Bladewatch.com

There’s a trend to outsource the grid infrastructure activities, whether in the form of renting datacenter space for you to plug in your network, your servers and blades, or the more popular deal is based around an on demand service.  I request 400 or 4000 servers for grid and plug those grid engines into my organization using a leased line type arrangement.

Ignoring the IT Security type issues and internal politics, what should I be looking for when discussing these deals.

What’s included? Including those chargeable services in the small print

Who buys the licenses?

Who has the capex?

How’s the billing done?

What’s the option to leave?

How are any energy costs increases dealt with?

What’s the SLA? What are your expectations internally?

What’s included
This is important, I don’t mind the blades/servers being monitored, but often the monitoring might be billed per item, so the more objects monitored, the higher the cost, also does this monitoring software have a license or support cost?  You may be paying for more than you need, and all these little things add up.  An example, you’ve got to pay £5 for an altiris license per unit per month, for your 1000 blades, that’s £5,000 you’ve got to pay or £15,000 a year, couple that with a monitoring license like MOM say £15 per unit per month and you’ve just spent an extra £20k per month.

Who buys windows/datasynapse for the boxes?
You will probably find that it’s cheaper to source the software licenses through your internal procurement process, check this, you might get a more significant discount than the vendor/service provider.
Check most inportantly the cost of any remote deployment licenses, you might find that you pay a significant cost for a license to remotely deploy windows/linux on the blades – check agains what your business can buy them at.

Who has the capex?
The main benefit from this kind of deal is that you can change your grid infrastructure from an ongoing depreciation model to a service, the kind that businesses love, but do check this is the case.

How’s the billing calculated?
If you’ve got an application used 9-5, do you pay a fixed cost per blade regardless of whether its in use, or is it a cpu/per hour used kind of arrangement?  You might find a fixed cost per blade gives you the incentive to encourage more users on to the grid and lower the marginal cost.

Whats the exit strategy?
Check you’re not tied into a three month or six month leaving period, you don’t know what your requirements could be, you may find that suddenly you need three times the capacity and a rival can provide that, do you want split vendors/split sites?

How are energy costs dealt with?
I only mention this because you don’t want surprises half way into the venture where the costs rise to the point where the benefit of your grid is competing with cost.

What’s the SLA? What are your expectations internally?

You need to balance risk against cost, what are your expectations with regards to an outage and how that relates to the service level agreement. Linking this to your existing support models, if the grid is tier1, and your teams are going to respond with tier1 support and expectations, how do you link that to a 9-5 support deal, or next day sla?  Do you really want to pay the premium for the 4 hour response?

No related posts.

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