Disaster Recovery and vmware

No disrespect, but that's a really bad idea above. It's a perfect recipe for ending up with a poor performing, unmanageable sprawling mess. For one thing, Neither VCB nor VDR are available in the free ESXi.

I've said it before and I'll say it again, as it applies to everything in corporate IT - do it properly or don't do it at all.
 
No disrespect, but that's a really bad idea above. It's a perfect recipe for ending up with a poor performing, unmanageable sprawling mess. For one thing, Neither VCB nor VDR are available in the free ESXi.

I've said it before and I'll say it again, as it applies to everything in corporate IT - do it properly or don't do it at all.

I didn't say VCB was free, but how is managing 10 servers (for arguments sake) any worse than managing 28 to create sprawl? Like anything in IT it needs to be managed, even a simple file system can become a nightmare if its not managed so I just don't see your argument as being valid.

The way to approach it would be to do a phased delivery with an initial testing phase. Going full hog, full tilt in one step is how you end up breaking things. Virtual a test bed and see how it fairs.
 
I didn't say VCB was free, but how is managing 10 servers (for arguments sake) any worse than managing 28 to create sprawl? Like anything in IT it needs to be managed, even a simple file system can become a nightmare if its not managed so I just don't see your argument as being valid.

The way to approach it would be to do a phased delivery with an initial testing phase. Going full hog, full tilt in one step is how you end up breaking things. Virtual a test bed and see how it fairs.

Virtualising adds another layer of management. You need to configure networking on host and guest, you need to configure storage on host and guest etc etc.

When you want to manage your exchange server, in a physical environment it's simple. In a virtual environment, you need to figure out which of your 10 non centrally managed hosts it's on before you can do anything.

Small scale VMWare projects like shouldnt have 10 hosts, they should have 2 or 3. Even in the absence of centralised management with vCenter, that would be a lot simpler than your suggestion.

As I keep saying, do it properly or not at all - middle grounds always cause more hassle than theyre worth
 
Damn I just wrote a reply and I lost it, damn browser.

In summary, your boss is a fool, virtualisation is a not a buzz word and I hope he's not an "IT professional" otherwise he is well behind the curve.

I agree with iaind above, your project is relatively small scale and doesn't have to be implemented over night. Start with one host and ESXi and get some testing done, if you can, get a SAN too. It doesn't have to have FC connectivity to start with as it won't be your primary solution you don't need multiple hosts if you are only testing.

Get a foot in the reigns and take it for a spin. It will give you an idea of what is capable on a virtualised solution. HA and DR alone are worth the money spent on a decent setup. Many trials and demos to be had and seen for you to put a case together for your management.

We started here relatively small time with a new server and and older one being commissioned as hypervisors we only had one Clariion SAN and no FC connectivity. Since then we own a range of software and tools, have more hosts, more SANS, multiple FC switches and excellent HA and DR capability.

The beauty of VMware is that it is totally scalable. Good luck OP!
 
Last edited:
should i be looking at the fibre route or gigabit ethernet?

Cost is ALWAYS an issue where I Work, so it needs to be feasible to save the company £££ in the long run.

if the SAN Dies, then what? (cause i know thatll be the response)

I would personally go copper gigabit.

My reason for this, is 10Gb Ethernet is rapidly increasing in use, our blades have 10Gbe cards on them, so all we need to do is swap a switch to a 10Gb one and all is good.

Copper is also cheaper, and I believe faster than FC.

Kimbie
 
I think the main issue is neither of us know enough to really push it forward.

Definately going to be spending more time on ESXi for now and see even if i can virtualise a few less important servers where downtime isnt an issue.

Im going to try and move in this direction in the coming months as alternatives to just buying servers that perform 1 role and wasting resources.

Thanks for all the responses, greatly appreciated ^^.

Currently starting the business case as we speak!
 
That is the best course of action as I see it Ashley. Get testing with some older server hardware and get ESXi on there and test/investigate/learn. Put some non-business critical applications/services on there and see what happens.

The more you practice and experiment, the more you will learn and be able to put forward a stronger case!
 
