to update of a VM such as version upgrade or addition of a VNF.
As described above, in a network including not only CPUs of servers but also programmable logic circuits such as FPGAs as a VM/VNF infrastructure, it is necessary to have a special control system in consideration of programmable logic circuits.
In view of the above, an exemplary object of the present invention is to provide a control method, a control apparatus, a network system, and a server, for efficiently controlling a network including programmable logical circuits as a VM/VNF infrastructure.
A control apparatus according to an exemplary aspect of the present invention is a control apparatus for a network including servers on which virtual network functions operate. The control apparatus includes a storage means that stores a correspondence relation among a programmable logic circuit included in a server, a virtual machine operating on the server, and a virtual network function implemented by the virtual machine of the server; and a control means that controls at least the virtual machine and the programmable logic circuit on which the virtual machine operates, based on the correspondence relation.
A network control method according to an exemplary aspect of the present invention is a control method for a network including servers on which virtual network functions operate. The method includes, by a storage means, storing a correspondence relation among a programmable logic circuit included in a server, a virtual machine operating on the server, and a virtual network function implemented by the virtual machine of the server, and by a control means, controlling at least the virtual machine and the programmable logic circuit on which the virtual machine operates, based on the correspondence relation.
A network system according to an exemplary aspect of the present invention is a network system including servers on which virtual network functions operate. The network system includes a network in which a plurality of servers, including at least one server supporting a programmable logic circuit, are connected with each other, and a control apparatus. The control apparatus that, based on a correspondence relation among a programmable logic circuit included in a server, a virtual machine operating on the server, and a virtual network function implemented by the virtual machine of the server, controls at least the virtual machine and the programmable logic circuit on which the virtual machine operates.
According to the present invention, it is possible to efficiently control a network including programmable logic circuits as a VM/VNF infrastructure.
According to exemplary embodiments of the present invention, in a network system in which virtual network functions (VNFs) can operate on servers, a network is controlled with use of a correspondence relation between a programmable logical circuit included in a server, a virtual machine (VM) operating on the server, and a VNF implemented by the VM of the server. For example, by referring to information indicating which server supports a programmable logic circuit, what program is installed in a programmable logic circuit, and which VM and which VNF operate on which server, it is possible to identify a VM and/or a programmable logic circuit that should be updated. As described above, by using information regarding a programmable logic circuit and VM/VNF information associated therewith, it is possible to control update of a circuit configuration program of a programmable logic circuit, addition/update of a programmable logic circuit according to addition/update of a VM, and the like, allowing efficient control of a network including programmable logic circuits.
First, an exemplary system configuration for explaining respective exemplary embodiments of the present invention will be described with reference to
As illustrated in
At least one of the servers in the lower-layer network 20 is a server including a programmable logic circuit. As described below, a programmable logic circuit is a hardware circuit capable of performing programmable routine processing at a high speed, and is operable as an accelerator of a connected CPU. Further, a programmable logic circuit is able to implement a user-desired logic function in a short period of time, and also has an advantage that it is rewritable. Hereinafter, an FPGA is shown as an example of a programmable logic circuit. A server in which a CPU and an FPGA are connected with each other is called an FPGA-equipped server, and a server having no FPGA is called an FPGA-non-equipped server.
Each VNF in the upper-layer network 30 is set on a physical server of the lower-layer network 20. For example, in the system illustrated in
In
In the network system as described above, the controller 10 is able to individually perform control such as addition, setting change, and deletion of a VNF, a VM, and an FPGA in an FPGA-equipped server and an FPGA-non-equipped server, in accordance with an instruction from the management apparatus 11. While the controller 10 can collectively manage the network system as described above, it is also possible to have a configuration including controllers for respective layers such that a controller controls the upper-layer network 30 (VNF layer) and another controller controls the lower-layer network 20 (NFVI layer). Hereinafter, the controller 10 according to exemplary embodiments of the present invention will be described in detail with reference to the drawings.
A controller 10 according to a first exemplary embodiment of the present invention controls servers and switches in a network system, and performs control of FPGAs, VMs, or VNFs and path control, according to an instruction from the management apparatus 11.
Referring to
The network management unit 101 manages information of a network including servers and switches, for instance topology information. The function management unit 102 manages functions operating on respective servers, that is, virtual machines VMs and virtual network functions VNFs. For example, the function management unit 102 manages which VM is operating on which server. The FPGA management unit 103 manages an FPGA of each server. For instance, the FPGA management unit 103 manages which server is FPGA-equipped and which FPGA program is installed.
