VMware ESX?
Author
Discussion

hutchingsp

Original Poster:

68,520 posts

241 months

Thursday 5th July 2007
quotequote all
Is anyone using this in a fairly small outfit i.e. up to a dozen non-critical servers?

If so what are you doing for hardware redundancy?

I'm looking at various scenario's from having a single box with as much onboard redundancy as possible and crossing fingers to looking at a couple of boxes with an entry level SAN behind them.

TIA.

kiwisr

9,335 posts

238 months

Thursday 5th July 2007
quotequote all
Only used it in dev&test environments but you'd need to look at VMWare HA, which is included in the Enterprise version. Does what it says on the tin, overall very impressed with it.

Guess it comes down to how flexible your environment is and whatever works out better from a cost perspective.

theboss

7,555 posts

250 months

Thursday 5th July 2007
quotequote all
An entry-level SAN will give you server redundancy but still leaves you with an obvious single point of failure, and storage redundancy is extremely expensive.

In these situations, i.e. smaller environments and money being an object, I tend to go for the approach that less is often better, based on the fact that more simple environments have far less to go wrong and are easier to recover, not to mention the fact that they cost a lot less in the first place.

If you have to have server redundancy, instead of a SAN perhaps you could invest in two idential well-specced servers, both running ESX, one of which is live and the other for recovery. Or consider using Virtual Server / VMware Server instead of ESX, which would allow you to use software replication (e.g. Double-take) at the host level.

Just my 2p's worth.

Edited by theboss on Thursday 5th July 22:38

hutchingsp

Original Poster:

68,520 posts

241 months

Thursday 5th July 2007
quotequote all
theboss said:
If you have to have server redundancy, instead of a SAN perhaps you could invest in two idential well-specced servers, both running ESX, one of which is live and the other for recovery. Or consider using Virtual Server / VMware Server instead of ESX, which would allow you to use software replication (e.g. Double-take) at the host level.
That is an option.

An HP DL380 hooked up to an external cabinet seems to look a reasonable option on the basis if the box fails we could hook the cabinet to a temporary box.

I did consider VMware Server, at the lower end though ESX starter seems a bit of a no-brainer for the lower overhead/additional performance - obviously once you start getting into "proper" redundancy you have additional hardware costs plus a lot of additional license costs.

theboss

7,555 posts

250 months

Thursday 5th July 2007
quotequote all
hutchingsp said:
theboss said:
If you have to have server redundancy, instead of a SAN perhaps you could invest in two idential well-specced servers, both running ESX, one of which is live and the other for recovery. Or consider using Virtual Server / VMware Server instead of ESX, which would allow you to use software replication (e.g. Double-take) at the host level.
That is an option.

An HP DL380 hooked up to an external cabinet seems to look a reasonable option on the basis if the box fails we could hook the cabinet to a temporary box.

I did consider VMware Server, at the lower end though ESX starter seems a bit of a no-brainer for the lower overhead/additional performance - obviously once you start getting into "proper" redundancy you have additional hardware costs plus a lot of additional license costs.
It depends how important performance is, you said these were largley uncritical, plus you may be surprised just how swiftly VM's run on modern kit with quad-core CPU's, hardware assisted virtualisation, and lots of cheap memory, even with a fat ugly host OS managing it all.

Plus the money you save on ESX (both those products are free) would easily buy you some double-take licenses, and there you have it, real-time bit-level replication of your VMs and even site redundancy if you fancy it.

Alternatively if you need the robustness of ESX you can combine both principles, run ESX on two seperate hosts (with local storage) and replicate at the Guest level, i.e. install Double-take in each of the VM's and replicate to 'standby' VMs on the other host. Double-take supposts such deployments.

Or there's the SAN route you already mentioned - you have lots of choices!!

_DeeJay_

5,057 posts

285 months

Friday 6th July 2007
quotequote all
kiwisr said:
Only used it in dev&test environments but you'd need to look at VMWare HA, which is included in the Enterprise version. Does what it says on the tin, overall very impressed with it.

Guess it comes down to how flexible your environment is and whatever works out better from a cost perspective.
We run about 500 VM's on ESX in 15 countries. However, one thing I'd never recommend is paying for HA in the current release of the product; it's terrible and VMWare pretty much acknowledge that. DRS is fine, but in the event of a total failure to all hosts, HA often makes a complete mess of recovering the service.

The choice you make will all depend on your requirements. Does the failover have to be seamless in the event of a hardware failure, or will you have the ability to recover the VM's manually? Also, what kind of failures are you attempting to protect against?

hutchingsp

Original Poster:

68,520 posts

241 months

Friday 6th July 2007
quotequote all
_DeeJay_ said:
The choice you make will all depend on your requirements. Does the failover have to be seamless in the event of a hardware failure, or will you have the ability to recover the VM's manually? Also, what kind of failures are you attempting to protect against?
Not seamless no.

