- From: Mark Baker <distobj@acm.org>
- Date: Fri, 29 Oct 2004 16:30:28 -0400
- To: "Mullins, Chalon" <Chalon.Mullins@schwab.com>
- Cc: 'Savas Parastatidis' <Savas.Parastatidis@newcastle.ac.uk>, www-ws@w3.org, Ian Foster <foster@mcs.anl.gov>, Carl Kesselman <carl@ISI.EDU>, Steve Graham <sggraham@us.ibm.com>, Steve Tuecke <tuecke@mcs.anl.gov>
Hi again, On Fri, Oct 29, 2004 at 12:50:07PM -0700, Mullins, Chalon wrote: > <MB> In my experience with scaling Web servers, what gets in the way is > exactly the mechanism you describe there; "storing it on the back end, > and passing information about how to get at it". By doing that, you > prevent the server from freeing up resources as it needs, and end up > requiring hacks like distributed garbage collection, and/or just > throwing more and more server resources at the problem. > > <CM> Exactly! We get the benefit that we can scale our applications by > adding or taking away servers without impacting clients, but, as things > stand, we have to take care of these issues ourselves. Ok, sure, but I claim that stateful apps have to upgrade much more often since they're maintaining state which isn't present for stateless apps (since it's all in the messages). > I would draw a > different conclusion, though: this is exactly why a standard way of > handling the inherent statefulness of the problem in a stateless protocol is > key -- then the server can understand what's going and free up resources, > etc. Not surprising that the Grid Community, which runs up against this > problem all the time, would be the one pushing the standard solution. Well, I personally feel that the Grid community has made a mistake by so fully embracing stateful comms with WS-RF. YMMV, of course. 8-) > <MB> I think you'd be surprised what you could handle. Most of the > interesting work on scalable services over the past 10 years or so > has been done with Web servers. > > <CM> Interesting claim, but I'm not sure what backs it up. I think you're > dismissing some really fascinating things that have been done, for example, > in the mainframe space. BTW -- we are always testing the limits of what we > can handle. No, I'm not too familiar with mainframes. I should have stated that most of the *published* work on scalability of network based services has been regarding Web server scalability; "interesting" is, of course, subjective. 8-) But if you had any pointers handy to the work you're referring to, I'd love to have a look. TIA! > <MB> There could very well be some scenarios for which that doesn't make > sense, but in my experience stateless should always be your first choice > unless there's a darned good reason not to use it. > > <CM> Well, I agree about going "stateless," but as the dialog that started > this point explained, I think we have differences over just what that means. > As far as whether we really do have a scenario that doesn't fit, I'm > comfortable that we have. And we may just have to disagree on that point. Yes, it seems so! 8-) Cheers, Mark. -- Mark Baker. Ottawa, Ontario, CANADA. http://www.markbaker.ca
Received on Friday, 29 October 2004 20:28:43 UTC