The World Has Changed But is it ready for the Criticloud?
Clouds are everywhere. Actually, one might think that not only are clouds everywhere, they can also be anything… I’ve attended a number of conferences and met with many IT professionals in the last few months, and in all of these encounters, without exception —in all of them, the cloud part of the discussion was prefaced with a joke, a movie clip, or an apology that basically said ‘there is no chance that we all understand the term cloud the same way, so let me explain what I mean by cloud…’. Amazing, isn’t it? This must be the first time that hype was so open to interpretation.
So now it’s my turn to try to explain what I see in all of these clouds clouding the IT skies.
For a cloud to be a real cloud, be it private, public, hybrid, or any other, it must adhere to a number of principles:
- Utility / elasticity — the notion of consuming applications and IT resources in an infrastructure agnostic manner. You need more, you consume more. Any dependencies and configuration changes are hidden from the customer
- Mobility — your workloads must be able to move around (for whatever reason) fairly easily
- Mix & match — a real cloud cannot support only one technology or one vendor. Just like you would not expect Google or Amazon to support only Windows and not iPhones. A cloud servicing an enterprise must be able to support all of the technologies running in that enterprise
- Performance — if you need high performance, you should get high performance from the cloud. You may be charged for it, but that’s ok — pay by the drink. You request it and you get it (and pay for it…)
So where do we stand today? The industry is making significant progress in each one of these principle areas. There are a good number of real public cloud providers, as well as internal clouds, providing services and creating offerings that are very close to what we expect from a cloud.
The challenge, as I see it, is in the Mission Critical Cloud, which we can nickname— the ‘Criticloud’
When I was in Kashya / EMC, we dealt mainly with the requirements of mission critical applications: applications with strict RTO / RPO requirements, driven by the business’s dependency on these applications. Take airlines for example, they can’t operate without their applications. If their applications are down, even for a short while, they can cause delays that in turn will trigger a domino effect: flight cancellations resulting in compensation for people on cancelled flights, who in turn rebook on alternate airlines, leading to the loss of customers who decide to fly with someone else next time.
Current cloud technology is not there yet. I know, everybody is trying to position themselves as if they support full enterprise applications rather than VMs. But try to implement mission critical applications in the cloud and you will see it won’t go the full distance. Not because of any single reason, but rather due to a combination of reasons rendering a mission critical cloud to be both highly complex to manage and inflexible:
- BC/DR (my favorite subject, considering my background) — when you take into consideration BC/DR requirements, at the application level, suddenly you see that mobility is disabled, or scalability is not as ‘effortless’ as it should be
- Performance — You need your high throughput application to spread across many hosts and LUNs, to ensure maximum performance. Can you configure it this way without manually drilling down to every single layer? Can you still maintain your flexibility?
- Heterogeneity — What if you have two data centers: one running EMC storage and the other NetApp (normally the result of an acquisition). How simple is it to manage them in a single cloud? In theory this should be easy, as applications use virtual disks and not real disks, but in practice…
One of the common threads here is, once again, the storage layer. At the server layer we have a single vendor, practically dominating the enterprise cloud space, providing a single platform that does a decent job at answering all our cloud requirements. Not perfect by any means but it looks like it’s heading the right way. However, in the storage layer we still don’t see any real shift in products architecture or design that will enable the storage to adhere to the cloud principles. And on top of that, different storage vendors have different approaches to closing these gaps…
Criticloud? Not yet, but we are on our way.
As everyone is saying, it’s a journey to the cloud. The journey is being slowed down, not just by implementation issues, but also by technology gaps. More and more vendors are addressing these gaps, and I believe we’ll soon have the whole puzzle figured out. How soon? Good question.
And as we journey to the cloud, all of us at ZeRTO would like to wish you all a happy holiday season and a great new year!