Right now we're in the boat of "one application/role = one physical server"

We have a large physical site, there's a main server room and we have a rack in another building on the end of a 2gbps fibre link.

What would be nice would be to be able to run VMware in the main server room but have our VM's replicated to the other location so that in the event of "something bad" we could reasonably quickly get up and running, the secondary system wouldn't need to be as fast as the main system as it would simply be a "get out of jail" system.

Of course there are various ways of doing this from replication at a SAN/brick level through to software, which is where I'm a little out of my depth in terms of specific product knowledge.

With direct-attached storage I'd assumed ESX Starter might be the way to go, but I'm unsure how the replication would work?

It goes without saying that the more seamless/resilient the solution the higher the cost, but the greater the odds of management buy-in because they could then consider running our MIS systems on it etc.

Put it this way, I'm not convinced all the hardware/ESX licenses needed to go the full SAN route is actually necessary.

Edited by hutchingsp on Friday 6th July 08:20

GhostyDog

464 posts

238 months

Friday 6th July 2007
quotequote all
depends on what you want and what applications your running. I've been thinking of using VMware Server for a while in some scenarios it ticks all the boxes in others it doesn't which leaves me half my stuff VMware and half my stuff unconsolidated like you have at the mo.

I'd start by doing a risk assessment of your systems, you probably have an idea what you think are the important systems but does that align to the businesses idea of what is mission critical? What level of downtime can you afford? Do you have metrics in place that can give you a relative cost to the business that downtime will cause?

IMHO without these in place you can't really design a system that provides maximum value versus TCO.

hutchingsp

Original Poster:

68,520 posts

241 months

Friday 6th July 2007
quotequote all
The specific roles at the moment would be:

1 x Intranet
1 x DMZ Web Server
1 x Domain Controller (we have another physical one)
1 x Light usage SQL Server
1 x Light usage Citrix

With the potential to add more VM's as other servers are replaced/outdated etc.

We're not a manufacturing/retailing company in the sense that down time doesn't directly equate to money lost.

At the basic end of things I'm starting to lean towards a single box running ESX Starter and something like ESXRanger to backup the VM's to disk (which would then be streamed to tape) and if the box breaks very badly I could bring up a temporary box with ESX and restore the VM's.

Where it starts to get interesting is when you look at the whole High Availability thing with redundant systems in our main server room as well as at the other end of our site with replication between the two. That's overkill for the above roles, but I'll be getting some prices and then my boss and the powers that be can decide if they think things such as our MIS systems could benefit enough to justify the cost.

It's very difficult to put a metric on our downtime as we're a R&D/Engineering company, we don't produce widgets and have many roles so we can't say "1 minute downtime = £x lost revenue" etc.

kiwisr

9,335 posts

238 months

Saturday 7th July 2007
quotequote all
hutchingsp said:
It's very difficult to put a metric on our downtime as we're a R&D/Engineering company, we don't produce widgets and have many roles so we can't say "1 minute downtime = £x lost revenue" etc.
It's not just revenue, it's other costs such as labour that you still pay for while the systems are down. Most companies will start to lose their patience when their staff have no IT for a day or two smile

Munter

31,331 posts

272 months

Saturday 7th July 2007
quotequote all
Those of you that are running VMware can I ask what you are using to do capacity management / planning?

hutchingsp

Original Poster:

68,520 posts

241 months

Saturday 7th July 2007
quotequote all
kiwisr said:
It's not just revenue, it's other costs such as labour that you still pay for while the systems are down. Most companies will start to lose their patience when their staff have no IT for a day or two smile
Yep that's a fair point.

My gut reaction is that with a setup such as the one I mentioned above that shouldn't happen.

We have all our kit under decent on-site warranty and the backup side of things is pretty well covered, without sounding complacent I like to think I have all the things covered that are under my control and can be covered with the money I have available.

I'm trying to be realistic in that I'm not thinking of putting our file/exchange servers on VMware as I don't believe there are sufficient benefits to justify not simply having a physical box for each role.

kiwisr

9,335 posts

238 months

Saturday 7th July 2007
quotequote all
Munter said:
Those of you that are running VMware can I ask what you are using to do capacity management / planning?
We used a product called Platespin PowerRecon.

Puggit

49,832 posts

279 months

Saturday 7th July 2007
quotequote all
kiwisr said:
Munter said:
Those of you that are running VMware can I ask what you are using to do capacity management / planning?
We used a product called Platespin PowerRecon.
That's good to know! How do you rate it?

I work for PlateSpin if anyone wants to ask questions smile