[talks] S Narayana general exam

[Melissa M. Lawson]

Srinivas Narayana will present his research seminar/general exam on 
Thursday Jan 19 at 10AM in Room 402.  The members of his committee 
are:  Jen Rexford (advisor), Mung Chiang, and Moses Charikar.  
Everyone is invited to attend his talk and those faculty wishing to 
remain for the oral exam following are welcome to do so.  His 
abstract and reading list follow below.
----------------------------
Abstract:

The end-to-end performance of user-facing online services (e.g.,
Facebook, Amazon) is critical to their operation, as small increases
in user-perceived latencies can have significant impact on
revenue. Such services typically run out of a number of geographically
distributed data centers, each connected by multiple ISPs to the
Internet, and a geographically distributed set of mapping nodes directing user
requests to data centers (e.g., based on proximity). Together,
request-mapping (i.e., assigning user requests to data centers) and
response-routing decisions (i.e., choosing an ISP to send responses)
determine the wide-area paths traversed by user traffic--which are a
significant contributor to end-to-end service latency. Yet in today's
systems, request-mapping and response-routing decisions are made
almost independently with coarse-grained, if any, coordination--due to
concerns of administrative separation and operational scale.

In this work, we investigate the benefits of coordinating
request-mapping and response-routing decisions, through an
optimization framework. We identify how suboptimal performance or
costs can result if (1) the mapping and routing systems have
incompatible objectives, or (2) if they are not sharing their
decisions, and even if (3) mapping and routing systems have compatible
objectives, share decisions, and optimize on top of each other. To
overcome these difficulties, we show how to perform jointly optimized
mapping and routing decisions while retaining their administrative
separation. In particular, we construct a provably optimal distributed
algorithm implemented through local computations at mapping nodes and
data center edge routers, with a lightweight exchange of
decisions. Evaluations of our solution using traffic traces from an
operational content distribution network shows that our system
converges quickly in practice, and offers much better performance and
lower costs than today's solutions.

and reading list:

(1) The Design Philosophy of the DARPA Internet Protocols
D. D. Clark

(2) End-to-end arguments in system design
J.H. Saltzer, D.P. Reed and D.D. Clark

(3) DONAR: Decentralized server selection for cloud services
P. Wendell, J. W. Jiang, M. J. Freedman, and J. Rexford

(4) Characterizing roles of front-end servers in end-to-end
performance of dynamic content distribution
Y. Chen, S. Jain, V. K. Adhikari, and Z. L. Zhang

(5) Optimizing cost and performance for multi-homing
D. K. Goldenberg, L. Qiu, H. Xie, Y. R. Yang, and Y. Zhang

(6) Optimizing cost and performance in online service provider networks
Z. Zhang, M. Zhang, A. Greenberg, Y. C. Hu, R. Mahajan, and B. Christian

(7) Cooperative content distribution and traffic engineering in an ISP network
J. W. Jiang, R. Zhang-Shen, J. Rexford and M. Chiang

(8) Traffic engineering vs. Content distribution: A game theoretic perspective
D. DiPalantino and R. Johari

(9) A Tutorial on Decomposition Methods for Network Utility Maximization
D. P. Palomar and M. Chiang

(10) Democratizing content publication with Coral
M. J. Freedman, E. Freudenthal and D. Mazieres

(11) iPlane: An information plane for distributed services
H. V. Madhyastha, T. Isdal, M. Piatek, C. Dixon, T. Anderson, A.
Krishnamurthy and A. Venkataramani

(12) Computer networks: a systems approach
Larry L. Peterson and Bruce S. Davie