Currently we have a server with several functions:
domain controller, SQL server, WSUS server, file server, print server, exchange server, backup exec
And in addition we still have an old server with a custom program for timekeeping, and then another one as an antispam server.For the new setup we want to have just one physical machine running a few virtual machines.
I’m thinking of two alternative setups, which one is better and why?Alternative 1
VM1: DC, file server, print server, timekeeping, WSUS, Backup Exec VM2: SQL server VM3: Exchange server VM4: Antispamwith VM1-3 running Windows Server, and VM4 running Debian.
Hardware RAIDAlternative 2
VM1: DC, print server, timekeeping, WSUS, backup exec VM2: SQL server VM3: Exchange server VM4: Antispam VM5: File serverwith VM1-3 running Windows Server, VM4 Debian, and VM5 FreeBSD (or something else allowing usage of ZFS)
Software RAID, SAS (or maybe even SATA) controller passthrough to VM5A few points to take in mind:
Data reliability is the first consideration, I’m thinking triple mirroring.Speed, we have about 30 concurrent users, and considering our economic situation I don’t see this increasing drastically in the next few years.
Is alternative 2 more reliable with regards to data, or does the more complex setup neutralize the advantage of ZFS.
When something goes wrong, how long will it take to get things up and running again?
Cost, I think both the hardware and software requirements of both are of similar cost, but am I correct in that?
Answer
Data reliability is the first consideration, I’m thinking triple mirroring.
Consider – cough – backups.
Speed, we have about 30 concurrent users, and considering our economic
situation I don’t see this increasing drastically in the next few
years.
Says nothing. I run a server with a lot of VM’s – about 30 – and only realistically 5 users using them and i still go all SSD for performance.
When something goes wrong, how long will it take to get things up and running again?
Depends. On the weather (delivery may be delayed), availability of hardware. Realistically you are supposed to have a separate machine as reserve READY – preferably NOT on premises (in case of fire). If you can do that – possibly with mirroring – then “getting back up” may take merely minutes.
My main machine raid 6 broke some week ago – then we moved now to all SSD. Thanks to backup every 15 minutes and decent network all it took me was a night to restore from backup. You say nothing about how prepared you are. YOu can read how things went for me at http://www.trade-robots.com/blog/technical-risk-are-you-prepared-for-a-disaster – but generally: This is a LOT less about hardware you have running than about having all the proper replacements ready to go.
A comapny with 30 users should have the funds to run single redundancy on base hardware without problems.
And btw., we are super happy with the new SSD lineup. Taking a full backup leaves me with hunderds of megabytes per second streadming out on the 10g interface and latencies on teh storage are still only 1-2ms. Which means no problems working on them 😉 Exchange – and particularly – SQL can get heavy.
You also seem not to run a lot SQL. I mean, that is either a very low requirements setup, or you are dead slow. SQL virtulization goes only that far – exspecially without SSD or dedicated high end discs. Make sure you can handle the load you expect onthe sql server (and no, do not tell me about user counts – that depends on USE…. my 2 people working on our main dev sql server use a 8 core machine up with reporting and analysis…. all depends on usage here).
Attribution
Source : Link , Question Author : Ernst , Answer Author : TomTom
