My two cents: Virtualization definitely has it’s merits; no doubt about that. There are already tons of white papers and position papers discussing this. There is a flaw in virtualization though: the virtual host is a single point of failure for more than one server. If your single-OS Exchange server’s motherboard dies, your SQL server keeps running. But if your virtual host’s motherboard dies, you lose the functionality of everything on that box. There are some excellent solutions for this to boost HA that wind up leveraging virtualization to a significant degree; that includes BDR with low SLAs (company I’m currently working with gives one hour SLA for full disaster recovery to the cloud). Depending on just how far one wants to take it one can design and build an infrastructrue that is near-indestructible using far, far less hardware than would be possible with a monolithic OS. Depending on how well such an environment is designed it’s okay for individual servers to have their MTBF hardware issues; while that one physical server is down (whether it’s monolithic or multi-guest) there are other servers capable of handling that downed machine’s load; one is not cursing some service guy to hurry up. If a tornado hits the datacenter dead center that enterprise could be back up and running in a short time period, possibly in limp mode, but operational (depending on how it’s configured). Then there is leveraging the application’s ability to provide it’s own HA. Later versions of Exchange have come a long way in addressing this. SQL with shared storage can also be configured to provide HA. Add in the HA available at below the guest OS, make sure you have a good crew of well-trained strategically-shaved monkeys and you’re set.
From: Ken Schaefer Sent: Wednesday, June 19, 2013 11:01 PM To: [email protected] Subject: RE: [NTSysADM] NAS or Server So, if the motherboard in your Exchange server failed, then your Exchange service would still be available to end users? If you put additional services on that physical host (say, your print service), and the motherboard failed, those services are still available to end-users? My understanding is that this is a “no” OK, so let’ say you put Exchange and your File Server on this physical host. You have “xyz” TB of storage. IN the case of catastrophic hardware failure (e.g. RAID controller failure that corrupts your data), you have a NBD hardware replacement warranty agreement, and it will take you 8 hours to restore all the data from tape. Your time-to-restore service (RPO or MTRS) is up to 2 business days. So, you have to ask your business – is this acceptable? If they turn around and say “no way – email and files are our most important services” then you probably need to look at: a) Faster warranty service from your vendor –or- b) Faster way to restore from backup –or- c) Split the service between multiple SPOFs, rather than concentrating on them on a single SPOF Instead, you can co-locate less important services (maybe printing, or DHCP) on the same physical host as the important ones Use business needs to drive technology decisions, so that there aren’t unpleasant conversations and/or recriminations when everything goes to cr*p. Cheers Ken From: [email protected] [mailto:[email protected]] On Behalf Of J- P Sent: Thursday, 20 June 2013 1:44 PM To: [email protected] Subject: RE: [NTSysADM] NAS or Server these are not a single physical running all guests 1 Esxi host running 2 guets - Physical Hardware a) guest one win2012 dc b) guest 2 citrix tx 2.windows 2012 dc physical (former DC/File/Print)-- Physical hardware 3. Windows 2012 hyper-V host - Physical hardware a) Windows 2012 guest running exchange 2013 3. Windows 2008 physical member server running sql 2005- physical hardware The DC's are all on separate physical hardware- Jean-Paul Natola -------------------------------------------------------------------------------- From: [email protected] To: [email protected] Subject: RE: [NTSysADM] NAS or Server Date: Thu, 20 Jun 2013 03:32:49 +0000 You don’t know what a single point of failure is? Or you don’t understand how a single physical server running multiple services is a single point of failure? Or something else? Cheers Ken From: [email protected] [mailto:[email protected]] On Behalf Of J- P Sent: Thursday, 20 June 2013 1:19 PM To: [email protected] Subject: RE: [NTSysADM] NAS or Server I'm not following the "single point of failure"
