Scale breaks everything. When you move from thinking, “10,000 servers is a lot,” to buying hundreds of thousands a year, your perspectives change. And that shift in perspective has changed the information technology industry. For hyperscale users, for vendors and even for enterprises and startups hoping to figure out how they can take advantage of this shift to build their own IT organizations or the next Angry Birds.
A parade of IT industry stars shared their perspectives on scaling gracefully during GigaOM’s Structure 2012. Here are some of the highlights.
Hardware woes and “strained” relationships
Intel previewed its next generation scale out Avoton chips while folks from AMD and Calxeda laid out the need for a data center fabric that would encompass the chips and the network. On Thursday we heard from Facebook’s Frank Frankovsky about his plans for possibly using a subscription model for buying CPUs and utterly deconstructing the server.
He also laid out Facebook’s efforts to boost page load times for users by putting Facebook hardware in various telecommunications points of presence. He also admitted that his company’s efforts may leave his relationship with some hardware vendors strained. And for those who aren’t designing their data centers and servers from the ground up to scale out most efficiently, a panel of data center providers laid out a variety of ways to deploy data centers that contains hundreds of thousands of servers without requiring a power plant.
Sadly, Debra Chrapaty the CIO of Zynga thinks hardware providers and data center owners could do more. She said the company has only seen “slight evolutions” in terms of things like server cooling. “We’ve yet to see true advances like alternative-energy-run data centers,” she said. “It’s certainly a topic worth a couple of hours of discussion.”
Fabrics are fabulous and OpenFlow takes a hit
The show had several panels that discussed fabrics and what the idea of a software-defined data center would mean. Steve Herrod CTO of VMware (s vmw) said it would mean the end of “server huggers,” people who like to know exactly which box is storing their data. In a deeper conversation Martin Casado of Nicira explained how far we have to go to get to any software-defined data center, while Dante Malagrino of Embrane said that when we got there, the types of IT jobs would change to reflect the programmatic nature of the new data center.
“Software-defined networking was so 2011, now we’re talking about software-defined data centers,” joked Malagrino onstage. But software-defined networking and the OpenFlow protocol did have their moment when Ken Duda of Arista got onstage and slammed the technology for having a limited use case (lawful interception of personal phone and data traffic). “It turns out OpenFlow 1.0 is a very good fit for that problem: I haven’t found any other problems for which I believe OpenFlow 1.0 is a good fit,” he said.
More snipes and gripes from API acrimony to database drama
A group hug ended the tensions of the API panel, but beforehand things were pretty tough as Marten Mikos of Eucalyptus fought it out with OpenStack co-founder Chris Kemp of Nebula. Kemp likened the OpenStack effort to the Linux of cloud. “Don’t you mean the Unix of cloud?” countered Mikos referring to the fact that many companies are working on their own OpenStack implementations. “Or maybe it’s the Soviet Union of cloud,” he quipped, likening it to a collective farm allegedly run for the benefit of its members, but which tends to be inefficient.
That wasn’t the only slam at the event. Rackspace’s Lew Moorman took the stage to caution developers on Amazon’s (s AMZN) closed infrastructure, while Amazon’s Werner Vogels made sure people in the audience understood that Amazon wasn’t about to stop innovating and moving forward. Hold onto your business plans, AWS-related services.
Meanwhile Google’s Sundar Pichai called Microsoft’s Windows ecosystem “ossified” and explained that apps would be the key to the Chromebook’s success as Google takes the device into the enterprise.
Unsettling ramifications for cloud and compliance.
Finally, the idea of the enterprise snuggling up to the cloud had its cautionary moment at the end of the show, as worries about compliance trickled in. It began with Alexei Rodriguez, VP of operations for Evernote pointing out that Amazon Web Services is a bit of a black box, which is why he built his own infrastructure. The topic of security and compliance also came up with the launch of Bromium, a new company aiming to isolate malicious threats on devices using micro-visors. And it ended with somewhat of a bang.
“There is a good chance that almost every organization that is out there that is using Dropbox or that is using Box is breaking the law,” proclaimed Puppet Labs CEO Luke Kanies (see disclosure) during the last panel of the day at GigaOM’s Structure conference Thursday.
Kanies wasn’t out to scare people, but he had a point: Most companies don’t even have internal rules for the use of data with cloud services, save for a clear understanding of the law. Fellow panelist and enStratus VP of Product Strategy James Urquhart agreed, pointing out that courts have yet to device whether Fourth Amendment rights apply to documents saved in the cloud.
And that was a wrap for the show. Scaling is hard. The cloud is still tricky to navigate. And our current security and regulations aren’t keeping up.
Disclosure: Puppet Labs is backed by True Ventures, a venture capital firm that is an investor in the parent company of this blog, Giga Omni Media. Om Malik, founder of Giga Omni Media, is also a venture partner at True.