MUTHIAN SIVATHANU
|
|
MUTHIAN SIVATHANU |
)
Hi! I graduated with a Ph.D. from the University of Wisconsin Madison in May 2005. I now work at Google Inc. A copy of my Ph.D. dissertation on "Semantically-smart Disk Systems" can be found here.
My old job application materials are still available online.
| Research |
My research interests span the broad boundaries of operating systems and distributed systems.
For my Ph.D., I worked with Prof. Andrea Arpaci-Dusseau and Prof. Remzi Arpaci-Dusseau .
My thesis research proposed and evaluated a new class of storage systems that are aware of how the file system or DBMS is using them, and exploits this knowledge to implement new optimizations and functionality that are currently impossible to implement. A straightforward way to making storage systems semantically-aware is to change the interface to storage to convey richer information; however, modifying an interface as basic as SCSI is not very pragmatic, since it requires broad industry consensus and raises various legacy issues. In contrast, semantically-smart storage systems automatically track higher level semantic information about the file system or DBMS, by careful observation of block level reads and writes from underneath an unmodified SCSI interface. By requiring no changes to the existing interface, semantically-smart disks enable seamless deployment and adoption of new functionality. I demonstrated the utility of semantic knowledge within a storage system by design and prototyping of a variety of new functionality aimed at improving the availability, security, and performance of storage systems. In each of these cases, semantic knowledge was shown to enable functionality that is otherwise impossible to provide in the storage hierarchy. More details are available in my research statement.
Earlier to this, I worked on Scriptable RPC, a framework that enables distributed system services to take advantage of the increasing level of intelligence in system components (be it disk, network interface or memory), resulting in extensible, robust and high-performance distributed systems.
In the past, I have also worked on a host of projects including distributed file systems for network attached storage, distributed RAID algorithms over network storage and security and anonymity architectures for network storage. Some of my old projects can be found here.
| Publications |
Journal Publications
Conference Publications
Technical Reports
| Selected Talks |
March 2004: "Improving Storage System Availability with
D-GRAID", FAST 2004, San Francisco.
August 2003: "Improving Storage System Availability with D-GRAID", HP
Labs, Palo Alto (Invited Talk)
August 2003: "Improving Storage System Availability with D-GRAID", IBM Research, Almaden(Invited Talk)
August 2003: "Performance Isolation on Google Servers", Google Inc.(Intern Talk)
October 2002: "Evolving RPC for Active Storage", ASPLOS 2002, San Jose
August 2002: "Evolving RPC for Active Storage", Hewlett Packard
Labs, Palo Alto (Invited Talk)
August 2002: "Scalable Resource Management for Wide-Area Storage",
IBM Almaden (Intern Talk)
August 2001: "Scalable Fault-tolerant Replication in Wide-Area Storage"
, HP labs, Palo Alto (Intern Talk)
June 2001: "The WiND Filesystem",
HP labs, Palo Alto (Intern Intro Talk)
| Internships |
Summer 2004
Microsoft , Redmond, WA
(Advanced Operating Systems group)
Details confidential.
Summer 2003
Google Inc. , Mountain View
(Infrastructure group)
Designed and implemented mechanisms for effective performance
isolation on Google
servers.
Summer 2002
IBM Almaden Research
Center , San Jose ( Storage
Systems group )
Designed an architecture for scalable, decentralized resource
management
in wide-area storage. The architecture, targeted at wide-area
storage service providers, facilitates, inter alia, optimal
replication of competing client virtual disks, both in terms of
the number of replicas and their placement across the globe.
The policies ensure that individual disks take mostly autonomous decisions,
while still working towards globally optimal resource allocation.
Summer 2001
Hewlett Packard Labs, Palo Alto
(Storage Systems
program )
Designed a protocol for scalable, fault-tolerant wide area replication
that permits concurrent conflicting updates from all replicas in a geographically
distributed storage system. The work resulted in a patent disclosure.
A prototype system was built with an iSCSI device driver at the client
talking to multiple disks interacting through rpc.
| Resume |
| About Me |
For the customary two-line history info, I was born in 1979 in Chennai (aka Madras), India. I did my schooling in St Johns MHSS, Chennai. A four-year stint at Anna University yielded me a B.E. in Computer Science and Engineering. Since August 2000, I have been a graduate student here at UW.
Besides Computer Science, which is my primary interest, I am interested in Law and Business, mainly from the influence of my father Mr M Sivathanu, a lawyer in India. This also partly explains my interest in choosing Business as my Ph.D. minor. My brothers Gopalan Sivathanu and Sankaran Sivathanu are also computer scientists.
I have been taking quite a few fun courses during my stay at UW, including Ice Skating, Swimming and Tennis. In addition, I also play badminton in my spare time.
My family is involved in some charity activities to the poor. Here is a page describing Sri Sivathanu Charitable Trust founded by my father to carry out some of this charity.
I like collecting quotes; Here is a collection of quotes I like.