|
|
My full name is Lakshmi Narayanan Bairavasundaram.
I'm a Ph.D. student in Computer
Sciences at University
of Wisconsin-Madison. My research area is Operating systems.
I work in the ADvanced Systems
Laboratory, and my advisors are Prof. Andrea C. Arpaci-Dusseau and Prof. Remzi H. Arpaci-Dusseau.
Resume: [PDF] (Last updated: Jan 2, 2008) |
ResearchMy research interests include File and Storage Systems, Operating Systems, and Fault Tolerance. Dissertation: My dissertation focuses on the characteristics, impact and tolerance of partial disk failures. Disk drives fail in complex ways. In most of these cases, the disk does not stop working entirely; the failures tend to be partial failures, where some disk sectors are unavailable due to a latent sector error or some disk blocks are silently corrupted. The goals of my dissertation are to understand the characteristics of partial failures, analyze how these failures impact various components of the storage stack, and develop solutions to tolerate such failures. As part of my dissertation, I have analyzed the occurrence and characteristics of latent sector errors in disk drives [SIGMETRICS07]. The study found that a significant percentage (20%) of nearline disks may be affected in 2 years of use, and that latent sector errors show high spatial and temporal locality. Given that partial failures could affect a significant percentage of disks, it is important to understand their impact on different elements of the storage stack. I (along with other members of my research group) have analyzed the impact of such errors on modern commodity file systems (IBM JFS, ext3, ReiserFS and Windows NTFS) [SOSP05, StorageSS06]. We found that even widely-used file systems have bugs in failure handling code, use illogically inconsistent policies and do not implement fault tolerance techniques like type-checking and replication effectively. In addition to file systems, I have also investigated the policies and mechanisms used by current virtual memory systems (Linux, FreeBSD,Windows XP) to tolerate disk errors [DSN06] and found that virtual memory systems are equally inconsistent and ineffective at handling partial disk failures. Given that even widely-used systems have bugs and cannot be trusted to handle disk failures, it is important to lower the trust we place in any single system. I am currently developing a file system architecture that eliminates the need to trust a single file system. Other Research: My initial research experience was built on studying how different layers of a system interact and how "gray-box" techniques can be used in such layers to overcome the limitations of narrow interfaces. A specific application of such techniques has been in semantically-smart disk systems. These are disk systems that leverage basic knowledge of file system operation to provide significant improvements in performance, availability and security. I have primarily explored the performance angle. I have developed a cache mechanism for disk arrays that utilizes basic knowledge of file system data structures to provide an exclusive cache i.e. a cache that retains blocks that are not present in the file system cache [ISCA04]. I have also extended this technique to work effectively with database systems [FAST05]. Finally, I have worked on using block liveness information in a semantically-smart disk [OSDI04]. In addition to my research at Madison, I have been on research internships at IBM T. J. Watson Research Center, NY (June-Aug 2004), where I developed a distributed caching mechanism for storage systems, Intel Corporation, OR (May-Aug 2005), where I developed techniques for hypervisor-based fault injection, and Network Appliance, CA (Jun-Aug 2006, Jun-Aug 2007) where I analyzed data on storage system errors. PublicationsAn Analysis of Data Corruption in the Storage Stack Lakshmi N. Bairavasundaram, Garth R. Goodson, Bianca Schroeder, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau To appear in the Proceedings of the 6th USENIX conference on File and Storage Technologies (FAST'08) San Jose, California. February 2008. Available as: Abstract PDF Postscript BibTex Parity Lost and Parity Regained Andrew Krioukov, Lakshmi N. Bairavasundaram, Garth R. Goodson, Kiran Srinivasan, Randy Thelen, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau To appear in the Proceedings of the 6th USENIX conference on File and Storage Technologies (FAST'08) San Jose, California. February 2008. Available as: Abstract PDF Postscript BibTex An Analysis of Latent Sector Errors