Grids, SLAs and the paper they're not written on

Wow. I've just come from the non-technical parallel session at SSOKU09... It covered everything from the similarities between social networks, bee swarms and the global economic crisis to the idea that we can pay our way out of the ICT skills shortage (wheee...pay hikes for all!?)

The first speaker was Davide Parrilli, a lawyer with an interest in SLAs (Service level agreements: contracts between the users and providers of compute services). The idea is that SLAs are a legally binding agreement between user and provider, guarenteeing payment by one, and provision of resources by the other. The other idea was that SLAs should become more and more favorable to users as the produce gets better and better. Weeeellllllll.....that was until question time anyway.

A few ideas were bandied around during question time, by both speaker and audience:
"Not worth the paper it's not written on" was one, and "Bottom line: read the contract" was another. The conclusion was that SLAs "can help, but they cannot do the whole job." Apparently, many SLAs include clauses that absolve the service provider from actually providing the service if it doesn't suit them at the time. So what about those 99% availability guarentees? They may only apply if and when your provider feels like it!

If you feel your provider really has done the dirty, and you wish to appeal for your rights in court, you may also find that the so-called "competent court", as selected by your service provider, is located in a country far, far away from your SME HQ. Not exactly practical.

So advice from the lawyers?

1) Read the contract
2) Ensure that its CONTENT is what you're looking for
3) Ensure that it is ENFORCABLE, in practise, not just in theory.
4) If all else fails, opt for TRUST. Grid computing has always been about sharing, and thus about trust. SLA or no SLA, trust between service users and providers is really the bottom line.