I think the main issue is neither of us know enough to really push it forward.

Definately going to be spending more time on ESXi for now and see even if i can virtualise a few less important servers where downtime isnt an issue.

Im going to try and move in this direction in the coming months as alternatives to just buying servers that perform 1 role and wasting resources.

Thanks for all the responses, greatly appreciated ^^.

Currently starting the business case as we speak!

Where are you based? You are welcome to visit us - South Wales - and see how we've implemented. It may give you a better idea of how things work. As I said, we spent £70k ish on our infrastructure - so we didn't hang about with it, but it'll certainly help you realise what virtualisation can achieve.
 
Where are you based? You are welcome to visit us - South Wales - and see how we've implemented. It may give you a better idea of how things work. As I said, we spent £70k ish on our infrastructure - so we didn't hang about with it, but it'll certainly help you realise what virtualisation can achieve.

Based in Redditch, Worcs. May take you up on that actually! SW isnt too bad and would definately be worth the visit!! Many thanks for the offer. if you could pm me your details i'll save them :) and take you up on it further down the line.
 
Just a query, why so hellbent with vmware, aside from support that is?

Xen or HyperV is fine. Server 2008 comes with a specific amount if "free" licenses for windows hosts

And Xen on a linux host just rocks.

- There's a cost saving alone there, especially if your on a budget and putting a business case together, you'll need to weigh up all possibilities
 
Wasnt particularly sold on any specific product, from impression on the web vmware is the market leader in virtualization?

Downloaded xen to test, have looked into Hyper-V but dont know a lot about it to be honest,

Licensing costs will obviously be a big factor, be interested to hear from people who use various other products aside from esx.

Ash
 
Features, plain and simple, vmware is so far ahead of the game that its not even funny.

Saying that, xenserver does seem to do a better job of TS/citrix
 
Features, plain and simple, vmware is so far ahead of the game that its not even funny.

Saying that, xenserver does seem to do a better job of TS/citrix

Mainly because Microsoft and Citrix are partners and obviously Citrix Xen Server is going to try and give an advantage to using it's own products.

Seriously though in a Production environment I would, at this moment in time, only consider VMware due to how far ahead they are technology wise and also the inovation, ease of use, stability, etc. that goes into there products it just works is a familiar phrase when using it.

Obviously cost is a massive factor. If VMware is double the price then Xen maybe a good solution but that doesn't mean it's as good as VMware and you then need to weigh up the additional features that using VMware would give you.

There's no easy answer overall though.


M.
 
Youd only consider vmware for every possible selling point? :)

For any reasonably sized business, the costs aren't that much of an issue, a few grand a host is nothing in the big scheme of things, especially when you can be making significant savings (especially by timing projects well) and intangible benefits become much more important
 
There's one on the vmware website but its liberal to say the least!

Not hard to do, you know what savings will apply to your business, you can look at before and after power consumption etc etc
 
Hiya mate,

Firstly, try not to confuse the terms Disaster Recovery with High-Availability, I will assume you refer to avoiding a problem within the same site (HA), rather than failing over to a different site (DR).

The other posters are in the right ball-park in terms of costs (£50-80K) I will try and help work out how you will get to those figures yourself and start to create a financial model for a business case.

I have done a few business cases for Virtualisation, the areas which to tackle in terms of your savings and expenditure are as follows.....

Please treat what I have written with a pinch of salt, there are so many assumptions and caveats, you will need to fill in the blanks and variables yourself :>

Savings

1. Capex Saving - Microsoft Licenses (For New Servers)
2. Opex Saving - Microsoft Software Assurance Saving (Current Estate)
3. Opex Saving - Saving On New Server Equipment For Projects
4. Opex Saving - Hosting Costs (Power & Cooling Servers) - Cumulative
5. Opex Saving- Hardware Refresh Cost (Current Estate) - Assuming 4 Year Cycle.

