The Future OSS, in Larry Dennison’s eyes

Larry Dennison calls himself a dreamer. In 1996 he helped dream up Avici Systems, a vendor whose high-reliability routers AT&T trusted with its core networks. When that routing business dried up, the MIT computer science and engineering Ph.D. helped dream up Soapstone Networks, which mined Avici’s control plane software for a pure play focused initially on carrier Ethernet networks. Since leaving Soapstone last fall, Dennison has gone on to found Lightwolf Technologies, a startup still officially in stealth mode that is predicated, according to its Website, on the assumption that the networks of the future will be a mix of optical and Ethernet transport and, atop that, software residing on commodity processing platforms. Lightwolf, which Dennison said should announce its first funding round sometime in the next six months, will focus on the need for a new breed of operation support systems (OSSs) made necessary by the emergence of several transformative industry trends, including the rise of cloud computing (or infrastructure outsourcing) and open-source software and the market for monetizing third-party applications. He spoke to Telephony’s Ed Gubbins last week.

On Lightwolf’s focus: It’s somewhere in the OSS space. It’s not an independent app that’s riding on top of someone else’s OSS. We’re definitely part of the OSS. The existing OSSs aren’t going to cut it in the new world. If I were trying to build an OSS play, that’s very difficult. I have to begin with a very simple unmet need, where you can become profitable, and then build. Over the next 10 years, I see lot of unmet needs. Right now, I’ve been building up the technological underpinnings that align with the market forces.

On the OSS of the future: Historically everybody has been buying [central processing unit (CPU)] capacity by the month or by the year. Cloud computing takes it a different way and says, ‘I need a lot of CPUs but only for an hour. I don’t want to buy for a month. I want to convert a thousand hours on one CPU into a thousand CPUs for one hour and do it interchangeably.’ That’s radically different than the existing virtualization model. You have to go from capacity-based accounting to usage-based accounting. All the performance monitoring, all the fine-grained usage things are going to get cranked way up -- the intensive scrutiny of did-you-actually-deliver-what-you-sold -- because it’s usage-based. From fixed virtual capacity, you’re starting to see more distributed parallel algorithms being applied. You’re also seeing people look at creating communities of independent developers who want to work on their platforms. Google, Amazon, Salesforce.com, IBM, Sun, HP – they’re all coming out with computation platforms that are richer than just a CPU with an operating system on it.