Linear programming without the matrix
STOC '93 Proceedings of the twenty-fifth annual ACM symposium on Theory of computing
On-line load balancing with applications to machine scheduling and virtual circuit routing
STOC '93 Proceedings of the twenty-fifth annual ACM symposium on Theory of computing
Wireless Networks
Online algorithms for selective multicast and maximal dense trees
STOC '97 Proceedings of the twenty-ninth annual ACM symposium on Theory of computing
Online througput-competitive algorithm for multicast routing and admission control
Proceedings of the ninth annual ACM-SIAM symposium on Discrete algorithms
Delayed Information and Action in On-line Algorithms
FOCS '98 Proceedings of the 39th Annual Symposium on Foundations of Computer Science
| Hi-index | 0.00 |
We study the problem of distributed online admission control and routing of permanent virtual circuits in a capacitated network. We assume that we have k distinct decision makers, each of which is responsible for gathering its own information about the state of the network. Through simulation, we demonstrate that an exponential based routing scheme will perform well in a distributed model provided granularity is sufficiently high. In order to ground these results theoretically, we prove that exponential-based schemes attain best-possible competitive ratios (same as for the centralized case) provided each edge can accommodate at least &OHgr;(k log n) requests. A matching lower-bound shows that no deterministic algorithm can attain best-possible competitive ratios without requiring the same level of granularity. In the randomized case, we present a modified exponential-based approach which obtains best-possible competitive ratios provided the granularity is at least &OHgr;(k + log n).Our results may be extended to the case where different requests have different profits, and where requests are allowed to be temporary. They also apply to admission control and scheduling for unrelated machines.