Deployment of client data compute kernels in cloud

• US 10,362,110 B1
• Filed: 12/08/2016
• Issued: 07/23/2019
• Est. Priority Date: 12/08/2016
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Matylitski, Ivan
• Primary Examiner(s): Book, Phyllis A
• Application Number:
  US15/373,367
• Time in Patent Office: 957 Days
• CPC Class Codes: H04L 67/1097 for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
• CPC Class Codes: H04L 67/60 Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources admission control or resource allocation H04L47/70

Abstract

A data storage service includes a client data compute kernel manager that receives and registers a client data compute kernel in a client kernel repository. The client data compute kernel may be a custom data compute kernel that is received from a client. The data storage service includes a client request handler that receives data access requests to a client data store. The client request handler may determine that a data access request is associated with the client data compute kernel. The client request handler may then deploy the client data compute kernel to one or more selected storage servers, and then forward the data access request to the client data compute kernel. A storage server may execute a storage service engine of the data storage service in one execution container on the storage server and the client data compute kernel on a second execution container on the storage server.

Claims

1. A system comprising:

one or more nodes that implement a data storage service, each node of the one or more nodes comprising one or more hardware processors and memory;

a client kernel manager node of the one or more nodes, configured to;

receive a client data compute kernel executable to access a client data store maintained by a data storage service, the client data compute kernel received from a client;

store the client data compute kernel in a client kernel repository; and

register in a registry that the client data compute kernel is associated with a type of data access request;

a client request handler node of the one or more nodes, configured to, in response to receiving a data access request to access the client data store;

select a storage service node of the data storage service to service the request;

determine based at least part on the registry that the data access request is associated with the client data compute kernel;

deploy the client data compute kernel on the storage service node; and

forward the data access request to the client data compute kernel.

2. The system of claim 1, wherein the storage service node is configured to:

execute a storage service engine of the data storage service in a first execution container;

receive the data access request; and

execute the client data compute kernel to service the data access request in a second execution container.

3. The system of claim 2, wherein the storage service node is configured maintain a shared memory of at least a portion of the contents of the client data store, and the shared memory is directly accessible to the first and second execution containers.

4. The system of claim 2, wherein the client request handler node is configured to:

deploy the client data compute kernel on a plurality of storage service nodes, each storage service node of the plurality of storage service nodes associated with a shard of the client data store;

forward the data access request to instances of the client data compute kernel on the plurality of the storage service nodes;

receive a plurality of results for the data access request from the plurality of the storage service nodes; and

aggregate the plurality of results to create a response to the data access request.

5. The system of claim 1, wherein the storage service node comprises a virtual machine instance operating on a virtual machine host.

6. The system of claim 1, wherein:

the storage service node comprises a client data compute kernel cache configured to cache client data compute kernels deployed to the storage service node; and

the client request handler node is configured to;

determine whether the client data compute kernel is cached on the storage service node; and

deploy the client data compute kernel on the storage service node in response to the determination that the client data compute kernel is not cached on the storage service node.

7. A method, comprising:

performing, by a data storage service, in response to receiving a data access request to access a client data store;

determining that the data access request is associated with the client data compute kernel based at least in part on a registry of client data compute kernels, the client data compute kernel being received from a client;

deploying the client data compute kernel to one or more storage servers of the data storage service; and

executing the client data compute kernel on the one or more storage servers to service the data access request.

8. The method of claim 7, further comprising:

prior to receiving the data access request;

receiving, by the data storage service, the client data compute kernel from the client;

storing the client data compute kernel in a client kernel repository; and

registering the client data compute kernel in a registry of client data compute kernels.

9. The method of claim 7, further comprising:

caching at the client data compute kernel on the one or more storage servers after the client data compute kernel finishes servicing the data access request;

receiving a second data access request for the data storage service associated with the client data compute kernel;

obtaining a first key value associated with a repository version of a component of the client data compute kernel in the client kernel repository;

obtaining a second key value associated with a cached version of the component on the one or more storage servers;

determining that the first and second keys indicate that the repository version of the component is newer than the cached version of the component; and

in response to the determination that the repository version is newer than the cached version, redeploying the component to the one or more storage servers.

10. The method of claim 7, wherein:

deploying the client data compute kernel comprises deploying the client data compute kernel to a plurality of storage servers including the one or more storage servers, each storage server of the plurality of storage servers associated with a shard of the client data store; and

executing the client data compute kernel comprises executing the client data compute kernel on the plurality of storage servers to service the data access request.

11. The method of claim 8, wherein registering the client data compute kernel in a registry comprises:

generating a unique identifier for the client data access request; and

sending the unique identifier to the client.

12. The method of claim 8, wherein registering the client data compute kernel in a registry comprises:

storing in the registry one or more version tags corresponding one or more components of the client data compute kernel.

13. The method of claim 8, further comprising:

receiving, by the data storage service, a second client data compute kernel from the client;

executing a test on the second client data compute kernel; and

accepting or rejecting the second client data compute kernel for registration based on results of the test.

14. The method of claim 7, further comprising:

initiating a session and launching the client data compute kernel to service the data access request; and

continuing to execute the client data compute kernel during the session to service additional data access requests.

15. The method of claim 10, further comprising:

receiving a plurality of results for the data access request from the plurality of storage servers; and

aggregating the plurality of results to generate a response to the data access request.

16. The method of claim 7, wherein executing the client data compute kernel comprises making a call from the client data compute kernel to a data storage service library stored on the one or more data storage servers via an application programming interface (API) exposed by the data storage service library.

17. The method of claim 15, further comprising:

operating a plurality of execution containers on the one or more storage servers; and

wherein executing the storage service engine comprises executing the storage service engine in a first execution container of the plurality of execution containers; and

wherein executing the client data compute kernel comprises selecting a second execution container from the plurality of execution containers in which to execute the client data compute kernel.

18. A non-transitory computer-accessible storage medium storing program instructions that when executed on one or more processors cause the one or more processors to:

perform, by a data storage service, in response to receiving a data access request to access a client data store;

determine that the data access request is associated with the client data compute kernel received from a client, the client data compute kernel executable to access a client data store maintained by the data storage service;

select one or more storage servers hosting the data storage service to host;

deploy the client data compute kernel to the one or more storage servers; and

forward the data access request to the client data compute kernel.

19. The non-transitory computer-accessible storage medium of claim 18, wherein the program instructions when executed on one or more processors cause the one or more processors to:

prior to receiving the data access request;

receive the client data compute kernel from the client;

execute a test on the client data compute kernel; and

based at least in part on results of the test, register the client data compute kernel in a registry of client data compute kernels.

20. The non-transitory computer-accessible storage medium of claim 19, wherein the program instructions when executed on one or more processors cause the one or more processors to:

store in the registry one or more version tags corresponding one or more components of the client data compute kernel.