Accelerator Loading Method, System, and Apparatus

Abstract

An accelerator loading apparatus obtains an acceleration requirement, where the acceleration requirement includes an acceleration function and acceleration performance of a to-be-created virtual machine, determines an image that meets the acceleration function and the acceleration performance, and determines a target host in which an available accelerator that can load the image is located, and then sends an image loading command to the target host. The image loading command includes a descriptor of the image, and is used to enable the target host to load the image for the available accelerator. In the method, a target host that can create the virtual machine may be determined based on the acceleration function and the acceleration performance of the to-be-created virtual machine, and an image used for acceleration is loaded to an available accelerator of the target host, to implement dynamic accelerator loading and deployment.

1 . An accelerator loading method implemented by an accelerator loading management apparatus, wherein the accelerator loading method comprises:
 obtaining an acceleration requirement comprising a first acceleration function and a first acceleration performance of a to-be-created virtual machine;   searching an accelerator image repository based on the first acceleration function and the first acceleration performance to determine a first image that meets the acceleration requirement and to obtain a first descriptor and a first accelerator model of the first image, wherein the first descriptor is a first identifier or a first location of the first image, wherein the accelerator image repository comprises images and image information, and wherein the image information comprises a descriptor, an accelerator model, an acceleration function, and acceleration performance of each of the images; and   sending, to a target host comprising a first available accelerator that can load the first image, an image loading command comprising the first descriptor and a second identifier of the first available accelerator,   wherein the image loading command is configured to enable the target host to load the first image for the first available accelerator based on the image loading command.   
     
     
         2 . The accelerator loading method of  claim 1 , further comprising determining the target host according to the first accelerator model. 
     
     
         3 . The accelerator loading method of  claim 2 , further comprising:
 determining candidate hosts corresponding to the first accelerator model;   determining an available accelerator of each of the candidate hosts; and   selecting a first candidate host of the candidate hosts as the target host.   
     
     
         4 . The accelerator loading method of  claim 3 , further comprising:
 obtaining a host list corresponding to the acceleration requirement, wherein the host list comprises identifiers of hosts that meet the acceleration requirement; and   searching an acceleration database based on the first accelerator model and the host list to determine the candidate hosts corresponding to the first accelerator model and the first available accelerator.   
     
     
         5 . The accelerator loading method of  claim 4 , further comprising:
 obtaining a request for creating a virtual machine, wherein the request comprises a computing resource requirement of the to-be-created virtual machine, a storage resource requirement of the to-be-created virtual machine, and the acceleration requirement;   searching all hosts managed by the accelerator loading management apparatus for hosts that meet the computing resource requirement and the storage resource requirement; and   generating the host list based on the identifiers.   
     
     
         6 . The accelerator loading method of  claim 4 , wherein the acceleration database comprises acceleration information of each host managed by the accelerator loading management apparatus, and wherein the acceleration information of each host comprises an identifier of the host, an identifier of each accelerator of the host, an accelerator model of the accelerator, an acceleration function of the accelerator, acceleration performance corresponding to the acceleration function of the accelerator, an attribute of the accelerator, and a status of the accelerator. 
     
     
         7 . The accelerator loading method of  claim 3 , further comprising further selecting the first candidate host as the target host based on a host filtering rule. 
     
     
         8 . The accelerator loading method of  claim 1 , further comprising:
 receiving from the target host an image request comprising the first descriptor;   obtaining the first image based on the first descriptor; and   sending the first image to the target host.   
     
     
         9 . The accelerator loading method of  claim 8 , further comprising:
 receiving from the target host a notification message comprising a third identifier of the target host, the second identifier, the first acceleration function, and the first acceleration performance;   updating an acceleration database based on the first acceleration function and the first acceleration performance; and   changing a status of the first available accelerator to locked.   
     
     
         10 . An accelerator loading management apparatus comprising:
 a communications interface;   a memory configured to store an accelerator image repository, wherein the accelerator image repository is configured to store images and image information, wherein the image information comprises a descriptor, an accelerator model, an acceleration function, and acceleration performance of each of the images; and   a processor coupled to the communications interface and the memory and configured to:
 obtain an acceleration requirement comprising a first acceleration function and a first acceleration performance of a to-be-created virtual machine; 
 search the accelerator image repository based on the first acceleration function and the first acceleration performance to determine a first image that meets the acceleration requirement and to obtain a first descriptor and a first accelerator model of the first image, 
   wherein the first descriptor is a first identifier or a first location of the first image; and
 send, through the communications interface to a target host comprising a first available accelerator that can load the first image, an image loading command comprising the first descriptor and a second identifier of the first available accelerator, 
 wherein the image loading command is configured to enable the target host to load the first image for the first available accelerator based on the image loading command. 
   
     
     
         11 . The accelerator loading management apparatus of  claim 10 , wherein the processor is further configured to determine the target host according to the first accelerator model. 
     
