Storage of bursty data using multiple storage tiers with heterogeneous device storage
View Patent ↗Improved techniques are provided for storing bursty data, such as checkpoints, in parallel computing environments. A parallel file system is provided comprising at least first and second storage tiers comprising respective disjoint subsets of storage; and at least one processing device configured to store burst data from a plurality of distributed processes for a given burst operation during the given burst operation on both of the at least first and second storage tiers. The given burst operation comprises a multi-phase input/output (IO) task, such as a checkpoint, having alternating periods of idle time and bursts of write and/or storage activity. A method is also provided for provisioning a parallel file system by obtaining a specification of burst data requirements for the parallel file system and determining an amount of storage required for at least first and second storage tiers comprising respective disjoint subsets of storage to satisfy the burst data requirements.
1. A parallel file system, comprising:
at least first and second storage tiers comprising respective disjoint subsets of storage; and
at least one processing device coupled to a memory and configured to store burst data from a plurality of distributed processes for a given burst operation, wherein the given burst operation comprises a multi-phase input/output (IO) task having alternating periods of idle time and bursts of one or more of write activity and storage activity, wherein the given burst operation requires a given bandwidth during said given burst operation, wherein said at least one processing device is further configured to (i) store said burst data on both of said at least first and second storage tiers at substantially the same time during said given burst operation such that a first portion of said burst data for said given burst operation is stored on said at least first storage tier and a second distinct portion of said burst data for said given burst operation is stored on said second storage tier, and (ii) migrate said first portion of said burst data from said at least first storage tier to said at least second storage tier, where said second distinct portion of said burst data is stored, substantially upon completion of said given burst operation, wherein the storage of said burst data on said at least first storage tier and said at least second storage tier, and the migration of said burst data from said at least first storage tier to said at least second storage tier is performed by a burst buffer appliance, and wherein the first portion of said burst data and the second portion of said burst data are dynamically allocated by said at least one processing device.
2. The parallel file system of claim 1 wherein the burst data comprises checkpoint data.
3. The parallel file system of claim 1 wherein the first storage tier comprises non-volatile electronic storage devices including one or more of flash, MRAM and PC-RAM.
4. The parallel file system of claim 1 wherein the second storage tier comprises disk storage devices.
5. The parallel file system of claim 1 wherein the parallel file system comprises a Lustre file system.
6. A method comprising:
obtaining burst data from a plurality of distributed processes for a given burst operation, wherein the given burst operation comprises a multi-phase input/output (IO) task having alternating periods of idle time and bursts of one or more of write activity and storage activity;
storing said burst data for a given burst operation requiring a given bandwidth during said given burst operation on at least first and second storage tiers comprising respective disjoint subsets of storage at substantially the same time such that a first portion of said burst data for said given burst operation is stored on said at least first storage tier and a second distinct portion of said burst data for said given burst operation is stored on said second storage tier; and
migrating said first portion of said burst data from said at least first storage tier to said at least second storage tier, where said second distinct portion of said burst data is stored, substantially upon completion of said given burst operation, wherein the step of storing and migrating are performed by a burst buffer appliance and wherein the first portion of said burst data and the second portion of said burst data are dynamically allocated by at least one processing device.
7. The method of claim 6 wherein the burst data comprises checkpoint data.
8. The method of claim 6 wherein the first storage tier comprises non-volatile electronic storage devices including one or more of flash, MRAM and PC-RAM.
9. The method of claim 6 wherein the second storage tier comprises disk storage devices.
10. A method for provisioning a parallel file system, comprising:
obtaining a specification of burst data requirements for said parallel file system of a given burst operation comprising a multi-phase input/output (IO) task having alternating periods of idle time and bursts of one or more of write activity and storage activity; and
determining an amount of storage required for at least first and second storage tiers comprising respective disjoint subsets of storage to satisfy said burst data requirements such that said at least first and second storage tiers can store burst data from a plurality of distributed processes for a given burst operation requiring a given bandwidth during said given burst operation at substantially the same time such that a first portion of said burst data for said given burst operation is stored on said at least first storage tier and a second distinct portion of said burst data for said given burst operation is stored on said second storage tier, wherein said first portion of said burst data is migrated from said at least first storage tier to said at least second storage tier, where said second distinct portion of said burst data is stored, substantially upon completion of said given burst operation, wherein the storage and migration are performed by a burst buffer appliance and wherein the first portion of said burst data and the second portion of said burst data are dynamically allocated by at least one processing device.
11. The method of claim 10 wherein the burst data comprises checkpoint data.
12. The method of claim 10 wherein the first storage tier comprises non-volatile electronic storage devices including one or more of flash, MRAM and PC-RAM.
13. The method of claim 10 wherein the second storage tier comprises disk storage devices.
14. The method of claim 10 further comprising the step of determining an amount of storage required for at least first and second storage tiers comprising respective disjoint subsets of storage to satisfy said migration of the burst data stored on the first storage tier to the second storage tier during a migration phase.
15. A non-transitory machine-readable recordable storage medium, wherein one or more software programs when executed by one or more processing devices implement the steps of the method of claim 6 .
16. The parallel file system of claim 1 wherein said at least one processing device is further configured to determine an amount of storage required for at least first and second storage tiers comprising respective disjoint subsets of storage to satisfy said migration of the burst data stored on the first storage tier to the second storage tier during a migration phase.
17. The method of claim 11 wherein said burst data requirements comprise one or more of a desired size, latency and interval of the checkpoints; a cost-per-Gigabyte, cost-per-bandwidth and number of checkpoints to be stored in said first storage tier; and a cost-per-Gigabyte, cost-per-bandwidth and number of checkpoints to be stored in said second storage tier.
18. The method of claim 6 further comprising the step of determining an amount of storage required for at least first and second storage tiers comprising respective disjoint subsets of storage to satisfy said migration of the burst data stored on the first storage tier to the second storage tier during a migration phase.
19. The parallel file system of claim 1 wherein said burst buffer appliance comprises a flash storage device.