Skip to main content

Exploiting Application-semantics for Self-* Management (Rx)

Problem Statement

Storage systems have evolved from being naive I/O devices to being a functionality-rich virtualization stack running on a hardware infrastructure that contains both compute and storage capabilities. Modern storage systems have the intelligence to radically simplify traditional application-oriented tasks such as provisioning, resiliency management, multi-tenancy policies, etc. Consider business continuity and disaster recovery management - IBM storage controllers (e.g., IBM's XIV controller) radically simplifies creation of an extremely large number of volume-level snapshots of the application at a fraction of the resource overheads and performance degradation an administrator might incur while doing the same task at the application-level.

The Rx project builds a platform to truly exploit the trends of modern storage architectures, server-network-storage virtualization, and cloud computing. The platform allows detailed discovery of application configuration and mapping to IT resources, as well as exploit the raw end-to-end data using advanced optimization and machine learning techniques. Rx is an ongoing effort based on years on research as well as interaction with storage and application administrators.

Selected Publications

  • Shivnath Babu, Nedyalko Borisov, Sandeep Uttamchandani, Ramani Routray, Aameek Singh, "DiaDS: Addressing the "my-problem-or-yours" syndrome with Integrated SAN and Database Diagnosis", in Proceedings of USENIX FAST, 2009.
  • Shivnath Babu, Nedyalko Borisov, Sandeep Uttamchandani, Ramani Routray, Aameek Singh, "Why Did My Query Slow Down?", in Proceedings of Conference on Innovations in Database Research (CIDR), 2009

People

Call for Collaboration

We are actively collaborating with academia in developing an application data model that can be exploited for systems management and eventually standardized. If you are an ISV or from academia and interested in complex real-world management problems, we would like to hear from you. Please contact Sandeep Uttamchandani (sandeepu@us.ibm.com).