Migrating workloads using networked attached memory
View Patent ↗A network system comprising a plurality of servers communicatively-coupled on a network, a network-attached memory coupled between a first server and a second server of the server plurality, and a memory management logic that executes on selected servers of the server plurality and migrates a virtual machine from the first server to the second server with memory for the virtual machine residing on the network-attached memory.
1. A method comprising:
changing workload data while stored in a first memory by processes executing on a first server having a first CPU;
moving said workload data to or from said first memory from or to, respectively, a second memory;
changing said workload data while stored in said second memory by processes executing on said first server;
changing said workload data while stored in said second memory by processes executing on a second server having a second CPU but not said first CPU; and
removing said workload data from said second memory to a third memory while said workload data is being changed by processes executing on said second server, said moving involving moving said workload data from said first memory to said second memory, said moving occurring while said workload data is changed by processes executing on said first server.
2. A method as recited in claim 1 wherein:
said moving involves directing memory accesses to said second memory that would have been directed to said first memory before said moving; and
said removing involves directing memory accesses to said third memory that would have been directed to said second memory before said removing.
3. A method as recited in claim 2 wherein said second memory is network-attached memory attached to said first and second servers via a network.
4. A method as recited in claim 3 wherein said network-attached memory is to serve as a paging device for said first and second servers.
5. A method as recited in claim 2 wherein said processes are virtual-machine processes, and said directings are performed by memory managers of said first and second servers.
6. A computer product comprising non-transitory computer-readable storage media encoded with code to, when executed by one or more CPUs:
change workload data while stored in a first memory by executing processes on a first server having a first CPU;
move said workload data to or from said first memory from or to, respectively, a second memory;
change said workload data while stored in said second memory by executing processes on said first server;
change said workload data while stored in said second memory by executing processes on a second server having a second CPU but not said first CPU; and
remove said workload data from said second memory to a third memory while said workload data is being changed by processes executing on said second server, the moving involving moving said workload data from said first memory to said second memory, said moving occurring while said workload data is being changed by processes executing on said first server.
7. A computer product as recited in claim 6 wherein:
the moving involves directing to said second memory memory accesses that would have been directed to said first memory before said moving; and
the removing involves directing to said third memory memory accesses that would have been directed to said second memory before said removing.
8. A computer product as recited in claim 6 wherein said second memory is network-attached memory attached to said first and second servers via a network.
9. A computer product as recited in claim 8 wherein said network-attached memory is to serve as a paging device for said first and second servers.
10. A computer product as recited in claim 6 wherein said processes are virtual-machine processes, and said directings are performed by memory managers of said first and second servers.
11. A system comprising:
a first server having a first memory and a first CPU;
a second server having a second memory and a second CPU;
a network;
network-attached memory communicatively coupled to said first and second servers via said network, said network-attached memory being to serve as a paging device for said first and second servers; and
a workload manager to migrate workload data from said first server to said second server by:
changing workload data while stored in said first memory by processes executing on said first server having said first CPU;
moving said workload data to or from said first memory from or to, respectively, said network-attached memory;
changing said workload data while stored in said network-attached memory by processes executing on said first server;
changing said workload data while stored in said network-attached memory by processes executing on said second server having said second CPU but not said first CPU; and
removing said workload data from said network-attached memory to said second memory while said workload data is being changed by processes executing on said second server, said moving involving moving said workload data from said first memory to said network-attached memory, said moving occurring while said workload data is changed by processes executing on said first server.
12. A system as recited in claim 11 wherein:
said first server has first memory management logic for directing memory accesses by said first processes to said first memory before said moving and for directing memory accesses by said first processes to said network-attached memory during said moving; and
said second server has second memory management logic for directing memory accesses by said first second processes to said network-attached memory before said removing and for directing memory accesses by said second processes to said second memory during said removing.
13. A system as recited in claim 12 wherein said first and second processes are virtual-machine processes.