We have both older IBM blades and newer HP blades to hold our virtualized environment. Roughly 150 + virtual servers. All SQL servers are true iron with 4 16 core CPUs and 256 gig of ram. Sans maintain both virtual systems as well as SQL luns.
Our licenses are probably more than the hardware. We move resources to help any app server if it is under too high a load.
On Wed, Apr 13, 2016 at 10:55 AM, Dave Crozier DaveC@flexipol.co.uk wrote:
Ken, Seven Virtualised servers on one box take a fairly beefy machine to give acceptable performance and certainly not a desktop type machine. We started with about 20 physical servers and we rationalised first (amalgamating servers where possible) and now have those resulting 7 VM's virtualised on one HP box using Hyper-V. The VM's run a mixture of server 2012 (hosting SQL 2012) and server 2008 hosting the dual exchange boxes. Probably your biggest decision is whether to add a SAN into the mix to hold your data which is cost effective where you need lots of data store.
Performance is excellent and the big advantage is that you can easily spin up a new server in minutes or clone an existing one very easily.
Biggest problem now is the reliability of the RDS/RDP server which controls all the Terminal Server sessions which seems to have a mind of it's own when it comes to going wrong. The problem seems to be with Microsoft RDS itself but various patches from them have improved the reliability.
All in all the transfer to VM status was fairly painless and we did it over two weekends.
In addition we have now installed a 3 cluster system with one of the clusters permanently doing backups of the live VM's using Veeam so we can, if we wish go back incrementally to any moment in time we wish on any server - Veeam is really good and we are impressed.
So, my advice, invest in one good hardware box, bags of memory and processing power and do it. If your hardware maintenance contract is good (HP are brilliant in their same day business hardware support) you won't look back.
Allocation of resources to the VM's once they are up and running is a breeze and you can really allocate processing power and memory where it is needed most. You also don't need clustering, but it adds a nice warm feeling - and really works well.
Dave -----Original Message----- From: ProFox [mailto:profox-bounces@leafe.com] On Behalf Of Ken Dibble Sent: 13 April 2016 15:57 To: profox@leafe.com Subject: [NF] Your Experience with Virtualized Networks
Have any of you had experience working with (rather) large virtualized networks?
It's been suggested that I virtualize my network. To me the advantages are not clear but the risks are. Since I am old-school and highly risk-averse when it comes to computer technology, I need to hear different perspectives from people who are not trying to sell me anything.
We have 7 servers that are candidates for virtualization on a single hypervisor.
- Domain controller (currently Linux but may be replaced with Windows)
- Heavily used Linux file server that includes two VFP databases and
serves up a large and growing number of network shares on which many users depend heavily
- Lightly used Windows Medicaid billing server running software that is
very slow and requires maximum throughput/speed
- Lightly used Windows Accounting server that is running software that is
very slow and requires maximum throughput/speed
- Windows RDP server that has about 15 authorized users; there's slow
growth on this, and it runs separate instances of my VFP application for some of those users.
- Windows Document management server
- Windows 7 Ultimate 64-bit Antivirus server (not a Windows server OS)
The Accounting and Medicaid boxes both have SQL Server Express databases, and there is client software on the workstations that access it, but the servers also have additional software installed as well. So it does not appear to me that this is either fully fat client-thin server or fully server-hosted software.
The network has about 130 workstations, connected through four gigabit switches, and there is moderate growth on that. We have a 25 Mbps synchronous internet connection and it is heavily used by many workstations. There are only a couple of network printers; most people have desktop printers.
There is a VoIP phone server on a separate physical network but which is connected to the computer network for remote-access purposes and to enable use of "switchboard" software on a few workstations.
There are two other servers on the network that are not candidates for virtualization:
- Fax server; has legacy dedicated hardware
- Backup server, which is almost constantly either running scripts to
backup and transfer data from the other servers, to itself and to a removable drive, or having data fed to it from other servers.
Total "live" data on the network servers is about 1 TB; we can expect slow-to-moderate growth on that.
The risks it seems to me are:
- Fail-over: If the hypervisor goes down, nobody, but nobody, can do any
work. Therefore I need a redundant mirrored system on a separate box, and a robust mechanism to continuously mirror the data without affecting performance. Is that really possible?
- What is going to be the real, day-to-day effect of using a virtualized
RDP server in a stack of other virtualized servers, some of which have heavy intranet traffic? I do not want to be in a situation where I'm told there are no worries and then, this system is installed, and the thing is dog-slow.
- In fact, I don't want that to happen in relation to any of the
applications we are using. What is a realistic expectation on this?
If the reality is that I will be assuming greater risks than I face now with separate physical servers, and those risks cannot be mitigated effectively, then are there any countervailing advantages to virtualization that would be great enough to justify making this change?
Thanks to all for the benefit of your experience.
Ken Dibble www.stic-cil.org
[excessive quoting removed by server]