TwinStrata

“If we’re not using the APIs, we aren’t using the cloud.” That’s what he said, and that’s what I can convince myself to believe every other Tuesday. (That’s when I like to exercise my self-skepticism.)

What’s the context? I was involved in a discussion with a prospect implementing a private cloud, and he was dead-set against using the cloud either as a file server or as block storage. And as I said, I can see his point of view, every once in a while. After all, block storage protocols like SCSI or ATA are somewhat limited in scope, and web APIs offer interesting capabilities, like tagging and versioning.

But the more I think about it, the more that I realize that the provocative declaration “REST API = cloud” is only telling half of the cloud story. Sure, it’s got the more public history to it; after all, the cloud only began to emerge as a buzz-worthy concept after big Web vendors like Amazon started providing programmatic interfaces. And there’s still plenty of confusion and uncertainty about what the term “cloud” even means, so putting a stake in the ground seems like it might be a valid way to break from the quaint old ways. Plugging a computer into a disk drive! That’s so 1990’s! ¹

In truth, the notions that the cloud embodies arose from two different camps, with two fundamentally separate sets of concerns. On the one side, you have the cloud as a natural extension of the WWW: a new way for people, communities, and companies to share information and work alongside each other, thousands of miles apart. That’s the kind of cloud that’s exemplified by companies like DropBox. On the other side, you have the cloud as a natural extension of the utility computing environment: an elastic resource that grows when you need it, shrinks when you don’t, and leverages the economies of scale to drastically reduce overheads. The best exemplar of the utility storage cloud, I’d say, is us.

There is a substantial distance between these two camps. Issues that are a minor concern for sharers (communitarians?) are absolutely unacceptable to utilitarians. Likewise, when we take security as a primary design concern, we deliberately lop off whole branches of possible sharing features. Sure, we’re working on ways to leverage in new and novel ways more of the massively scalable back-end that the cloud provides, but we always view the goal as “plug it into the wall and forget about it”. You don’t have to buy a new generator every time you get a new dishwasher; why should you need a new array every time you get a new server?

I won’t spend much time rehashing the differences between sharers and utilitarians. There are valid reasons to integrate both models into businesses and personal lives. There’s no reason they can’t coexist and even cooperate. But to declare either model as the definitive cloud is short-sighted and a little silly.

Except on every other Tuesday.

¹Here’s an easy way to tell that you’re talking to an old storage guy: we don’t plug disk drives into computers, we plug computers into disk drives.

Bookmark to:
Del.icio.us | digg | Twitter | MySpace |
FaceBook | reddit | Stumble Upon |

Nicos was giving a talk recently, and made a statement along these lines: “With the hybrid data center, the whole is greater than the sum of parts. If your data is stored locally and augmented with cloud storage, then you’ve got a system that’s much more reliable than either one on its own.” (He uses a lot more words in Part I of this series.)

Now, since I’m a geek (no, really, it’s true!) I can’t just take an adage at face value. And I’m thinking, “Well, of course not, if you’re talking about reliability. Really, the whole is much more dependent on the product of its parts.” And so I wrote down this equation:

Where did this equation come from? ¹

In our hybrid data center, to lose a chunk of data, you’ve got to have simultaneous failures in both the local data center and in the cloud. If you’re using a cheap SATA drive locally, with something like a 99.9% uptime (three 9s), but you’re using a cloud provider that advertises 99.999999999% (eleven 9s) durability, what does the whole system look like? Plug each part into the equation, and you get 99.999999999999% (fourteen 9s) durability: much better than the sum of its parts.

That equation also points to a reason that I am uncomfortable with the marketing category of “cloud gateway”. A true gateway product wouldn’t have the same multiplicative effect. At best, it would simply pass on the reliability of the cloud, but at worst it would introduce new failure dependencies. The whole actually would be the sum of its parts:

The hybrid data center, with local and cloud storage working together to ensure greater data availability and durability, is the target that we’ve always been working towards with CloudArray, even if that’s not what we’ve always called it.

And you know what’s really cool? That’s what we’ve done.

¹ Where did this equation come from? Well, we’re used to talking about uptime and reliability in terms of 9s, e.g. “this array has five 9s of reliability, with 99.999% uptime!” That is, at its heart, a probabilistic statement, and means that the probability that the array will not experience a failure is .99999, and the converse is that the probability of failure is 0.00001.

Now, one of the principles of probability theory is this: if you have two independent events that might occur, and you want to measure the probability that either one happened, you add them together. But on the other hand, if you want to measure the probability that both events happened, you’ve got to multiply them.

Bookmark to:
Del.icio.us | digg | Twitter | MySpace |
FaceBook | reddit | Stumble Upon |

I ’m sure you’ve seen the recent negative press on cloud providers.  There have been several newsworthy events in the past few weeks and I wanted to take the opportunity, in my inaugural post on the TwinStrata blog, to add a bit of perspective.

Perhaps overlooked in all the discussions of the Amazon outage was the fact that Amazon’s Cloud storage services (S3 and RRS) were not affected, no data loss, no customer impact.  One of the fundamental characteristics of S3 and RRS, that which makes it such a viable target for offsite storage, is that there are multiple, geographically disbursed, copies of the data.  A single site outage, such as the one in Virginia, isn’t sufficient to bring the entire system offline.  The bottom line, for cloud storage users like TwinStrata customers, is that the redundancy of cloud storage helped them weather the outage without a hint of a failure.

Lucas Mearian stated in :”What happens to data when your cloud provider evaporates” that there is no way to directly migrate customer data from the cloud in the case of a major outage. While the statement regarding “directly migrate customer data” is true there are products available that can do so indirectly. While it may not be possible to migrate data directly from Vendor A to Vendor B it is certainly possible to move that same data via an intermediary.  By using an intermediary the whole point regarding data outages and data mobility becomes moot since you do have the ability to select multiple providers and/or move data from provider to provider or location to location.

Cloud is just the latest delivery model for outsourcing.  Thus the term “CloudSourcing” , which I rather like.  There’s no magic here, folks.  It’s just a delivery model.  Yes, it’s a Utility Model but in IT we’ve been talking about a utility model for close to a decade.  A rose by any other name…  Grid, virtualization, outsourcing…  Cloud is evolutionary, NOT revolutionary, and it’s here to stay.  The trick is to keep your eye on the shifting sands.  Use the technology where it makes sense and don’t use it where it doesn’t.  There are many compelling reasons for using Cloud technology, regardless of how you define “cloud”, and I would argue that it is a question of when, not if, you will adopt it.

It’s inevitable that there’s fallout in the Cloud space.  Remember what I said about shifting sands?  There’s always fallout when technology is changing and, in IT, change is a constant.  This is a good thing: a sort of technology Darwinism.  It’s a simple fact that there will always be casualties in an evolutionary model.  That does not mean the sky is falling.  It means that another competitor did a better job or didn’t make some particular mistake.  It means that the survivor is stronger than the fallen, which is good for you.

Hype and exaggeration can be bothersome, but the overall message is a good one.  The Cloud is maturing.  The Utility delivery model is becoming reality.  Instead of being slaves to our equipment, we in IT can spend a bit more time solving the business problems our equipment and services are designed to address.  Instead of worrying about how much storage I need to allocate to this LUN, how much memory and how many processors do I need for this production server; instead of worrying if a server I build today will last three years; instead of all that, I can focus on getting the job done.  This is the value that cloud brings to your business.

Bookmark to:
Del.icio.us | digg | Twitter | MySpace |
FaceBook | reddit | Stumble Upon |