Perhaps I'm missing something, but in your description I think you've just scaled out my description to more nodes. If I have an 8 node cluster with 4 active nodes and 4 passive nodes you have n+0 A/A cluster model. If any physical node of this cluster fails you now have 2 active clusters (admittedly virtualized) running on the same piece of physical hardware. One of (many) problems with A/A clusters was customers were poor at capacity planning and they often had 2 A nodes running at > 50% capacity. When failover occurred the single node remaining physical node was at or beyond capacity and performance suffered significantly.
Would it be fair to say you're trying to get some of the benefits of scale-out for free in a scale-up model? I'm not sure that free doesn't have a cost. One of the costs is certainly that in the spectrum of supportability you've moved further off the center line than the configuration was previously. What other costs there might be are mostly hypothetical speculation at this point. If someone wants to send me into a lab with a machete and a 12 pack I can provide some more practical insights. > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:bounce- > [EMAIL PROTECTED] On Behalf Of Alex Alborzfard > Posted At: Thursday, August 31, 2006 9:23 AM > Posted To: swynk > Conversation: E2k3 cluster question (general) > Subject: RE: E2k3 cluster question (general) > > Chris, I completely agree with you about the perils of running 2 nodes > on a single physical box, but what do you think of the following: > > If I understand MS's best practice for EXCH clustering, they support > scale-up & -out models. With scale-up you need to build both A & P > nodes for each EXCH cluster. With scale-out you have multiple A nodes & > 1 P node in the cluster. > Now with VM(ware) you're supposedly can have same HA as the scale-up > model with fewer servers than the scale-out. Instead of the MS's scale- > out, with VM you could use 2 servers each containing both A and P > nodes. This works out better because it doesn't rely on a single P node > to cover multiple A nodes in a cluster. So you could build an 8-node > cluster using only 4 servers, with each server, running 2 separate > clusters. > > I'm not sure if it would be supported by MS, but would it make sense to > run an EXCH cluster like that? > > Alex > > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of > Chris Scharff > Sent: Wednesday, August 30, 2006 4:57 PM > To: Exchange Discussions > Subject: RE: E2k3 cluster question (general) > > The purpose of a cluster is ostensibly to provide high availability. As > such it's my belief that one needs to understand the boundaries of > supportability and best practices for a cluster and draw a line as > close to the center as possible. > > So in a scenario where one had a 4 node cluster where each of 2 nodes > exists on a single physical machine it would seem to me that you have a > pretty basic design flaw. If you are running an N+1 cluster you lack > sufficient nodes per physical host to run 3 nodes in the event of a > single physical hardware failure. If you're running N+2 you now have > all the complexity of a 4 Node system with all the benefits[1] of an > A/A cluster. > > A FE or connector Exchange server on virtualized HW seems like a > perfectly reasonable deployment scenario for many organizations. I am > not convinced that deploying a mailbox server on same is reasonable and > even less convinced it increases availability. I guess it really > depends on what problem you're trying to solve and I didn't see any > evidence of a problem or a solution in the OP comments. > > [1] Do I really need to clarify that "benefits" is tongue in cheek when > discussing A/A clusters? > > > -----Original Message----- > > From: [EMAIL PROTECTED] [mailto:bounce- > > [EMAIL PROTECTED] On Behalf Of Alex Alborzfard > > Posted At: Wednesday, August 30, 2006 3:01 PM Posted To: swynk > > Conversation: E2k3 cluster question (general) > > Subject: RE: E2k3 cluster question (general) > > > > Dare I ask: what is specifically wrong with virtualizing Exchange or > is > > it virtualzing an Exchange cluster? What are the major issues and > > pitfalls? > > > > Alex > > > > -----Original Message----- > > From: [EMAIL PROTECTED] > > [mailto:[EMAIL PROTECTED] On Behalf Of > > Chris Scharff > > Sent: Wednesday, August 30, 2006 3:35 PM > > To: Exchange Discussions > > Subject: RE: E2k3 cluster question (general) > > > > In a lab sire. In production.... can't think of a reason other than > > hating your job and wanting a new one. > > > > > -----Original Message----- > > > From: [EMAIL PROTECTED] [mailto:bounce- > > > [EMAIL PROTECTED] On Behalf Of Wehner, Paul > > > (wehnerpl) > > > Posted At: Wednesday, August 30, 2006 2:22 PM Posted To: swynk > > > Conversation: E2k3 cluster question (general) > > > Subject: RE: E2k3 cluster question (general) > > > > > > I'm going to assemble extra ips, quorum, store volumes and give it > a > > > shot. > > > Might be moderately useful to have two virtual servers on a two > node > > > cluster. > > > > > > -----Original Message----- > > > From: [EMAIL PROTECTED] > > > [mailto:[EMAIL PROTECTED] On Behalf Of > Ed > > > Crowley [MVP] > > > Sent: Wednesday, August 30, 2006 2:04 PM > > > To: Exchange Discussions > > > Subject: RE: E2k3 cluster question (general) > > > > > > I suppose you could install Microsoft Virtual Server and run the > new > > > Exchange servers in virtual machines, but I don't recommend it for > a > > > production environment. > > > > > > Ed Crowley MCSE+Internet MVP > > > Freelance E-Mail Philosopher > > > Protecting the world from PSTs and Bricked Backups!T > > > > > > -----Original Message----- > > > From: [EMAIL PROTECTED] > > > [mailto:[EMAIL PROTECTED] On Behalf Of > > > Wehner, Paul (wehnerpl) > > > Sent: Wednesday, August 30, 2006 10:34 AM > > > To: Exchange Discussions > > > Subject: E2k3 cluster question (general) > > > > > > > > > I have a two node active/passive exchange cluster working fine. > > > Is it possible to add a second exchange cluster to these two boxes? > > > (assuming new ips, different store volumes, etc) Thanks, Paul > > > _________________________________________________________________ > List posting FAQ: http://www.swinc.com/resource/exch_faq.htm > Web Interface: http://intm-dl.sparklist.com/read/?forum=exchange > To subscribe: http://e-newsletters.internet.com/discussionlists.html/ > To unsubscribe send a blank email to > [EMAIL PROTECTED] > Exchange List admin: [EMAIL PROTECTED] > To unsubscribe via postal mail, please contact us at: > Jupitermedia Corp. > Attn: Discussion List Management > 475 Park Avenue South > New York, NY 10016 > > Please include the email address which you have been contacted with. > > > _________________________________________________________________ > List posting FAQ: http://www.swinc.com/resource/exch_faq.htm > Web Interface: http://intm-dl.sparklist.com/read/?forum=exchange > To subscribe: http://e-newsletters.internet.com/discussionlists.html/ > To unsubscribe send a blank email to [EMAIL PROTECTED] > dl.sparklist.com > Exchange List admin: [EMAIL PROTECTED] > To unsubscribe via postal mail, please contact us at: > Jupitermedia Corp. > Attn: Discussion List Management > 475 Park Avenue South > New York, NY 10016 > > Please include the email address which you have been contacted with. _________________________________________________________________ List posting FAQ: http://www.swinc.com/resource/exch_faq.htm Web Interface: http://intm-dl.sparklist.com/read/?forum=exchange To subscribe: http://e-newsletters.internet.com/discussionlists.html/ To unsubscribe send a blank email to [EMAIL PROTECTED] Exchange List admin: [EMAIL PROTECTED] To unsubscribe via postal mail, please contact us at: Jupitermedia Corp. Attn: Discussion List Management 475 Park Avenue South New York, NY 10016 Please include the email address which you have been contacted with.
