The ‘eh,’ the bad and the ugly: Cloud Service Level Agreements

You’d think that cloud Service Level Agreements (SLA) would finally be well-written and consistent. Nah!

As users, we’d like to not have to worry about SLAs. What we really want is for clouds to be as reliable as the electrical grid…

Sorry about that. I just had a brown out and my Internet connection went out.

Nothing is error-proof. Sure, there may be tens of thousands of physical servers at any cloud data center, but even at the best of times, things go wrong.

All the major public cloud services go out from time to time. In November, Google Cloud europe-west1 region Compute Engine lost connectivity for about 70 minutes. The Compute Engines? They were fine. The network engineers had decided to manually add a new peering connection instead of using the automated system that checked the link for problems.

So would the Google Cloud SLA cover this? Nope.

Google defines downtime as:

“For Instances: Loss of external connectivity and/or persistent disk access for all running Instances that are hosted across two or more zones combined with the inability to launch replacement Instances in any zone.”

Only one zone lost connectivity in this instance, so those customers were out of luck.

SLAs often don’t provide much coverage, and you have to carefully decode the language.

For example, an SLA that guarantees annual availability sounds good. But is it really? Over a year, the service can be for 4 hours and 23 minutes a year before any kind of vendor penalty comes into play.

If, on the other hand, your cloud vendor offers 99.95% monthly availability, the allowable downtime is suddenly cut down to 22 minutes. That’s because, by applying the downtime metric to each month instead of averaging it over year, the service provider has less slack.

HP Helion, which is now closing down, started this measurement of downtime. Today, Google, Microsoft and Amazon also use a monthly method.

You get the idea. In SLAs, small details make a big difference.

The Cloud Standards Customer Council, a cloud user advocacy group, has been working on setting cloud SLA standards and best practices. But, as they point out in the latest document, Practical Guide to Cloud Service Agreements, Version 2.0, “cloud customers should be aware that there may be a mismatch between their expectations and the cloud providers’ actual service terms.”

In addition, “It is common for disputes to arise over the structure of the agreements, thus cloud customers must pay close attention to the language and clauses of the CSA. Large cloud providers can be inflexible with their CSAs, while small cloud providers may seem more flexible, but tend to over-promise in order to obtain clients.”

Bottom line: Before putting your business on a cloud, make certain it’s not just your CIO and CTO making the deal. Bring your in-house counsel as well. When it comes to getting the most from your SLA, you’ll need both technology and legal expertise.

Illustration by Bob Reselman,


  1. I think you meant … as reliable as POTS telephone service, only nobody has that anymore 😉


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: