sponsors
usenix conference policies
Impression Store: Compressive Sensing-based Storage for Big Data Analytics
Error message
You are not authorized to post comments.Jiaxing Zhang, Ying Yan, Liang Jeff Chen, Minjie Wang, Thomas Moscibroda, and Zheng Zhang, Microsoft Research
For many big data analytics workloads, approximate results suffice. This begs the question, whether and how the underlying system architecture can take advantage of such relaxations, thereby lifting constraints inherent in today’s architectures. This position paper explores one of the possible directions. Impression Store is a distributed storage system with the abstraction of big data vectors. It aggregates updates internally and responds to the retrieval of top-K high-value entries. With proper extension, Impression Store supports various aggregations, top-K queries, outlier and major mode detection. While restricted in scope, such queries represent a substantial and important portion of many production workloads. In return, the system has unparalleled scalability; any node in the system can process any query, both reads and updates. The key technique we leverage is compressive sensing, a technique that substantially reduces the amount of active memory state, IO, and traffic volume needed to achieve such scalability.
Open Access Media
USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.
author = {Jiaxing Zhang and Ying Yan and Liang Jeff Chen and Minjie Wang and Thomas Moscibroda and Zheng Zhang},
title = {Impression Store: Compressive Sensing-based Storage for Big Data Analytics},
booktitle = {6th USENIX Workshop on Hot Topics in Cloud Computing (HotCloud 14)},
year = {2014},
address = {Philadelphia, PA},
url = {https://www.usenix.org/conference/hotcloud14/workshop-program/presentation/zhang},
publisher = {USENIX Association},
month = jun
}
connect with us