One of the oft-stated promises of cloud-based applications is more up-time, more application availability than those deployed via on-premises servers. Some might take the statement at face value; others may question it. I’d like to make the argument that both sides are right.
First let’s agree on a definition of cloud (if just for the time it takes for you to read this blog). Some argue that a single enterprise provisioning applications to their own operations from an owned and operated data center is cloud. I think this is best labeled “private cloud.” Here when I talk about cloud I’ll be referring to multi-tenant applications delivered from a provider’s data centers.
As it turns out, this distinction between private and more public clouds helps answer the question of whether cloud application deployments can deliver more up-time than on-premises deployments. Up-time has costs associated with it. 99% uptime is less expensive to achieve than 99.9, 99.99 or 99.999%.
Private, premise-based uptime: No small investment
A company creating and deploying applications in their own data center makes a conscious decision about uptime as they construct their private cloud plan. In order to get more availability, more reliable and expensive equipment needs to be put in place. Delivering higher uptime also means buying not only the best equipment but also the easiest to repair. Then there is the cost of building redundancy, buying multiple sets of the best equipment. Finally, there’s the cost of deploying, housing and operating multiple sets of equipment in geographically diverse and secure locations.
Sound expensive? It is. Financial services firms (think American Express or Goldman Sachs for example) have the in-house expertise and financial wherewithal to build private clouds that can deliver all the bells and whistles of cloud-delivered applications, e.g., the best equipment, secure data centers staffed with highly-trained personnel, geographi- cally dispersed around the globe. Governments also have the resources to create private clouds with peerless availability, like the US military or Internal Revenue Service.
But let’s say that the cost of equipment is a surmountable obstacle. Finding the right people, with the right expertise, to deliver superior uptime can be an even greater challenge. Many initially assume that the existing IT organization will build and maintain a private cloud. Often, however, companies find themselves having to engage specialized consulting firms because their in-house personnel don’t have the required knowledge or experience.
Another issue with delivering acceptable uptime is the location of the private cloud deployment. Again, it is often assumed that the private cloud will just take over part of the existing enterprise data center. But the required networking connections often drive companies to third-party facilities, e.g., from Savvis or Terremark.
Public clouds: shared uptime, shared costs
A public cloud application provider, like Five9 or Salesforce.com, has all three of the elements required to deliver industry-leading uptime: best-in-class, redundant equipment, cloud IT professionals working 24-7, already working in certified data center facilities. And their services can be delivered at affordable prices because it is built as a shared resource across many customers.
So can a cloud application provider deliver better uptime and availability than an on- premises deployment? If we include in the question the notion of “at a reasonable cost,” then the answer is definitely yes. Most organizations don’t have the financial and human resources or physical plant to deliver a five-nines-uptime guarantee. But an organization dedicated to delivering cloud-based applications can, because they amortize the cost across a large number of customers.
One word of caution
Not all cloud application service providers are created equal. As you investigate taking applications to the cloud, be sure that the provider you choose has deployed the best equipment, has the most qualified personnel and offers the geographic redundancy and in-production upgrade procedures that deliver on the promise of best-in-class uptime.