     
         12 . The accelerator loading management apparatus of  claim 11 , wherein the processor is further configured to:
 determine candidate hosts corresponding to the first accelerator model;   determine an available accelerator of each of the candidate hosts; and   select a first candidate host of the candidate hosts as the target host.   
     
     
         13 . The accelerator loading management apparatus of  claim 12 , wherein the memory further comprises an acceleration database configured to store acceleration information of each host managed by the accelerator loading management apparatus, wherein the acceleration information of each host comprises an identifier of the host, an identifier of each accelerator of the host, an accelerator model of the accelerator, an acceleration function of the accelerator, acceleration performance corresponding to the acceleration function of the accelerator, an attribute of the accelerator, and a status of the accelerator, and wherein the processor is further configured to:
 obtain a host list corresponding to the acceleration requirement, wherein the host list comprises identifiers of hosts that meet the acceleration requirement; and   search the acceleration database based on the first accelerator model and the host list to determine the candidate hosts corresponding to the first accelerator model and the first available accelerator.   
     
     
         14 . The accelerator loading management apparatus of  claim 13 , wherein the processor is further configured to:
 obtain a request for creating the to-be-created virtual machine, wherein the request comprises a computing resource requirement of the to-be-created virtual machine, a storage resource requirement of the to-be-created virtual machine, and the acceleration requirement;   search all hosts managed by the accelerator loading management apparatus for hosts that meet the computing resource requirement and the storage resource requirement; and   generate the host list based on the identifiers.   
     
     
         15 . The accelerator loading management apparatus of  claim 13 , wherein the acceleration requirement further comprises a computing resource requirement of the to-be-created virtual machine and a storage resource requirement of the to-be-created virtual machine, and wherein the processor is further configured to:
 search the accelerator loading management apparatus for hosts that meet the computing resource requirement and the storage resource requirement, and   generate the host list based on the identifiers.   
     
     
         16 . The accelerator loading management apparatus of  claim 13 , wherein the acceleration database is configured to store acceleration information of each host managed by the accelerator loading management apparatus, and wherein the acceleration information of each host comprises an identifier of the host, an identifier of each accelerator of the host, an accelerator model of the accelerator, an acceleration function of the accelerator, acceleration performance corresponding to the acceleration function of the accelerator, an attribute of the accelerator, and a status of the accelerator. 
     
     
         17 . The accelerator loading management apparatus of  claim 12 , wherein the acceleration requirement further comprises a host filtering rule, and wherein the processor is further configured to further select the first candidate host as the target host based on the host filtering rule. 
     
     
         18 . The accelerator loading management apparatus of  claim 10 , wherein the processor is further configured to send the image loading command to an acceleration agent of the target host, and wherein the image loading command is configured to enable the acceleration agent to load the first image for an accelerator driver corresponding to the first available accelerator. 
     
     
         19 . The accelerator loading management apparatus of  claim 18 , wherein the processor is further configured to:
 receive, from the target host through the communications interface, an image request comprising the first descriptor;   obtain the first image based on the first descriptor; and   send the first image to the target host through the communications interface.   
     
     
         20 . The accelerator loading management apparatus of  claim 18 , wherein the processor is further configured to:
 receive, from the target host through the communications interface, a notification message comprising a third identifier of the target host, the second identifier, the first acceleration function, and the first acceleration performance;   update an acceleration database based on the first acceleration function and the first acceleration performance; and   change a status of the first available accelerator to locked.   
     
     
         21 . An accelerator loading system comprising:
 a plurality of hosts comprising a target host; and   an accelerator loading management apparatus connected to the hosts and comprising a memory comprising an accelerator image repository, wherein the accelerator image repository is configured to store images and image information, wherein the image information comprises a descriptor, an accelerator model, an acceleration function, and acceleration performance of each of the images, and wherein the accelerator loading management apparatus is configured to:
 obtain an acceleration requirement comprising a first acceleration function and a first acceleration performance of a to-be-created virtual machine; 
 search the accelerator image repository based on the first acceleration function and the first acceleration performance to determine a first image that meets the acceleration requirement and to obtain a first descriptor and a first accelerator model of the first image, wherein the first descriptor is a first identifier or a first location of the first image; and 
 send to the target host an image loading command comprising the first descriptor, 
   wherein the target host comprises a first available accelerator that can load the first image and is configured to:
 receive the image loading command; 
 obtain the first image based on the first descriptor; and 
 load the first image for the first available accelerator. 
   
     
     
         22 . The accelerator loading system of  claim 21 , wherein the accelerator loading management apparatus is further configured to determine the target host according to the first accelerator model. 
     
     
         23 . The accelerator loading system of  claim 21 , wherein the accelerator loading management apparatus is further configured to:
 determine candidate hosts corresponding to the first accelerator model;   determine an available accelerator of each of the candidate hosts; and   select a first candidate host of the candidate hosts as the target host.