Bladewatch.com

So what are you working on?

I’ve recently being trying to get approval to decommission all Windows 2000 servers, and combine with it, that servers that are not running Windows 2000 be replaced. Windows 2000 is shortly coming out of support and I want us to start replacing them with Windows 2008.

What number of servers does this involve?

We’ve got a server estate of 1900, 900 in London, about 400 in Chicago and 300 in our offices in Switzerland and Hong Kong. From this, we’ve got about 570 servers running Windows 2000, all of which should be replaced and the hardware refreshed.

I’m actually trying to enforce that anything older than 18 months be replaced.

 Can I ask about the hardware element? Yes, in what respect?

Anything that’s running Windows 2000 is going to be pretty old and not worth the bother to rebuild.

Is this an internal IT or a customer driven project?  Has someone in the business said “let’s dump Windows 2000″ or is this an IT initiative?

Oh It’s and IT one, the business need to be made to understand that Windows 2000 comes out of support next year! So we need to dump it, move on and bring our infrastructure into the year 2009. Also at the same time, we need to refresh our hardware and bring it up to date, this should improve energy efficiency and performance, I was thinking we could virtualize everything we could.

Have you got an anticipated project cost?

So far it’s about 1800 man days plus hardware. That includes the project manager/server/network/os time. I though by bundling it in, we can do everything in one go.

It might be a bit of a challenge but it’s a sales pitch we just need to do for funding, and I suspect the support teams will be behind it, so they can ditch the legacy systems. I’m doing a presentation to the head of IT about it next week.

Do you know how much you will need for hardware?

Well I was thinking if we manage to virtualize 20%, that leaves us with 400 servers or so to upgrade, (on the basis that we should be able to re-use some servers or decommission some roles via consolidation).

Some thoughts on his comments which we discussed.

So in the example above. Mike is asking for:

1800 man days.

• Let us say that a man day costs £300. 
  • Half the cost of asking a service provider to provide me a Wintel/Storage/Networks guy for the business day (9-5)
• That’s £540,000 assuming everything is done on time.

400 servers.

• Let us say mid range 2u server, mid price £3500?
• That’s £1.4 million.
• These figures exclude – storage, network infrastructure (deploying ILOs? 10GB Ethernet/operating system licenses/application upgrades for those applications which aren’t validated to the new operating system).

Remember, I cannot simply say there’s Windows 2008, I have to allocate a test system, have the os configuration validated, the application loaded and tested. I then have to discuss and approve the server configuration, ratify it against internal standards.

Five years ago, a stand alone Proliant 1600 might have been ok, but do we need dual power supplies? Clustering? Does the application need 1GB Ethernet which it might not have right now? We’re asking for £2 million on the basis that everything goes to plan and that there are no exceptions.

Let’s take as Mike’s example MSPA000876, his server of which he was particularly unhappy about. A large file server with for volumes ranging in size from 400gb – 800gb, a DL380 G3 with 2GB RAM, with Emulex LP9000 cards, running Windows 2000.

His concerns:  (perfectly valid ones)

• It’s running unsupported (out of date) Emulex cards
• What cards have we internally validated against each hardware, what drivers are validated, what cards have been validated to our storage switch, and our SAN? What do the vendors support and how does that fit in with our standards?
• The server hardware is out of support – have we ever been in support? Do we care?
• It’s got an out of date version of PowerPath (storage software) and the Emulex card drivers.

My questions:

• Does it work? Does the current system meet end user requirements?
• Could we further enhance what we have?
• Are we currently supported, and would the new configuration be in support – what do we mean by out of support?
• Would upgrading the hardware and the operating provide a direct user benefit?
• Does this solution need a new server or could we move it to a filer or something similar?
• Are we re-architecting each platform to check that we are only deploying what needs to have a server – are we consolidating roles – who decides that?
• Have you seen your ‘man day’ requests and thought about the costs?
• Have you decided what elements are in scope or not?
• That DL380 with 5 array shelves attached – who pays to cover it to SAN?  The project, or the business sponsor who really should have done it anyway years ago?
• The servers ‘out of scope’ as they’ll be replaced/decommissioned – who manages that?
• When servers are replaced and the old ones to be decommissioned – how do we ensure that they do go out the door and they aren’t rebuilt?

Could we not go on the basis of:

Windows 2008 is the new organizational Windows platform for production and development.

• All systems should be upgraded to Windows 2008 with those currently Windows 2000 being a priority.
• Issue below specification guidelines based on minium features for example let’s say all servers need to have lights out and shall we go on anything older than a G3.
  • Technically speaking, the G3 is fine, but on the basis that it is over five years old, we’ve had our value for money from it. Ideally we’d be G5 but, let’s look at what is possible.

First of all before we make any recommendations, we need to know:

• How many Windows 2000 installations we have – Mike has said 570 From those 570 servers:
  • How many can be virtualized?
  • How many are production/development How many are on ‘unsupported hardware’
  • How many are IT based (we own) How many are customer based
  • How many are customer facing and ‘technically have no outage’ (require swing servers).
  • How many will involve a SQL/IIS upgrade and therefore application validation?
  • How many will need a re-design – sorry it needs to be a cluster or have replication?
  • How many need an indirect capacity upgrade and who funds that?
• How do we prevent the free for all approach?

My Pentium 200 needs replaced, I’ll have the absolute MOST EXPENSIVE SERVER IN THE UNIVERSE BECAUSE IT ARE PAYING AND NOT ME, IT’S A FREE LUNCH, I NEED 19 PROCESSORS 128GB RAM AND 7TB OF STORAGE. Only to discover it’s an FTP server which sends three files a month and could be virtualized – but that need analysis, debate and approval/acceptance.

Once we’ve got that down, we can then make a presentation based on value for money, what is achievable, what is core, what we can do based on need and availability.

Mike’s not wrong. Of course he’s not. However, what we need is business buy in and we need data. Data that is filtered in such a way not to hide information or create walls, but to say this:

• Club class would be to replace everything – 570 servers, SAN boot it all, using virtualization where possible.
• Premium class would be to replace all servers with Windows 2008 and using SAN where possible.
• Economy class would be to rebuild everything Windows 2008, upgrading systems where appropriate, accepting minimum production standards and buying new servers only where necessary.

Key to this is figures and estimates of cost. I remember one CIO saying to me, “this guy came into my office and said to upgrade my SAN would cost 300 man days, is he nuts? How does he think I’m going to get that signed off?

No related posts.

Related posts brought to you by Yet Another Related Posts Plugin.