Alistair Croll Is In(to) the Cloud

We had the opportunity to meet with Alistair this past spring, and have continued to value his perspective on trends in IT operations. Today we asked him some questions on his views of cloud computing and where it’s headed. Alistair is a lead blogger for Bitcurrent, and covers emerging web technologies, networking, and online applications. He is also the chair of the cloud computing track at Interop NY held in September.

StackSafe: Is cloud computing right for every organization? What do companies need to know, understand, and have in place before getting into the cloud?

Alistair Croll: Ultimately, a utility model for computing is right for everyone.

Of course, that’s a contentious statement. There are caveats. It’s like electrical power: Sometimes, you can’t use it; but it’s rare that it makes economic sense to build your own dam.

Ten years from now, I suspect people will be asking, “what do I have to run in-house?” instead of “what can I run in the cloud?” But it will take time.

One reason for this is that cloud apps are built very differently. To take advantage of true cloud computing technologies you need to rewrite your apps. This is one of the reasons why most public cloud applications are new ones; very few people are porting their enterprise ERP to the cloud as is.

Look at storage, for example: Instead of a relational database, which you scale by adding more machines, you have a service like SimpleDB, BigTable or CouchDB. These scale indefinitely; but they don’t work like a traditional database.

For “legacy” applications that won’t run natively in the cloud, companies need a software layer like Elastra, Rightscale, 3Tera, Enomalism, or many others to run their apps in a cloud environment.

So I would say companies need a five-year plan for cloud adoption, and they need to start training operators on the cloud with non-production, transient applications like training and testing first. They also need to give developers cloud sandboxes and be opportunistic about letting them build new things on Amazon, Google, and so on.

StackSafe: How does cloud computing affect downtime? Does it help or hinder?

Alistair Croll: If you build on resilient, scalable cloud systems, then in the long run, it can make it better. But there are still lots of moving parts, as recent outages at high-profile cloud operators have shown. There’s also a lot of debate over centralization — having one big cloud means it’s one big target, and there’s no “biodiversity” on the Internet.

StackSafe: What advice would you give Amazon on their recent entrance into cloud computing, and the downtime that followed? How can they work better in the cloud?

Alistair Croll: I think they’re doing extraordinarily well, to be honest. I don’t think I can give them advice; Werner and the team are very smart people. In the recent data corruption event, the people who were affected were those who were taking shortcuts and not checking data hashes.

A lot of developers don’t think like operators. Amazon needs to make people aware of things like global server load balancing. They could also open up their distributed servers a bit so people could make sure an application was located in several places, and handle failover that way. Today they offer Availability Zones for this.

I wouldn’t call Amazon a recent entrant — much of what it’s doing is based on years of in-house operations. What they’ve done is opened things up, and that reduces standardization that always leads to problems. Google, on the other hand, is limiting what people can do with its cloud (you have to use SimpleDB, you must write in Python, you have to go through the Google API) and this affords them better control. It’s a question of whether you want to adapt to your users, or have your users adapt to you.

StackSafe: What can companies learn from Amazon’s EC2 experiences?

Alistair Croll: Unless they’re also clouds, not much. But they can learn about operations and availability, and make sure they have a good networking person in house to help with stuff like load-balancing.

StackSafe: You gave five reasons why cloud computing is more than just hype, why do you think people are treating it as the latest trend?

Alistair Croll: Data center hosting has been a commodity business for a long time. This is revitalizing a traditionally boring infrastructure sector, which is great. Much of that revitalization is justified, but there are a lot of people co-opting what was once a fairly specific term to their advantage.

There’s also a lot of backlash by site operators. Clouds mean the engineer can deploy most of the app themselves, without asking someone in ops to help. Which is why so many people jump all over outages: They’re looking for reasons to prove they’re needed.

I think operators are needed, and that clouds introduce different kinds of challenges as I mentioned above. But if an operator thinks he’s going to be racking servers and checking MAC addresses in five years, he’s in for a big surprise.

StackSafe: What other trends are you seeing in the data center?

Alistair Croll: Now that it’s easy to make virtual machines, there are too many of them. Corralling those machines is a good idea, and ripe for development. Also, the prevalence of Internet-enabled mobile devices is causing new problems and eliminating the need for a desktop. I also think the shrinkwrap license model is in for some surprises — people are using a computer for a day, not a year, and don’t want to pay traditional license fees. It’s one reason open source has been so big in the cloud world.

StackSafe: How did you come up with the idea for UnConference at Interop, and what results have you seen?

Alistair Croll: It’s been a great event, partly because it’s unstructured. Lenny Heymann and I are always looking for ways to keep conference content fresh and engaging, and Unconference does that really well.

By bringing together end users, vendors, experts and educators — getting them off the podium and into the crowd — we have some great discussions. We wanted to offer just enough structure to get a conversation started, but enough latitude to let people take the discussion where they wanted. We’re really excited about doing the event in New York, particularly with the new themes of mobility, social networks, and cloud computing.

Popularity: 5% [?]

Filed Under: Cloud Computing, Downtime, IT Operations, Interviews, Interviews-Bloggers