1. Depending on who talk to, Microsoft Datacentre Edition is £1400 per socket, £2800 for a standard 2-CPU Server. This license allows you to Virtualise as many machines as you can fit onto the hardware. While this sounds expensive, it will save you money as you will not require separate licenses or maintenance i.e if you are paying 30 Server x £750 (£22,500) for a Microsoft Standard/Enterprise licence per year then you will then pay 3 x £2800 (£8,400) instead, assuming a 3 node Virtualisation Cluster.

2. These savings are usually cumulative, per year, over a 5 year period due to the way modern Microsoft Enterprise agreements work. Big bucks to be saved here.

3. The Visualisation cluster will be specced to host and absorb new requirements, how big do you intend to grow? 1 Server per year? 5? Factor these savings into the business case.

4. Power & Cooling. Tricky one this, it is the popular one that most calculators go for. Unless you rent your facility from a provider, these calculators can be mis-leading. Why? Well if you have to provide your own facility, there is a considerable capital and operating expenditure in doing so, this negates some, if not all of the savings you will experience due to lower power requirements and these numbers are not factored into the equation.

There is merit in doing this if you are building a £m Datacentre, will it make the business case on your scale? Doubtful. If you do need a yard-stick, then I usually use £250 per server, per year.

What will may make a big difference, is if you rent your space from a provider, going down from 10 racks to say 3, will save you a small fortune, depending on your contract.

5. Refresh costs. Now for me, this is where the big money is saved and can raise the eyebrows of your finance man.

If we assume a 4 year refresh cycle and an average cost of £5000 to procure, install and transfer to new kit, you will bank £150k, into your calculated savings, every 4 years. These savings increase, the more the business grows.

There are associated savings in Network Ports and fewer switches, although with 28 servers, it is possible that you only use 4 x 48 Port Switches anyway, even with physical servers so, the savings will be small.

There are plenty more areas where you will save costs, these are very tough at trying to prove or dis-prove, such as decreased effort to provision and maintain the environment. I would avoid these although, some software firms will insist they can calculate these savings reliably. I don't believe them.

All in you are probably looking at saving around £200k over 5 Years.


The Costs

1. Servers. In my humble, you will need a minimum of 3 Server Hosts and an external backup/restore box. The servers at the moment are very powerful in terms of RAM capabilities. I have experience with all of the latest hardware platforms and I would look for a server with large numbers of RAM slots in a 2U or 4U rack form factor. The Dell R810 with 128GB RAM (8GB DIMMS) is an excellent cost/performance box, assuming 3GB per VM and a 60% utilisation ratio across the cluster, this gives you a 75+ server capacity.

Assuming £11,500 per server for the hosts and £5000 for the backup/restore host, the cost for Server Hardware is around £40k.

2. Software. As previously stated, you will need 6 x CPU of Microsoft Datacentre Edition Licenses and an Enterprise license. £10k.

If you want to use VMWare, you are looking at £800 per CPU for Advanced Edition (this gives you the nice backup/restore software) plus vCentre. This would place the cost in the region of £7000 for VMWare Licenses.

£17k.

3. Storage Hardware. Given your scope, I would go for a shared storage system using 1GbE iSCSI or NFS. This should be plenty fast enough for your needs.

If we assume 70GB Per Server, with a 40% De-duplication ratio, we can esti mate 2.1TB and 1.2TB respectively. I think for a nice Equalogic array, maybe £6-10k? I usually deal in the larger enterprise arrays (Hitachi, Netapp, EMC) so am not too clued up on costs on the smaller ones.

£10k.

4. If your switches are reasonable and currently run at 1GbE with Jumbo Frames then you can probably re-use these.

All in all, I reckon you are looking at a Capital outlay of around £70k.

In theory your "profit" over 5 Years will be in the region of £130k.
 
Last edited:
Besty i could kiss you.

Apologies the title is a tad misleading, I am more concerned with same site failover.
Im going to give this a thorough read and more thought tomorrow.

One Q: With vmware am I going to pay £800 per core (physical?) so if i have a single quad core xeon, do i need to shell out £800x4? or just £800.

It seems the company wants High Availability, but doesnt want to pay the premium for it, getting this across to The Board is always going to be an issue, I think your post is just the ticket!
 
Last edited:
Back
Top Bottom