Storage QoS (PriorityMeister)

Providing for storage QoS in dynamic heterogeneous networks and storage environments in the face of workload interference is an important and challenging problem. We continue to push the state-of-the-art in this space.

PriorityMeister is a novel QoS controller that combines per-workload priorities and rate limiting to provide tail latency QoS for shared networked storage servicing bursty workloads. The combination is important: priority enables latency differentiation, while rate-limiting controls for workload burstiness and throughput interference. PriorityMeister automatically and proactively configures the priorities and rate limits, even for networked storage that involves multiple stages, to provide superior tail latency QoS.

PriorityMeister dataflow diagram.

For earlier PDL research on storage QoS, see: Argon, and Self-* Storage.

People

FACULTY

Greg Ganger
Mor Harchol-Balter

GRAD STUDENTS

Alexey Tumanov
Timothy Zhu

INDUSTRY COLLABORATORS

Michael A. Kozuch, Intel

Publications

• PriorityMeister: Tail Latency QoS for Shared Networked Storage. Timothy Zhu, Alexey Tumanov, Michael A. Kozuch, Mor Harchol-Balter, Gregory R. Ganger. ACM Symposium on Cloud Computing 2014 (SoCC'14), Seattle, WA, Nov 2014.
  Abstract / PDF [940K]

Acknowledgements

We thank the members and companies of the PDL Consortium: Amazon, Bloomberg, Datadog, Google, Honda, Intel Corporation, IBM, Jane Street, Meta, Microsoft Research, Oracle Corporation, Pure Storage, Salesforce, Samsung Semiconductor Inc., Two Sigma, and Western Digital for their interest, insights, feedback, and support.