Messages in this thread | | | From | Fred Douglis <> | Subject | Re: Process Migration on Linux - Impossible? | Date | Tue, 30 Sep 1997 15:48:59 -0400 |
| |
I'm not on this list, but reference.com turned up my name and pointed me at the thread about process migration.
I must take exception to Larry's comment that migration is a bad idea but remote execution is great. I agree that remote execution (parallel make etc.) is wonderul. But just because migration is difficult doesn't mean it shouldn't be done. I had a full-functioned implementation of transparent process migration in Sprite almost 10 years ago. It made the system useable for a wider range of applications than you can get with something like Condor -- which, contrary to some other messages in this thread, is not a process migration facility, it's a remote excution facility with checkpointing. Much of the extra work you need to go to in order to get that level of transparency is exactly what you need to support full migration rather than just remote execution.
By the way, the reference to my own thesis was about Zayas's work on copy-on-reference, if I understand correctly -- anyone wanting more info should check out SOSP-11 (1987).
Fred Douglis <http://www.research.att.com/~douglis/> ----------------------------------------------------------------- PGP Fingerprint: E9 F8 BF BE 52 D6 07 A6 D9 D5 28 B8 20 46 D5 1E
| |