Colloquia: Arun Iyengar/IBM TJ Watson Research Center Efficient Techniques for Serving Dynamic Web Data with Scalability and Reliability in ACES 6.304

Contact Name: 
Jenna Whitney
Jul 6, 2006 3:00pm - 4:00pm

Speaker Name/Affiliation: Arun Iyengar/IBM TJ Wat

son Research Center

Talk Title: Efficient Techniques for Serving Dy

namic Web Data with Scalability and Reliability

Date/Time: July 6

2006 at 3:00 p.m.

Coffee: 2:45 p.m.

Location: ACES 6.304

Host: Mike Dahlin

Talk Abstract:
This talk will present tech

niques for improving performance
and reliability of commercial applicat

ions which typically
have Web front ends. While Web servers can general

ly sustain
high throughputs for serving static content many of the
real challenges occur in serving dynamic data requiring
significant ba

ck-end processing. The talk will show how
commercial systems we have b

uilt have achieved good performance
under very high loads. One importa

nt technique we have
used is to break up Web pages into fragments based
on lifetimes
and personalization characteristics. This can increase <

br>the amount of data which can be cached. I will present
techniques t

hat we have developed for automatically detecting
fragments in Web page


Intelligent caching and load balancing are key components
improving performance and geographic distribution of
servers improves
availability and can also reduce latency.
The trade-offs between repli

cating and partitioning data
across multiple servers will be discussed.
The talk will
also describe techniques we have developed for predicti

the performance of Web applications which have been successfully
used for capacity planning for several major customers.

A number of
the techniques presented in the talk are applicable
to distributed com

mercial applications in general and not
just those with Web front ends.
I will describe some of
the open problems that we are currently worki

ng on and suggest
good research problems in these areas. I will also i

some research areas in Web performance which are likely

have less impact.

Speaker Bio:
Arun Iyengar does research and dev

elopment into Web performance
and capacity planning caching storage a

llocation and
distributed computing at IBM''s T.J. Watson Research Cen

His work has been incorporated into products by several

ies including IBM HP and Microsoft and has improved
performance at ma

ny highly accessed Web sites serving considerable
dynamic content. Aru

n is Co-Editor-in-Chief of the ACM
Transactions on the Web Chair of th

e IEEE Computer Society''s
Technical Committee on the Internet the Cha

ir of IFIP Working
Group 6.4 on Internet Applications Engineering and

Co-Chair for the 2006 International World Wide Web Conference.

He is also an IBM Master Inventor. He has PhD and Masters
both in Computer Science from the Massachusetts
Institute of Technolo

gy where he was a National Science
Foundation Graduate Fellow and a BA

Summa Cum Laude and
with Distinction from the University of Pennsylvani