Andy Brown, one of the leading visionaries of IT infrastructure on Wall Street, joined us for a conversation on topics including the current state of network design, the hyperscale versus enterprise mindset, some signals that network innovation is sinking in and navigating networking career paths.Below are some transcript highlights from a fun afternoon. Andy, thank you.
You’ve been a vocal leader in the infrastructure community for a decade now, and have been a part of some of Wall Street’s most advanced designs. From the CTO chair, what is the state of innovation in networking today?
Having run one of the first virtualization initiatives on Wall Street, one of the things that I can tell you is, we kind of did networking but not really. We did server virtualization, and we introduced this whole notion of thin provisioning of storage back in 2006 in that particular implementation. In terms of the networking, we really didn’t implement a cloud centric or virtualization centric networking solution.One of the things that has been both frustrating and interesting since then is that server virtualization and storage virtualization have continued, one might argue that the innovation in server virtualization has moved a little (20-25%) maybe since ‘06 or ‘07, storage virtualization has definitely improved and some of things that 3par did and now kind of common features found in EMC, Netapp and other products. Both in network and security, you could argue that they are lacking significantly the level of automation and let’s say the amount to which policy is driving the outcome over server and storage. That is a bit of a frustration and in a way it’s allowed the complexity that is existed in enterprise data centers to persist.
We’ve spoken a number of times on the mindset shift from enterprise IT to hyperscale design. Care to riff on that?
A: So there is almost an east/west divide, and as someone who roughly half my time in west coast, roughly half on east coast, I see it really clearly. It is almost a paradox. All these east coast companies want auto scaling and all the other capabilities of cloud that you can get from building from on top of cloud but then there is still some view underneath that somehow or the other that the cloud can’t support transactions. There is definitely still a view on that, east coast mentality perspective. The west coast guys look at the east coast guys and they say “Why aren’t you adopting this? It is so much cheaper, faster and better…”I’m not saying it is an easy bridge to cross. If you look at any of places where I’ve worked recently you are taking about 10000 applications they are all different. They have been designed differently; they have a different ways of doing disaster recovery, different ways of doing state and storage replications. The challenge I think is, how you squeeze out the cost from those kind of applications, automate as much of them as you can from both the infrastructure and the application side, but stop adding more, still digging the hole, stop building more apps that look like infrastructure and start building apps that use a consistent infrastructure for areas like disaster recovery, data replication, common storage and so on.
So what are some of the first steps we’ll see from the enterprise side as the hyperscale design philosophy grows to embrace a broader community?
Networking at the enterprise level has not yet gotten integrated in to the process of how you provision apps or cloud. However, in the hyperscale-style designs [centralized controllers], you have far few boxes to manage and you can manage it like a wholesaler then a retailer. That is kind of the way I think about networking -- it has not really become integrated into the process of how you provision apps or cloud.
And on the vendor side?
Someone will figure that out how to take application design patterns and network blueprints and network templates and create automated provisioning and whoever does that is going to be very successful. One interesting point is that I see the large vendors doing the opposite of that.
In the midst of all of the change going on in data center design and on the networking landscape, any parting words of advice for network engineers today thinking about their careers?
If data-center networking is the thing that you want to be good at, then you have to start learning about storage, and service, and security and get out of your silo and start talking to people. Communication is still one of the biggest challenges, I think, in running a large organization and when you are embedded in the line (silo) and you are being given a thousand things to do every day, taking the time to start to understand what ecosystem you are part of, who is part of your supply chain, who is upstream from you, who is downstream from you is important.