Battery Ventures has promoted Itzik Parnafes to general partner as the firm continues to make investments in the enterprise IT sector. The venture firm and Parnafes is betting that the wave of change in IT brought about by large-scale data centers and cloud computing is still providing solid investment opportunities.
Parnafes, who is based in Israel, was previously a partner who led Battery’s investment in Delphix, a company that makes software to virtualize databases. Parnafes was also on the board of XtremIO prior to its acquisition by EMC. “The guys keep asking me if our investment in enterprise IT is done yet, but there’s still a lot of opportunity here,” says Parnafes.
Battery has been one of the more active venture firms investing in the hardware and infrastructure side of our broadband-influenced and consumer web-obsessed culture with investments in Calxeda, the ARM-server company, Anobit, the flash memory firm that Apple purchased, and Cumulus Networks, a company rethinking switches. It has made more than 20 investments over the past five years in enterprise IT, and put more than $240 million in.
What’s going on in enterprise IT–or, frankly, IT.
For every Facebook and Pinterest there are hordes of servers and IT personnel trying to figure out how to scale these web-based businesses at costs that still enable them to make money. On the web, your infrastructure and software are the raw materials that comprise your cost of goods sold, so many of these firms are looking to make their infrastructure cheap and elastic.
Some buy from cloud providers like Amazon, while others graduate to hosting their own servers and building their own data centers. But this trend hasn’t exactly been a boon for the legacy IT guys as open-source software and commodity hardware have been deployed across hundreds of thousands of feet of raised floor inside data centers. Scale breaks things, and in the last five years venture firms have been investing in software and hardware that helps bridge the gap between legacy and a new hyperscale architecture.
But as Parnafes says, there is still opportunity, and that opportunity is in rethinking the architecture itself. Google is doing it with a team of secretive scientists, as are some in the cloud and high performance computing world. Once you build a compute architecture that’s elastic and on-demand, the physical infrastructure constraints — which remains inelastic and isn’t deployed on demand — become a source of pain.
Peace, love and interoperability in next-gen IT
It’s like once you move to a house in the suburbs you realize that you love the lawn and schools, but you hate the commute. Computing and data centers are now realizing the time they spend provisioning and futzing with the hardware is their equivalent of a long commute. Hence the current vogue for software-defined everything, the rise of Mellanox, which can support both Infiniband and Ethernet networks, and even the purchase of Xsigo by Orcale. The end goal here is a compute architecture that can be just as adaptive as the virtualized services built on top of them. What’s more, workloads and data will be able to flow from one data center (or cloud) to another seamlessly.
“These are the prime motives that direct our thinking: what will be in the next generation data center,” Parnafes says.”Look at every constraint you encounter today and try to eliminate them, and if you identify such a technology you have a possible winner on your side.”
Parnafes acknowledges that we may never get to this boundary-less world, but he suggests that regulations as opposed to technology or business models will be what stands in the way. To bolster this idea, he points to the CIOs of large banks which have seen the cost efficiencies of running an Amazon-style compute cloud and are trying to emulate Amazon’s costs structure and elasticity as opposed to buying into some vendor-defined private cloud model.
“It’s kind of an ironic twist,” Parnafes said. “In the past the public clouds wanted to emulate the private data centers to make people feel comfortable and now the private ones want to emulate the public clouds which have very efficient model for people to run their [compute] jobs.” So as these worlds not only collide, but are coming together in some sort of software-defined and virtualized lovechild, startups that can help ease that transition should give Parnafes a call.