The Washington Post reports that Google is offering Google Apps Premier Edition paying customers a 15 day credit to make up for the service outage that occurred on the 24th of February. If you look at the Google Apps SLA, this means Google is offering compensation as if the February uptime percentage will be under 95%.
Now, for a short month like February, a 4 hour outage would mean 99.4% uptime and Google’s SLA says you only get 7 days/user credit for that level. Google may just being a bit generous to deal with the negative public relations of a 4-hour outage – or Google could be tacitly admitting that less than 95% uptime is what you really got in February. And that might be exactly what you can reasonably expect.
Self-run or traditionally outsourced enterprise email services may have total uptime that isn’t much higher than that – but how much of the downtime is 4 hours long? Sixteen 15-minute outages with email is not a big deal – a four hour outage is a very big deal. The uptime % calculation is the same – but the business acceptability is not.
Another note – Google’s SLA has the caveat “The Google Apps SLA does not apply to any service s that expressly exclude this Google Apps SLA (as stated in the documentation for such services) or any performance issues: (i) caused by factors outside of Google’s reasonable control;….” Hmm, I’m pretty sure internal IT shops running email don’t get such a nice wide loophole. If attackers launch a denial of service attack against Google that impacts uptime, is that outside their reasonable control? I don’t think it should be, because DDoS protection should be a standard capability for every Software as a Service provider – but I guess that will be up to the lawyers.
The bottom line comes down to: is a 4-hour outage acceptable for your business email services? Is 15 free days of service that may have 4 hour outages useful compensation?
6 responses so far ↓
1 Paul Lopez // Feb 26, 2009 at 4:33 pm
John – Good call on the SLA calculation. As you point out, a large block outage is more severe than many smaller ones. In fact, how often do we get intermittent interruptions with Exchange? After all, this is asynchronous, best efforts. What troubles me is the expectation of some enterprise class service from an Internet search company.
Paul Lopez
2 Russ Malz // Feb 26, 2009 at 5:02 pm
John –
Excellent point, and I agree with Paul Lopez (above). Google is a good choice for a consumer, and their mail product has some nice functions. That said, Google email and aps were not engineered for the enterprise.
3 Lydia Leong // Feb 27, 2009 at 2:34 am
Well, service providers like IBM have the “reasonable control” clause in their contracts, too, and a lot more verbiage besides.
Google doesn’t even get the grace of maintenance windows. A 1:30 am outage for folks in the US would probably not draw blinks in the internal enterprise — but because Google is 24×7 global, they get zero breathing room. Their outage was the result of a software update, in the end. In the enterprise (or in most outsourcing arrangements), you get a window in which to do such upgrades in which downtime is expected.
I think the irony of running an email infrastructure the size and scope of Google’s is that you are actually engineering to a heck of a lot higher standards that you do within the enterprise. You can unplug a rack full of Gmail servers and the service will keep chugging along seamlessly. Unplug an Exchange server in the enterprise, and not only will your email come to an immediate halt, but there’s some non-negligible chance it won’t come back up smoothly, either. And it’s not like the code quality of commercial email packages is necessarily any higher than Gmail’s. It’s just that your tiny software bug or even quirk can get magnified a millionfold into a massive outage when it happens to be what runs Gmail.
4 John Pescatore // Feb 27, 2009 at 9:42 am
Lydia – all well and good, but if Google wants to be in the business of selling services to enterprises, vs. putting ads in front of consumer eyeballs, it will have to do better. It really doesn’t matter to an enterprise employee that Google is engineering their plumbing to higher standards – if the toilet doesn’t flush when I need it, it is broken.
5 Andrew Barker // Mar 12, 2009 at 6:22 am
Hi
From a European perspective you are on the wrong tact entirely. The problem is less the outage (although it hit us when we are awake!). But the service that is left behind is very basic HTML. So if you have the temerity to want to read an email, the likelyhood of it failing is 90% and forget anything else.
So the outage is not really 4hrs, it continues as we speak for most of us. Google may be protecting their core North American region. But to all intents and purposes the rest of the World is out!
Andy
6 Google Apps: Consumer SLAs for Enterprise Business | Cloud Computing and Bad Behaviour // May 11, 2009 at 9:29 pm
[...] John Pescatore from Gartner highlighted some interesting aspects to the Google Apps Service Level Agreement – for instance “The Google Apps SLA does not apply to…any performance issues: (i) caused by factors outside of Google’s reasonable control” (clipped for efficiency, read the full extract and John’s comment at his blog at Gartner.com) [...]
Leave a Comment