FPGA cloud platform acceleration resource allocation method, and system
View Patent ↗An FPGA cloud platform acceleration resource allocates and coordinates accelerator card resources according to delays between a host of a user and FPGA accelerator cards deployed at various network segments. Upon an FPGA usage request of the user, allocating an FPGA accelerator card in an FPGA resource pool that has a minimum delay to the host. A cloud monitoring management platform obtains_transmission delays to a virtual machine network according to different geographic locations of various FPGA cards in the FPGA resource pool, and allocating a card having a minimum delay to each user. The cloud monitoring management platform prevents unauthorized users from accessing acceleration resources in the resource pool. The invention protects FPGA accelerator cards that are not authorized for users, and ensures that the card allocated to a user has a minimum network delay, thereby optimizing acceleration performance, and improving user experience.
1. A Field Programmable Gate Array (FPGA) cloud platform acceleration resource allocation method, comprising the following steps:
S 1 : requesting to use an FPGA acceleration card by a user to initiate a resource allocation request to a cloud monitoring management platform;
S 2 : notifying a virtual machine of sending a data packet only for testing a delay in a broadcasting mode by the cloud monitoring management platform;
S 3 : identifying the data packet by an FPGA acceleration card in a resource pool, and feeding FPGA identification information back to the virtual machine which initiates the broadcasting of the data packet;
S 4 : writing, by the virtual machine, the FPGA identification information fed back into a memory cache in order, storing a feedback data packet which responds first to a first position in a queue, and feeding the feedback data packet back to the cloud monitoring management platform; and
S 5 : parsing the feedback data packet by the cloud monitoring management platform according to a protocol, acquiring information of an FPGA board having the minimum network delay with the virtual machine, generating an authorization file which is sent to the user, and then operating the FPGA board by the user according to the authorization file.
2. The FPGA cloud platform acceleration resource allocation method according to claim 1 , wherein, the FPGA identification information comprises a virtual ID and a virtual physical address.
3. The FPGA cloud platform acceleration resource allocation method according to claim 1 , wherein, when the FPGA acceleration card is in a non-idle state, the data packet which is sent by the virtual machine and is for testing the delay is discarded without response.
4. The FPGA cloud platform acceleration resource allocation method according to claim 1 , wherein, the memory cache is a First In First Out (FIFO) cache.
5. An FPGA cloud platform acceleration resource allocation system, comprising:
a resource request module, configured to request, by a user, to use an FPGA acceleration card to initiate a resource allocation request to a cloud monitoring management platform;
a test data sending module, configured to notify, by the cloud monitoring management platform, a virtual machine of sending a data packet only for testing a delay in a broadcasting mode;
an FPGA feedback module, configured to identify the data packet by an FPGA acceleration card in a resource pool and feed FPGA identification information back to the virtual machine which initiates the broadcasting of the data packet;
a minimum delay determination module, configured to write, by the virtual machine, the FPGA identification information fed back into a memory cache in order, store a feedback data packet which responds first to a first position in a queue, and feed the feedback data packet back to the cloud monitoring management platform; and
an authorization module, configured to parse the feedback data packet by the cloud monitoring management platform according to a protocol, acquire information of an FPGA board having the minimum network delay with the virtual machine, generate an authorization file which is sent to the user, and operate the FPGA board by the user according to the authorization file.
6. The FPGA cloud platform acceleration resource allocation system according to claim 5 , wherein, the FPGA identification information comprises a virtual ID and a virtual physical address.
7. The FPGA cloud platform acceleration resource allocation system according to claim 5 , wherein, when the FPGA acceleration card is in a non-idle state, the data packet which is sent by the virtual machine and is for testing the delay is discarded without response.
8. The FPGA cloud platform acceleration resource allocation system according to claim 5 , wherein, the memory cache is an FIFO cache.