Quota Enforcement for High-Performance Distributed Storage Systems

Appeared in Proceedings of the 24th IEEE Conference on Mass Storage Systems and Technologies (MSST 2007).

Abstract

Storage systems manage quota to ensure that each user gets the storage they need, and that no one user can—even by accident—use up all available storage. This is diffi- cult for large, distributed systems, especially those used for high-performance computing applications, because resource allocation occurs on many nodes concurrently. We present a scheme where quota is enforced asynchronously by intelligent storage servers: storage clients contact a shared management service to get vouchers, a capability-like certificate that the clients can redeem at participating storage servers to allocate storage space. This approach produces low load on the shared management service, promotes good scaling, and allows the client to make decisions about which storage server(s) to use without communicating with the management service for further approval. Storage servers and the management service periodically reconcile voucher usage to ensure that clients do not cheat by spending the same voucher at multiple storage servers. We report on a simulation study that shows that this approach gives performance nearly as good as not enforcing quota at all, and that the load on the shared management server is remarkably low.

Publication date:
September 2007

Authors:
Kristal Pollack
Darrell D. E. Long
Richard Golding
Ralph Becker-Szendy
Benjamin C. Reed

Projects:
Secure File and Storage Systems
Ultra-Large Scale Storage

Available media

Full paper text: PDF

Bibtex entry

@inproceedings{pollack-msst07,
  author       = {Kristal Pollack and Darrell D. E. Long and Richard Golding and Ralph Becker-Szendy and Benjamin C. Reed},
  title        = {Quota Enforcement for High-Performance Distributed Storage Systems},
  booktitle    = {Proceedings of the 24th IEEE Conference on Mass Storage Systems and Technologies (MSST 2007)},
  pages        = {72-84},
  month        = sep,
  year         = {2007},
}
Last modified 15 Jan 2023