The Promise of High Availability in the Cloud
- February 25, 2014
After setting the High Availability and describes how it could be measured, I propose to conclude this first series of articles with Cloud Consulting website by a small overview of the promises of Cloud Computing providers.
And as for the previous items, if you have any comments questions please leave your questions in the comments.
Cloud Computing providers guarantee 99.9%, or 8 hours of downtime (Amazon S3, SimpleDB, Azure, eNlight cloud computing service providers in India) and 99.95%, or 5 hours of downtime (Amazon EC2, Azure Compute) per year. GoGrid offers about an offer cloud Guaranteed 100% with a payback of 10000%. Beware though, this figure of 100% does not correspond to a measure of GoGrid, since it is by definition unattainable. Rather, it is a contractual arrangement (SLA – Service Level Agreement) as some other providers offer, which pays in this case up to 100 times (10,000%) downtime costs.
In cloud computing, these figures must be interpreted in different ways:
In the case of ITaaS (IT as a Service: Amazon EC2, GoGrid Compute, Azure Compute, eNlight Cloud Services), main service around which offers Cloud Consulting, HA is the availability of the hardware platform provider: in clear failure in the datacenter. Since your IT deport its own servers in the cloud, it remains responsible for software crashes mentioned above. However, the correction of a software crash is generally much faster (minutes to hours) than a hardware crash (several hours to several days).
In the case of PaaS (Platform as a Service: force.com, Google AppEngine), HA is the combination of the availability of data center provider and the software framework affixed above. HA is significantly worse (example: 99.9% for Azure Services against 99.95% for Azure Compute). What’s more, your IT should always assume management of crashes deployed over the PaaS software.
In the case of SaaS (Software as a Service: salesforce.com, Google Apps, Oracle OnDemand), HA is the combination of three levels of availability: the data center providers in India, the software framework affixed above, and Finally, the SaaS itself at the top of the pyramid. Despite this complex combination of HA SaaS is often better than a PaaS (if one includes the rate of HA software deployed over the framework), since SaaS is developed specifically by the supplier for its own platform, with optimum knowledge constraints. Conversely, your own deployed over a PaaS software incompatibilities may suffer.
In conclusion, we must remember that the promise of HA Cloud Computing is certainly required, but it does not become completely transparent or abstract for your IT .