Theoretical Computer Science - Special issue on dynamic and on-line algorithms
SIAM Journal on Computing
How Useful Is Old Information?
IEEE Transactions on Parallel and Distributed Systems
An Opportunity Cost Approach for Job Assignment in a Scalable Computing Cluster
IEEE Transactions on Parallel and Distributed Systems
The Power of Two Choices in Randomized Load Balancing
IEEE Transactions on Parallel and Distributed Systems
Improved Strategies for Dynamic Load Balancing
IEEE Concurrency
FOCS '02 Proceedings of the 43rd Symposium on Foundations of Computer Science
On the performance of TCP splicing for URL-aware redirection
USITS'99 Proceedings of the 2nd conference on USENIX Symposium on Internet Technologies and Systems - Volume 2
SPAND: shared passive network performance discovery
USITS'97 Proceedings of the USENIX Symposium on Internet Technologies and Systems on USENIX Symposium on Internet Technologies and Systems
On fully distributed adaptive load balancing
DSOM'07 Proceedings of the Distributed systems: operations and management 18th IFIP/IEEE international conference on Managing virtualization of networks and services
On cost-aware monitoring for self-adaptive load sharing
IEEE Journal on Selected Areas in Communications
Hi-index | 0.00 |
We consider load sharing in distributed systems where a stream of service requests arrives at a collection of n identical servers. The goal is to provide the service with the lowest possible average waiting time. This problem has been extensively studied before, but most previous models have not incorporated the monitoring costs explicitly. This paper focuses on a rigorous study of maximizing the utility of monitoring. We extend the Supermarket Model for dynamic load sharing by explicitly incorporating monitoring costs. These costs stem from the fact that the servers have to answer load queries, a task which consumes both CPU and communication resources. This Extended Supermarket Model (ESM) allows us to formally study the tradeoff between the usefulness of monitoring information and the cost of obtaining it. In particular, we prove that for each service request rate, there exists an optimal number of servers that should be monitored to obtain minimal average waiting time. Based on this theoretical analysis, we develop an autonomous load sharing scheme that adapts the number of monitored servers to the current load. We evaluate the performance of this scheme using extensive simulations. It turns out that in realistic scenarios, where monitoring costs are not negligible, the self-adaptive load balancing scheme is clearly superior to any load-oblivious load sharing mechanisms.