Demand for Internet services can vary over a wide
range over many different timescales. It is not cost-effective to statically
provision the service infrastructure with enough resources to meet service
level requirements under the worst-case demand. Therefore, services that
must meet specific service level objectives must be deployed on an adaptive
service infrastructure that can allocate and dynamically adjust
the amount of resources at the right place and time to meet requirements at
minimal cost.
The adaptive infrastructure must control service placement, demand
routing and resource allocation in an integrated way to permit valuable
trade-offs across their solutions. We study the dynamic characteristics of
distributed demand to examine their implications on policies for service
placement, load balancing and resource allocation.
Papers and Reports
Presentations
|
![](http://welcome.hp-ww.com/img/s.gif) |
|