David Mitchell SmithNot that there isn’t already too much in this world, but hey its a cool headline…
Recently there was another short gmail outage. and it sparked the usual outrage over how unreliable cloud computing is and how OMG nobody should use it etc., etc. Of course people wasted more time complaining about the outage that actually experiencing it, but I digress.
There was also recently an internal email outage that affected us here. What’s notable is that there was no outrage. No claims that ‘on premises email is unreliable’ and how everyone should move to the cloud. I’m not saying there should have been outrage, just noting that there wasn’t any that I saw. Every system has outages. Ours here is generally very reliable and I don’t complain when there is a glitch. But then I don’t complain about a minor outage with a cloud service either. Do you? When your internal email is down for an hour, is there outrage?
I’m not our email or high availability expert. I’m “just a country analyst” and dumb user (of course I’ve been called worse). But some digging shows that most enterprises expect an SLA of about 99.5% (maybe even less), which means downtime of about 3.5 hours per month. Can you imagine the blog traffic if gmail were down for 3.5 hours per month? Twitter would croak (oh wait, it already is croaking..).
Typical uptime for cloud email services is in the 99.9% range. Last time i checked, that looked a lot more reliable. Is there a double standard here? Is internal IT cut a lot more slack? Are expectations unrealistic of cloud providers?
Category: Uncategorized Tags:
6 responses so far ↓
1 Benoit Lheureux September 10, 2009 at 10:30 am
Your post has the potential to spark over-analysis, but stated simply IMO you hit the nail on the head. 99.5% is good enough for a lot of Apps and while that’s the public SLA from many of the B2B vendors I cover, though many offer 99.9%. For most, GMail is free — so you’d think 99.5% would be enough. But public tolerance is thin.
But regardless of what is “just”, by sheer virtue of its size and impact on the public overall you gotta figure that GMail, like the Google website itself, is held to a higher standard. Without meaning to offen our own Gartner IT staff and their generally reliable IT operations, they are not running a Googleplex and they don’t impact millions of people, nor do we here at Gartner expect them too.
Having said that (and I’ll admit this is a bit odd) I’ll add in closing that I have used Google as my homepage for years. Not only because I often open my browser to do a search, but also because to me Google.com is the equivalent of a “Digital Dial Tone” for the Web — if their page pops, I know I’m connected to the Web. Perhaps sad that I need to know that — but with variances in hi-speed ISP service quality from my cable provider, and given my diverse travels and methods for connecting, its always good to know.
2 Ed Emmerson September 10, 2009 at 10:42 am
I thought the same exact thing myself David. After some time though, I would have to argue that the Gmail outage has some contrast to an enterprise outage that in most cases would affect a few thousand people. Most of whom are not going to speak ill of their own corporate infrastructure publicly, even if the outage lasted for much longer. Whereas the Gmail outage affects millions of users whom without that email simply switched over to Twitter or the actual service they were “popping” with Gmail. Also, nobody would care if someone “tweeted” that ABC Widget Co.’s email is down. I would ask if the Gmail outage created real outrage or viral outrage.
Gmail is also ad supported in most cases and as an AdWords client, I can tell you that Gmail users are much more likey to click through an ad than any other web mail service. Therefore the outage becomes a time of non-exposure for that medium to advertisers and seen as a revenue hit. ‘Another country heard from’.
I would conversely argue, to your point, that an enterprise email outage will sometimes bring productivity to a trickle in business units where it is the main form of communication both internally and externally. Again, a revenue hit. I also think that this has a deeper psychological effect on knowledge workers as frustration with email becomes frustration with IT and has a longer term and more direct impact. The discontent that is expressed is verbal, face to face and more powerful and contagious. It may be a silent outrage, but it is there and much more damaging in my opinion.
I look forward to your upcoming tweet that the ‘Exchange server in Trumbull’ is down, but I don’t think it will go viral.
3 Scott Olson September 10, 2009 at 11:01 am
The title of your post reminded me of Seth Godin’s blog yesterday about righteous indignation, http://sethgodin.typepad.com/seths_blog/2009/09/righteous-indignation.html, and how we should do away with it all together. Good advice.
You are right of course that people do over react and that cloud service providers are held to an unfair standard. The icing on the cake is that it is a free service. Ed points out that there are lost exposure time for advertisers, but of course those advertisers are only paying when people are clicking. Gmail is a great service as evidenced by the millions of users they have. Anyone complaining can simply move onto another service if they feel strongly about it.
4 Nick Jones September 10, 2009 at 11:08 am
Maybe there’s no outrage because we have so many alternate ways to contact our colleagues so we only lost one channel out of many. I could still use IM, cellular, SMS, and a few personal email systems. So there wasn’t much real inconvenience.
5 Eric Knipp September 10, 2009 at 2:43 pm
David, the sad fact of the matter is that 95% availability is good enough for most apps.
In truth I would probably be better off if our email were down 50% of the time. More time for research. Of course I’d have to unlearn my pavlovian conditioning.
6 Gordon Van Huizen September 16, 2009 at 1:53 pm
I think Benoit’s comments get to the crux of why there’s so much noise made when a public service goes down. It’s public! The outage affects more people, and there’s natural interest in understanding why it went down and how robust the service may in fact be. Journalists, pundits and users will all be compelled to have something to say about it, and readers will take note…and join the fray.
But there is a very real issue that we can illuminate, though, regarding service levels. As several people have pointed out, 99.999% uptime isn’t a true requirement for most services. But for some services–and some consumers of those services–a specific service level may indeed be warranted. I’d argue that the Cloud can help here. Why? Because real contracts exist between business entities. A consuming party can contractually mandate a service level, with penalties if that service level isn’t met. Quite a bit harder to do with an internal resource!