Arik Hesseldahl

Recent Posts by Arik Hesseldahl

Accel’s Ping Li Compares the Cloud to the Mainframe

Ping Li has a theory about cloud computing in the enterprise. Every time a new computing platform emerges, all the basic building blocks that made the first mainframe computer systems successful have to be there. He refers to this idea as the “cloudframe,” and it’s something he’s been thinking about a great deal in his role as a partner at the venture capital firm Accel Partners. Known best for its investments in Facebook and, more recently, Groupon, Li has been involved in several of Accel’s investments in enterprise-focused companies. Among them are Cloudera, the company that’s popularizing the use of the open-source software Hadoop to handle big database applications, and, more recently, Nimbula, which is building private clouds. He also sits on the board of Lookout, a mobile security firm that Accel invested in last year.

I caught up with Li last week to talk about the way he sees the cloud shaping up in the enterprise, and where he’s seeing opportunity as it develops. If cloud computing turns out to be the fundamentally new computing platform that many think it is, then there’s a lot of different pieces–Li refers to these as layers–that have to be assembled to make it work. And it’s in those layers where he looks for opportunities.

NewEnterprise: So what do you mean when you talk about the “cloudframe”?

Li: “I keep telling everyone that every time there’s a new computing platform, you basically have to re-create all the pieces of the mainframe. The basic building blocks of computing don’t change. You still need provisioning, management, security, and you need networking. The pieces may change, and different layers get merged into others. I’ve been calling it the cloudframe, for lack of a better word, which allows you to redefine all the layers, from the data layer to the storage layer to the provisioning management layer. Some of the layers that may have been separate before are being merged into one. So I’m spending my time figuring which are the layers that you can build a company around and which are the ones that become features rather than products.

So what are you finding?

There are two layers I’ve spent a lot of time time on, and where we’ve made investments. One is the data layer, and one of the companies I’ve spent a lot of time on is Cloudera, which is built around the Hadoop ecosystem. With the new types of data and applications that are out there, it’s really challenging the idea about whether the relational database is the right data management solution of the future. If you look at the big Internet data centers, say at Google or Facebook, the answer is no. There are no Oracle databases running those applications. Instead they’re using things like Hadoop and MapReduce and distributed file systems on commodity hardware. So I think the whole data structure is moving because the types and volumes of data are changing so much. The data is a lot more complex, and there’s a lot more of it.

There’s been a lot of suspicion and resistance in the enterprise toward letting critical data and applications out the door to run on someone else’s hardware. Is that changing?

I’m less rigid with my definition of the cloud. It doesn’t have to all run on Amazon or something like that. This cloudframe idea is really about extracting applications and services from underlying hardware and resources. It’s about having things available to you anywhere, and having them scale up quickly. They can exist on both sides of the firewall. When we launched Cloudera we thought everyone was going to want to run things on Amazon’s cloud. But most enterprise customers are running large installations of Hadoop on their own servers.

So the data layer is clearly one place where you think businesses can be built. What are some of the others?

Once you have the data layer set, there will be a lot of innovation on the storage side. We’re doing a lot of things around flash memory storage. We’ve invested in Fusion-io, which uses flash memory to build high-performance storage systems. Before, you had to have a storage area network to get the performance and availability you needed to run a big database application. Adding flash memory to your storage lets you expand a lot and do it efficiently. If you look at a lot of the cloud data centers, there’s not a lot of EMC and NetApps gear running in them. There’s a lot of commodity white boxes with Fusion-io starting to get deployed in them.

Last year you joined the board of Nimbula after Accel made an investment in it. That must be part of another layer you like, right?

That’s the application layer. Nimbula is from the guys who first built Amazon EC2. Now they’re actually building a private version of EC2, for the enterprises who for one reason or another can’t go to the public cloud. If an IT manager wants to deploy an app these days they have to put in a request and then he may get servers in six months. With EC2, if you want to deploy something for testing or whatever, all you need is a credit card and you can get it going right away. The Nimbula guys are building a technology that provisions and manages cloud services out of your own IT resources. A lot of enterprise IT guys I talk to say their bosses point to Amazon EC2 and ask, “Why can’t you build me one of those?” Now they can.

All the layers are all getting reinvented. Some will be a service, some will be on-premise. For a new platform to emerge, a lot of things need to come together, and I think a lot of things have been in the making for a while. The environment feels as ripe as it ever has.