The company, whose software automates the tedious tasks IT workers have traditionally undertaken to make sure networks and servers are running up to speed, cracked the $100 million revenue mark in 2006.
Opsware has achieved this plateau by winning such customers as GE, Target and EDS (Quote) , as well as through a key reseller deal with Cisco Systems (Quote) that gives it deeper penetration into territory eyed by larger rivals IBM (Quote), HP (Quote), CA (Quote) and BMC Software (Quote).
Ben Horowitz, Opsware CEO and co-founder, also put his own prints on the industry, having spent time at AOL before joining Andreessen at Netscape. Now the two have their eyes on the next stop in the world of the Internet. Can anyone say Web 3.0?
Internetnews.com caught up with Andreessen and Horowitz at Opsware's analyst day here to discuss the company's market opportunities and plans.
Q: Marc, you mentioned in your opening discussion that Opsware is having success versus the larger vendors -- IBM, HP, CA, BMC -- because Opsware has a more modern architecture suited for today's computing. But those vendors are also advocating the service-oriented architecture (SOA) model. What's so different?
Andreessen: They're fine companies and they have fine products, but the challenge is IBM has Tivoli, HP has various products, CA has UniCenter, BMC has Patrol. They have other products that do service-oriented architecture but those products that we would in theory sell against were architected in the 80s.
In the 80s, the typical IT infrastructure was a fairly small number of servers running primarily database software, supporting desktop applications, running primarily Windows 3.1: the fat client desktop architecture on a relatively simple network, with relatively simple network routers and switches -- NetBEUI, NetBIOS, Novell Netware. That was the state of the world. That's when those products were architecture. Fast forward 20 years. Here we are in 2007 and what's the world like today?
Large-scale commodity servers running 64-bit multi-core CPUs running Linux and Microsoft OS, which didn't exist 20 years ago, running application stacks like Java and PHP, which didn't exist 20 years ago, running Web-based applications, which didn't exist 20 years ago, being used by people on browsers over complex networks with wireless, VoIP, firewalls and all these new networking technologies and with complicated storage technologies that never existed.
Historically, what's happened in this industry -- when there is a re-architecture -- there is a complete turnover in the products. That's happened before.
Today's products took over from the mainframe products of the 60s and 70s. Basically, the exact same thing is happening again. We say this, and it sounds ballsy when we say it because IBM and HP are big important companies ... we say this because we made more than $100 million in revenue last year from a standing start four years ago.
The vast majority of the customers we sold to that are more than 10 years old themselves already have licenses to Tivoli, or UniCenter or Patrol and yet they're buying Opsware. That's why.
Q: You also mentioned virtualization in your presentation. IDC recently said the number of physical machines is decreasing thanks to virtualization. Does Opsware see that trend?
Andreessen: That's ridiculous. I totally disagree with them. I use the term "elasticity" from economics. That is, there is a more than one-to-one increase in the things that you'll do.
If the cost of running an application on a server goes from $30,000 to $3,000, people are going to rollout more than 10 times as many applications. It'll be the same thing when multi-core kicks in, when virtualization kicks in.
The physical server price will start the fall; the price of power consumption will start to fall. Applications are getting way more complicated and consisting of more servers per application than before and operate at a much larger scale. I think you're going to see a huge rise in virtual servers and physical servers to handle all of the data.
Next page: Horowitz on the value of Opsware
Q: You also discussed how this will mean more virtual servers for Opsware to manage, but does Opsware fear push-back from IT guys who don't want to pay for management of each operating system instance on a machine?
Horowitz: Well, [the IT manager's] cost is driven by how many operating systems he's running. So how many versions of Windows do you need patched is what his cost is based on, so it's pretty natural for us to price against that.
The value of Opsware is how many fewer operations you have to do on, for example, Windows. So how do you measure that? And whether there are six instances on a virtual machine or six on six different physical machines doesn't really figure into the customer's mind in terms of the pricing. Our customers have been pretty comfortable with us pricing on a per virtual instance basis.
Q: Is Web 2.0, with all of its mashups and wikis, driving this application explosion?
Andreessen: Perfect examples. Wikis, mashups, mapping applications. There's the new generation of all of these video-sharing sites. These new generation of startups doing video editing as a Web service. Doing the equivalent of Adobe Premiere or Final Cut Pro but doing it in a Web site. All of this social networking is generating content.
These are all applications that, speaking as someone who was in the dot-com bubble, that were not economically feasible to do in 1997. You had to buy huge expensive Sun servers with Oracle databases and EMC storage. They are economical now with Intel or AMD multi-core CPUs running virtualization software, running Linux. It's a great example of elasticity; these ideas are taking off in huge volume that have become economical to do.
Q: What's the value in buying the run-book automation provider iConclude?
Horowitz: Run-book is the first application of process automation. The run book is the set of processes the networks follow when something goes wrong. Something comes in from the network operations center and you've got to react to it. So, different things that happen if you're rolling out a new application. You've got a security vulnerability alert or you're making some kind of performance change. All of these are processes throughout IT that require coordination among networks, servers and storage.