The control unit 106 controls a forwarding path of the network service based on information managed by the management units 101 to 103. For example, the control unit 106 identifies a path of the upper-layer network 30 based on function information from the function management unit 102 and topology information from the network management unit 101. The control unit 106 also identifies a correspondence relation between a VM and an FPGA to be updated, based on information related to FPGAs from the FPGA management unit 103 and function information from the function management unit 102.
Hereinafter, description will be given on a control method according to the present embodiment with reference to the system illustrated in
As illustrated in
As illustrated in
Further, as illustrated in
The controller 10 of the present embodiment can use the aforementioned management tables of the function management unit 102 and the FPGA management unit 103 to control the network/VNF/VM/FPGA.
It should be noted that in the controller 10, the functions of the network management unit 101, the function management unit 102, the FPGA management unit 103, and the control unit 105 as described below may also be realized by executing programs, stored in the program memory 107, on the CPU. Hereinafter, examples of the aforementioned VM/FPGA control will be described with reference to the drawings.
As illustrated in
Addition or deletion of a VM is also controllable similarly. For example, in the case of adding a new VM to the server X, the control unit 106 instructs the target server X to add the new VM, via the network interface 104. Meanwhile, in the case of deleting the VM2 from the server X, for example, the control unit 106 instructs the target server X to delete the VM2, via the network interface 104.
As illustrated in
Addition or deletion of an FPGA is also controllable similarly. For example, in the case of adding a new FPGA to the server X, the control unit 106 instructs the target server X to add (install) the new FPGA program, via the network interface 104. Meanwhile, in the case of deleting the FPGA program F-A1.1 from the server X, for example, the control unit 106 instructs the target server X to delete FPGA program F-A1.1, via the network interface 104.
The control for the CPU to perform the data processing of the FPGA as a substitute for the FPGA may be performed autonomously when the server X receives an FPGA program update instruction from the controller 10. Alternatively, the controller 10 may instruct the server X to cause the CPU as the substitute to perform the processing. For example, a processing substitution instruction may be included in an FPGA program update instruction from the controller 10.
As illustrated in
Hereinafter, FPGA program update control as illustrated in
As illustrated in
When the server X receives the FPGA program update instruction from the controller 10, the server X causes the CPU to autonomously perform data processing performed by the FPGA as a substitute for the FPGA. First, data processing of the VM2 performed by the FPGA is autonomously performed by the CPU, resulting in that the CPU performs data processing of VM1 and VM2 (operation S403). Then, update processing is performed with use of the update FPGA program received from the controller 10 (operation S404). When program update on the FPGA is completed, the FPGA takes over the data processing of the VM2 from the CPU and performs it on the updated program F-A1.2 (operation S405). In this way, during the time of version upgrade of the FPGA program, data processing that should be performed by the FPGA during FPGA update is continued by the CPU. Accordingly, the server X is able to update the FPGA program without interrupting the processing.
As described above, the control to cause the CPU to perform data processing performed by the FPGA as a substitute for the FPGA may be performed by the server X autonomously. Alternatively, the controller 10 may perform the control under the initiative of the controller 10 by including a processing substitution instruction in the FPGA program update instruction.
As described above, according to the network control by the present embodiment, it is possible to identify a VM and/or an FPGA to be updated, with use of a correspondence relation among an FPGA included in a server, a virtual machine (VM) operating on the server, and a VNF implemented by the VM of the server. Thereby, it is possible to perform control such as update of an FPGA program, and addition/update of an FPGA associated with the addition/update of a VM, allowing efficient control of a network including FPGAs.
While the aforementioned embodiment illustrates the case where the management apparatus 11 collectively manages a network system via the controller 10, it is also acceptable that a management apparatus in which the management apparatus 11 and the controller 10 are integrated collectively manages a network. For example, as illustrated in
Further, the present invention is not limited to collective management described above. The present invention may have a configuration in which respective layers of a multilayer system are managed by different management units in cooperation with each other.
As illustrated in
The management units 12a and 12b that manage respective layers may be configured such that individual devices communicably connected with each other perform the management operation of the respective exemplary embodiments in cooperation with each other, or that they perform the management operation under management of a host device. It is also acceptable to have a configuration in which the management units 12a and 12b that manage the respective layers or a host management unit that manages the management units 10a and 10b may be in one management apparatus while being separated functionally.
The present invention is applicable to a system in which virtual network functions (VNF) are deployed on a network.
Number | Date | Country | Kind |
---|---|---|---|
2016-070569 | Mar 2016 | JP | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/JP2017/012223 | 3/27/2017 | WO | 00 |