STOC '97 Proceedings of the twenty-ninth annual ACM symposium on Theory of computing
httperf—a tool for measuring web server performance
ACM SIGMETRICS Performance Evaluation Review
"Balls into Bins" - A Simple and Tight Analysis
RANDOM '98 Proceedings of the Second International Workshop on Randomization and Approximation Techniques in Computer Science
Probability and Computing: Randomized Algorithms and Probabilistic Analysis
Probability and Computing: Randomized Algorithms and Probabilistic Analysis
Theory, Volume 1, Queueing Systems
Theory, Volume 1, Queueing Systems
Ursa minor: versatile cluster-based storage
FAST'05 Proceedings of the 4th conference on USENIX Conference on File and Storage Technologies - Volume 4
Dynamo: amazon's highly available key-value store
Proceedings of twenty-first ACM SIGOPS symposium on Operating systems principles
Using Black-Box Modeling Techniques for Modern Disk Drives Service Time Simulation
ANSS-41 '08 Proceedings of the 41st Annual Simulation Symposium (anss-41 2008)
Scalaris: reliable transactional p2p key/value store
Proceedings of the 7th ACM SIGPLAN workshop on ERLANG
Informed data distribution selection in a self-predicting storage system
ICAC '06 Proceedings of the 2006 IEEE International Conference on Autonomic Computing
A Performance Model of Zoned Disk Drives with I/O Request Reordering
QEST '09 Proceedings of the 2009 Sixth International Conference on the Quantitative Evaluation of Systems
Cassandra: a decentralized structured storage system
ACM SIGOPS Operating Systems Review
Pesto: online storage performance management in virtualized datacenters
Proceedings of the 2nd ACM Symposium on Cloud Computing
Hi-index | 0.00 |
We model and evaluate the performance of a distributed key-value storage system that is part of the Spotify backend. Spotify is an on-demand music streaming service, offering low-latency access to a library of over 16 million tracks and serving over 10 million users currently. We first present a simplified model of the Spotify storage architecture, in order to make its analysis feasible. We then introduce an analytical model for the distribution of the response time, a key metric in the Spotify service. We parameterize and validate the model using measurements from two different testbed configurations and from the operational Spotify infrastructure. We find that the model is accurate---measurements are within 11% of predictions---within the range of normal load patterns. We apply the model to what-if scenarios that are essential to capacity planning and robustness engineering. The main difference between our work and related research in storage system performance is that our model provides distributions of key system metrics, while related research generally gives only expectations, which is not sufficient in our case.