Application Performance Management 

Cloud 99.999

Gmail is currently down for some of the service’s users – and so the world anxiously holds its breath waiting for the resolution.

Gmail is currently down for some of the service’s users – and so the world anxiously holds its breath waiting for the resolution.  This follows on the heels of a larger outage two weeks ago.

It should be worth noting that Google’s paid userbase are covered by a 99.9 percent uptime commitment.  It’s not the usual 5 nines associated with most network applications – although in light of recent events…

So, let’s say that we’re talking a total of two days of downtime over the past two years since it was launched to the public on February 7, 2007.  The last outage had Gmail skating on that 99.9% edge, this outage pushes it down to 99.7%.

But wait, that outage was the third downtime in six months.

Now, don’t get me wrong, Gmail is a wonderful application.  It’s taken the portability of its predecessors and added amazing flexibility – I can get my Gmail email through the Web, on my iPhone, browse headlines through an RSS feed, download to my home Thunderbird client, and even use it as a way to store all those darn game serial numbers in case I lose the sticker that comes on the case.

Cloud apps provide anywhere, everywhere access, with zero-brain-required “installation,” configuration, and maintenance.  It does a great job of separating content (the stuff you want) from context (how it is presented) so that you can have it presented in a number of formats.  These are all great things, and explains the appeal of the cloud platform.

For a home user like me, 98% is reliable enough.  But that’s just it – cloud models (currently) can’t provide the same kind of reliability that we’ve come to expect from LAN and WAN applications.  Google didn’t promise “five nines” – they promised three, and delivered one.   I don’t think that it’s a matter of being oversubscribed or of improper maintenance – Google is Google,for crying out loud – but that a cloud application, by definition, contains all the problems associated with any app accessible over a wide area network (in this case, the Internet) as well as the problems associated with serving multiple customers from multiple locations.  The number of mail users a Google or a Yahoo has to support dwarfs the number of users supported at even the largest of enterprises.

Moving to the Cloud is basically putting your application into a shared resource pool, and it seems to be part of a larger trend.  Virtualization lets us consolidate multiple servers onto one machine.  WAN Optimization lets us consolidate multiple machines into one data center.  What the cloud does is consolidate multiple companies IT departments into one shared pool.

The point is – when you go to the cloud, you sacrifice a little reliability for greatly increased flexibility.  This is not to discourage you from making a switch over, but just be aware of the risks and, more importantly, be aware of your needs.  In a cloud environment, monitoring network bandwidth remains important because cloud providers will need tools to assess traffic and end-user responsiveness, so that they can adjust their computing capacity to handle the traffic without expensive unnecessary overprovision.

Comments

rewrite

Insights from the app driven world
Subscribe Now >
RECOMMENDED
The Sociology of Software >How (Not) to Lie with Data Visualization >DevOps and Cloud Computing: Exploiting the Synergy for Business Advantage >