The “Magical Cloud”
- Date: 9 May 2011
- Author: Scott
- Category: Services
Does the world really need another blog in response to AWS’ EC2 outage? No, not really. The media blitz that has ensued is quite disappointing especially given the fact that ANY service provider can have an outage. Amazon, Google, Rackspace and Microsoft are all incredibly innovative and reliable service providers. Many lessor providers would wilt under the scrutiny that Amazon is being exposed to currently. News flash – Amazon never promised that they would not ever experience an outage. Truth be told, the previously mentioned organizations are better at providing availability than 99% of their customers. A recovery plan can’t be: “Well, it’s in the Cloud, so it’s all protected and redundant – right?” Wrong. Have you read the SLA? Did you understand the terms and conditions of the contract?
The magical properties of the Cloud are just about as real as unicorns!
Underneath the billowing fields of cottony soft Cloud marketing, there are real data centers, with real equipment, managed by real people. The Clouds’ omnipresence has billowed up to a point where perception is: it cannot fail – which is completely disingenuous at best and an outright lie at worst. Most IT professionals already know this. ANYTHING can fail and ANY provider can have an outage. The old saying, “It’s not a matter of if, but when…” is not made obsolete due to the Cloud.
Investors are pushing for valuation while consumers are clamoring for near-free ubiquitous computing, storage and bandwidth. Free is not a sustainable business model – it costs real money to provide these highly available, never-fail platforms. In some cases, Cloud providers have other means by which they derive their revenue, but in the long run, free is NEVER sustainable.
Somehow, this has created a situation whereby some IT professionals, ignoring their experience and common sense, go all in on moving to the Cloud. It is fine to move to the Cloud to leverage a service provider’s ability to provide quality infrastructure, expertise and reliability on a scale that you would not be able to afford or replicate yourself – that is exactly why you should consider it. However, moving your infrastructure, apps or data to the Cloud does not magically provide you with a recovery plan.
David Chapa says it well in his recent blog post: “if cloud will be an extension of your operation and if all of your data resides in it, you should have a contingency plan in place and you should test it regularly.”
In my mind, there are two primary levels of availability that IT professionals need to contemplate:
1.) High-Availability (HA): A 0-1 hour recovery time objective (RTO). An HA solution requires replication… period. If you have 20TB of production data and expect to recover to a local appliance (what happens when your “local” goes away? – think tornado or Katrina) or worse, recover from tape-based media, you’re going to be faced with a harsh reality. There is physicality to data – no behind-the-scenes magic or pixie dust is going to help you recover 20TB quickly without replication.
2.) Business Continuance (BC): A 2-72 hour RTO. This will require a blended approach utilizing replication, virtualization and DR-Grade data protection.
Anything longer than 72 hours is what we consider a disaster RTO. Both of the above approaches are part of a recovery PLAN. PLANS have to be TESTED. Without TESTING, plans are meaningless. Obviously, we could talk forever about the finer details of either of the two approaches, but I hope my point is clear: The Cloud does not provide a PLAN. The Cloud does not magically mean you will never experience an outage. If it absolutely cannot ever go down, expect to pay your Cloud provider for replication and/or redundancy.
Actually, on that note, expect to pay your cloud provider. Why? Because the Cloud is not free and repeat after me, “Free is not a sustainable business model!”

Comments
Comments are currently closed.