I have seen this headline a couple of times in the last month: Cloud Customers at the Mercy of Providers! It's just ridiculous. We left a five-nine world a while ago. Redundancy does not fix everything. And to put it into perspective, to run redundancy on your own Hosted Exchange server would be expensive from a labor and hardware standpoint. It would also be complex and not automatic.
Even when youtry to build in redundancy (like Netflix did by utilizing different sectors of AWS), it sometimes fails. We have seen outages this year by Google, Yahoo, Netflix, Amazon, Facebook, twitter, Rackspace, Salesforce and probably others I am unaware of. I don't think this will slow down cloud adoption. People choose cloud for reasons that have nothing to do with redundancy. Cloud is financially efficient (as Savvis puts it), flexible, and available from any authorized and enabled device. It also removes a required skill set off the books. In other words, businesses can focus on their own business and not on tech or IT. In addition, the remote/virtual/mobile workforce grows every year, driving more cloud adoption. There is no going back.
Think about doing it yourself. You would need the following:
- data center room or NOC;
- generator that has to be tested and maintained;
- battery backup - tested, maintained and environmentally sound;
- servers, switches, routers, fiber-channel, power channels;
- duplicate gear;
- fire suppression system;
- compliance certificates;
- licensing for any software;
- client software or apps for every O/S - mobile and desk;
- Internet capacity for remote access;
- redundant Internet pipe;
- VPN or other security device with RADIUS for access authorization;
- staff that knows how to handle all of this stuff, 24x7;
- power usage;
- air conditioning;
The CAPEX would be large (which is one reason buyers choose cloud) and the labor costs - hiring, retaining, training, benefits, etc. - would be high - and in some cases scarce. And despite the outlay of capital - human and otherwise - there is no guarantee that you can keep it up 99.99% of the time - which means about 1 hour of downtime per year.
I'm not defending the outages, just saying that this will be expected behavior, just like dropped cellphone calls and faxes that required three or more retries.