Given the ubiquity of multiserver systems, it is surprising that even at this late date so little is understood regarding their performance as compared with their single-server counterpart. While for complex single server systems, we have closed-form solutions for busy periods, response times (sojourn times), and much more, we cannot make similar statements for most two-server systems. Furthermore while the M/G/1 queue is easily analyzed under a range of scheduling and prioritization policies, for the M/G/2 queue most scheduling policies aren't even close to analytically tractable.
The intractability of multiserver systems under different scheduling policies is particularly unfortunate, since "smart" scheduling algorithms have proven extremely effective at reducing mean response time for single-server queues, and are likely to be even more effective in the multiserver regime, due to the added flexibility multiple servers offer. There is a growing body of evidence suggesting that choosing the right task assignment policy when assigning jobs to hosts in a server farm can improve mean response time by orders of magnitude. Yet many important task assignment policies are still analytically intractable. Even the simplest questions regarding multiserver systems, such as ''how does one powerful server compare with many less-powerful servers'' and ''what is the benefit of allowing cycle stealing between the servers'' remain largely open.
However recent advances in the area of multiserver systems (including moment results, tail asymptotic results for some specialized systems, methods for reducing 2-dimensionally infinite Markov chains to 1-dimensionally infinite Markov chains, time and unit-scaling techniques, and new results in scheduling theory and task assignment) lead us to believe that we are now finally in a position to solve some of these central multiserver questions. The purpose of this workshop is to bring together select researchers from around the globe to present their research and exchange thoughts, ideas, conjectures, and analytical techniques related to multiserver systems and their applications. We believe that significant progress on these difficult problems will be made through the exchange of ideas, and the collaborations borne out from this workshop.
If you plan to attend the workshop, you must REGISTER HERE . Laptop projectors, overhead projectors, and a white board will be available. Speakers with electronic presentations should email these to us by APRIL 5. Please click on EMAILING PRESENTATIONS for more information.
For directions from the Pittsburgh Airport to the Holiday Inn Hotel click here: DIRECTIONS TO HOTEL. Please save your transportation receipts and give these to Charlotte Yano for reimbursement.
For current weather in Pittsburgh: WEATHER
For a list of popular attractions in the Pittsburgh area: ACTIVITIES
Back to Mor's Home Page .