pnm@goanna.cs.rmit.OZ.AU (Paul Menon) (12/06/90)
Hi,
Not so long ago, the Time Warp and Virtual Time concepts in handling
deadlock avoidance, parallelism and even as a basis for an operating
system were hot research topics. Have they disappeared? Lost favour?
Or are the people involved real busy? David Jefferson (et al) was one
of the main researchers. I'd appreciate hearing from anyone with any
recent info in this field.
Thankyou,
Paul Menon,
Dept of Computer Science,
Royal Melbourne Institute of Technology,
124 Latrobe Street,
Melbourne 3001,
Victoria, Australia.
pnm@goanna.cs.rmit.oz.au
PH: +61 3 660 3209reiher@onyx.jpl.nasa.gov (Peter Reiher) (12/07/90)
In article <9808@darkstar.ucsc.edu> pnm@goanna.cs.rmit.OZ.AU (Paul Menon) writes: > > Not so long ago, the Time Warp and Virtual Time concepts in handling > deadlock avoidance, parallelism and even as a basis for an operating > system were hot research topics. Have they disappeared? Lost favour? > Or are the people involved real busy? I guess we're not publishing in the right place, or something. JPL has been developing the Time Warp Operating System for eight years or so, with the help of David Jefferson. In the past year, we've published 5-6 papers, mostly at parallel/distributed simulation conferences, but also at other conferences. If you don't keep up with parallel simulation, you might not have seen the most recent stuff. One high-profile place where Time Warp has recently gotten exposure was Fujimoto's article in the recent CACM surveying parallel simulation methods. There are upcoming articles from the last Principles of Distributed Computing conference and, at some point there will be an article on the Time Warp Operating System in one of the major systems journels. Basically, Time Warp has proven successful in the field of parallel simulation. Other parallel simulation methods exist, but Time Warp has certain advantages for many types of simulations that the other methods can't offer. Time Warp has also had much success in its basic goal, speeding up simulations. We've gotten speedups in excess of 40 for our system, and have gotten up to 75% of the critical path speedup for some simulations. We aren't alone. Richard Fujimoto, at Georgia Tech, is doing a lot of Time Warp related research, including special hardware for Time Warp support and database issues. Jade, a Canadian simulation company, is marketing a commercial parallel simulation system using Time Warp as a base. Rand has its own version that runs on a network of personal workstations. (A recent issue of Operating Systems Review had an article on their system.) Mitre has done some related work, as has a group at Rockwell. I recently saw a PhD dissertation by a student who wrote his own version of Time Warp for circuit simulation. Another person I know just received a PhD from Washington for his theoretical Time Warp work. He's at Bellcore now, and is still interested in Time Warp. A student of Kleinrock's at UCLA is doing queueing theory work related to Time Warp. I've probably forgotten several others, and there are many I don't know about, I'm sure. Up to this point, Virtual Time has largely been used to speed up discrete event simulations on parallel hardware. We believe it has much broader application, and hope to have the chance to demonstrate it. One potential area is distributed databases. Time Warp also seems to have fruitful possibilities for failure recovery. Jefferson is working on a Virtual Time programming language. Our project at JPL is doing work on dynamic load management, and shortly will work on checkpoint/restart and data management issues. There's lots to do, yet. Oh, by the way, the Time Warp Operating System (our version) is available for experimental use. NASA's Cosmic software distribution system can get it for any group in the U.S. There's some fee attached, which was supposed to be pretty nominal for universities. I don't know if groups outside the U.S. can get it from Cosmic or not. If you have any further questions, I'll be happy to try to answer them. Peter Reiher reiher@onyx.jpl.nasa.gov . . . cit-vax!elroy!jato!jade!reiher