Workloads Separation
Abstract:
Identifying the characteristics of a storage workload is critical for resource provisioning for metrics including performance, reliability, and utilization. Although multi-tenant systems are increasingly commonplace, characterization of multiple workloads within a single system trace is difficult because workloads are highly dynamic and typically not labeled. We show that, by converting a block I/O workload to a signal and applying blind source separation, we are able to successfully separate many application workloads.
Workloads Counting
Abstract:
Understanding how many simultaneous tenants are interacting in a shared storage system is essential for SLA satisfaction and resource provisioning. However, due to the volatility of multi-tenant system behavior, existing approaches fail to distinguish interleaved storage workloads on shared systems. We introduce CENSUS, a novel classification framework that combines time series analysis with gradient boosting to identify the number of tenants in a storage workload by projecting its trace into a high-dimensional feature representation space. We show that Census can distinguish the number interleaved workloads in a real world trace segment with an average error of 5 to 28%.
Fast Automatic Tuning
Abstract:
Automatically tuning the database systems is desirable to achieve good performance. However, even with the help of sampling and searching methods, finding the best configuration is still time-consuming and intractable while the numerous knobs have a huge space of configuration. How to get the best configuration within limited training time and computing cost is an open question. Pre-selecting the most important knobs or do meta-learning might be a possible approach.
Publications & Presentation
Si Chen, Omar Aaziz, Jeanine Cook, Avani Wildani,
Similarity Measurement for Proxy Application Fidelity
The International Conference for High Performance Computing, Networking, Storage, and Analysis (SC21) (Poster), November 2021
Si Chen, Jianqiao Liu, Avani Wildani,
Census: Counting Interleaved Functional Tenants on Shared Storage
36th International Conference on Massive Storage Systems and Technology (MSST 2020), October 2020
Si Chen, Jianqiao Liu, Avani Wildani,
Census: Counting Interleaved Functional Tenants on Shared Storage
FAST'20 Work-in-Progress (Talk + Poster), February 2020
Si Chen, Avani Wildani,
Chasing the Signal: Statistically Separating Multi-Tenant I/O Workloads
FAST'19 Work-in-Progress (Talk + Poster), February 2019
Si Chen, Avani Wildani,
Chasing the Signal: Statistically Separating Multi-Tenant I/O Workloads
ML for Systems (co-located with NeurIPS 2018), December 2018