in Disk Drives Lakshmi N. Bairavasundaram, Garth R. Goodson, Shankar Pasupathy, Jiri Schindler. Proceedings of the International Conference on Measurement and Modeling of Computer Systems (SIGMETRICS'07). San Diego, California. June 2007. Kenneth C. Sevcik Outstanding Student Paper Award! Available as: Abstract PDF Postscript BibTex Limiting Trust in the Storage Stack Lakshmi N. Bairavasundaram, Meenali Rungta, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau Proceedings of the 2nd International Workshop on Storage Security and Survivability (StorageSS'06) Alexandria, Virgina. October 2006. Available as: Abstract PDF Postscript BibTex Dependability Analysis of Virtual Memory Systems Lakshmi N. Bairavasundaram, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau Proceedings of International Conference on Dependable Systems and Networks (DSN'06) Philadelphia, Pennsylvania. June 2006. Available as: Abstract PDF Postscript BibTex Semantically-Smart Disk Systems: Past, Present, and Future Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau, Lakshmi N. Bairavasundaram, Timothy E. Denehy, Florentina I. Popovici, Vijayan Prabhakaran, Muthian Sivathanu Sigmetrics Performance Evaluation Review (PER) Volume 33, Number 4. March 2006. Available as: Abstract PDF Postscript BibTex Database Aware Semantically-smart Storage Muthian Sivathanu, Lakshmi N. Bairavasundaram, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau Proceedings of the 4th USENIX conference on File and Storage Technologies (FAST'05) San Francisco, California. December 2005. Available as: Abstract PDF Postscript BibTex IRON File Systems Vijayan Prabhakaran, Lakshmi N. Bairavasundaram, Nitin Agrawal, Haryadi S. Gunawi, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau Proceedings of 20th ACM Symposium on Operating Systems Principles (SOSP'05) Brighton, United Kingdom. October 2005. Available as: Abstract PDF Postscript BibTex Life or Death at Block Level Muthian Sivathanu, Lakshmi N. Bairavasundaram, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau Proceedings of 6th Symposium on Operating Systems Design and Implementation (OSDI'04) San Francisco, California. December 2004. Available as: Abstract PDF Postscript BibTex X-RAY: A Non-Invasive Exclusive Caching Mechanism for RAIDs Lakshmi N. Bairavasundaram, Muthian Sivathanu, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau Proceedings of the 31st Annual International Symposium on Computer Architecture (ISCA'04) Munich, Germany. June 2004. Available as: Abstract PDF Postscript BibTex Dynamic Path Profile Aided Recompilation in a JAVA Just-In-Time Compiler R. Vinodh Kumar, B. Lakshmi Narayanan, R. Govindarajan Proceedings of the International Conference on High Performance Computing (HiPC'02) Bangalore, India. December 2002. Available as: PDF Functional Unit Usage Based Thread Selection in a Simultaneous Multithreaded Processor Deepak Babu M.I, Lakshmi Narayanan B, Madhu Saravana Sibi G, Ranjani Parthasarathi Poster Session of the International Conference on High Performance Computing (HiPC'01) Hyderabad, India. December 2001. Available as: PDF InternshipsSummer Intern. Advanced Development Group, Network Appliance. Sunnyvale, CA. June - August, 2007. I analyzed the occurrence of silent data corruption in disk drives, identifying important characteristics that would be useful for corruption-proof system design. Summer Intern. Advanced Development Group, Network Appliance. Sunnyvale, CA. June - August, 2006. I analyzed data on latent sector errors (disk errors wherein a sector becomes inaccessible), examining the dependence on factors such as disk drive age and capacity, and identifying characteristics such as spatial and temporal locality. Summer Intern. Core Virtualization Research Group, Intel Corporation. Hillsboro, OR. May - August, 2005. I developed techniques for hypervisor-based fault injection and applied these techniques to study operating system behavior when memory and disk errors occur. Summer Intern. IBM T.J. Watson Research Center. Hawthorne, NY. May - August, 2004. I designed and implemented an extensible and scalable distributed cache architecture for an enterprise storage system. My scheme involved the use of remote memory (on machines over a high-speed LAN) for caching disk blocks. Summer Research Fellow. Jawaharlal Nehru Centre for Advanced Scientific Research (JNCASR) Bangalore, India. May - June, 2001. I developed an instruction scheduler for a Java Just-in-Time compiler. The scheduler was built to utilize path profile information. |