Storage bypass driver operation in a highly available computer system
A highly available computer system has a primary compute node and a secondary compute node, both of which operate to support at least one similar process using the same instructions and the same information at substantially the same time, and in the event that the secondary compute node generates a read request that is not able to be serviced by cache memory local to it, then the secondary compute node redirects the read request to the primary compute node for servicing.
1. A method for operating a prototype, highly available computer system during a time that the computer system is not running a check-point operation, comprising:
prior to an initial checkpoint operation after starting the computer system, detecting a read request by a secondary compute node comprising the highly available computer system, determining that the read request is not able to be serviced from local cache memory and servicing the read request from mass storage dedicated to the secondary compute node;
after the initial and all subsequent checkpoint operations are completed, a read request generated by the secondary compute node that not able to be serviced from local cache memory is redirected to a primary compute node comprising the highly available computer system; and
receiving, at the primary compute node, the redirected read request and servicing the read request from mass storage dedicated to it, and sending a read completion message to the secondary compute node;
wherein each of the primary and secondary compute nodes support at least one similar process using the same instructions and the same information at substantially the same time.
2. The method of claim 1 , further comprising the secondary compute node having a storage driver that operates to detect the read request and to determine that the read request is not able to be serviced from the cache memory that is local to the secondary node.
3. The method of claim 2 , wherein the storage driver receives the read request from an input/output subsystem comprising the secondary compute node.
4. The method of claim 3 , wherein the storage driver generates a read message that is used to service the read request.
5. The method of claim 1 , wherein the redirected read request is sent to a storage driver comprising the primary compute node over a messaging bus connecting the primary to the secondary compute node.
6. A method for operating a prototype highly available computer system during a time that the computer system is not running a checkpoint operation, comprising:
prior to an initial checkpoint operation after starting the computer system, detecting a read request by a secondary compute node comprising the highly available computer system, determining that the read request is not able to be serviced from cache memory that is local to the secondary compute node and servicing the read request from mass storage dedicated to the secondary compute node;
after the initial and all subsequent checkpoint operations are completed, a read request generated by the secondary compute node that is not able to be serviced from local cache memory is redirected to the primary compute node; and
receiving a response from the primary compute node indicating that the redirected read request is successfully completed;
wherein each of the primary and the secondary compute nodes support at least one similar process using the same instructions and the same information at substantially the same time.
7. The method of claim 6 , further comprising the secondary compute node having a storage driver that operates to detect the read request and to determine that the read request is not able to be serviced from cache memory that is local to the secondary node.
8. The method of claim 7 , wherein the storage driver receives the read request from an input/output subsystem comprising the secondary compute node.
9. The method of claim 6 , wherein the redirected read request is sent to a storage driver comprising the primary compute node over a messaging bus connecting the primary to the secondary compute node.
10. A prototype highly available computer system, comprising:
primary and secondary compute nodes each of which have separate mass storage systems and support at least one similar process using the same instructions and the same information at substantially the same time;
the secondary compute node having a storage driver that operates during a time that the computer system is not running a checkpoint operation and prior to an initial checkpoint operation after starting the computer system, to detect read requests, determine that a detected read request is not able to be serviced from cache memory that is local to the secondary compute node, and redirecting the read request to the primary compute node; and
the primary compute node receiving the redirected read request, servicing the read request from the mass storage dedicated to it and generating and sending a read completion message to the secondary compute node.
11. The system of claim 10 , further comprising the secondary compute node having a storage driver that operates to detect the read request and to determine that the read request is not able to be serviced from cache memory that is local to the secondary node.
12. The system of claim 11 , wherein the storage driver receives the read request from an input/output subsystem comprising the secondary compute node.
13. The system of claim 10 , wherein to a storage driver comprising the primary compute node received the redirected read request over a messaging bus connecting the primary to the secondary compute node.