OCP's Carter: Open-Source Agility Begets Web-Scale Agility

Web-scale IT leads to agility, but how do you get agile enough to even deploy web-scale IT in the first place? Bill Carter, CTO of the Open Compute Project Foundation (OCP), has an answer -- and it starts with the word "open."

Previously, in Part I of this two-part, lightly edited Q&A, Carter talked about how open-hardware helps the deployment of web-scale operations and services. (See OCP's Carter: Open Hardware for Faster Web-Scale IT .)

Now, in Part II, below, Carter discusses the importance of web-scale IT in the carrier market, what telcos can learn from other enterprises about web scalability, and how web-scale agility is fed by a perpetual cycle of agility.

Telco Transformation: Why is web-scale IT important for network carriers right now?

Bill Carter: Well, I think it comes down to the learning that we've gone through with efficiency, scale-out, resiliency, cost containment, manageability and deployment; all the things that we learned from doing things at scale can directly apply and should directly apply. I think it's critical that they adopt that and that they embrace that. You look at open-source software and the disaggregation of hardware and software -- particularly disaggregation of the switches -- there's some pretty key learning. There are capex savings and opex efficiency and responsiveness to issues when you disaggregate the network switch. You can deploy things, you can manage it better, you can deploy quicker…That's the learning that web-scale went through, and I think that that type of learning is critical to the telco business, the carriers. (See Carter: OCP Seeks to Broaden Reach at OpenStack.)

For example, if you asked a web-scale guy about the number of people that it takes to operate their datacenter, you typically get an answer that is in terms of the number of servers that they can support o] the number of pieces of equipment that every person can support. The web-scale guys will say, "Hey, I have one technician for every 500 or 1,000 servers." And they're headed to a goal of one tech for the datacenter for every 10,000 pieces of equipment, and I've seen those numbers thrown out. On the other hand, you look at the carriers, and the amount of people that it takes to manage their infrastructure, it's not at that scale at all. And so they have to figure out how they get to that scale to be competitive.

TT: What might telcos be able to learn from the general enterprise in building out their web-scale IT operations -- if anything?

BC: The carriers for years have had facility requirements that were pretty strict in planning their datacenters -- and rightfully so. They didn't want to have hardware failures that resulted in dropped calls. The enterprise guys are trying to manage their business and they have business continuity goals. They have basically accomplished the same things in terms of business continuity without having such a strict set of requirements. I think, over the years, that the enterprises got to the same level of reliability in that they have chosen two different paths, and I think the carriers could learn something from that.

I think the other thing that the carriers can learn from enterprise IT involves the fact that, in some cases, they actually operate a public cloud service -- and I'm thinking of when Terremark was owned by Verizon. They were managing a public cloud set of services and at the same time they had people managing their network infrastructure. And from my viewpoint, they really managed it quite differently and they had different approaches to the types of hardware that they purchased, the companies that they worked with, the overhead, the resources they put in place -- all those things. I think that they could learn a lot by evaluating their management structure and saying, "Do I really need to have two separate organizations?" -- because enterprise IT doesn't have that. The enterprise IT guys are running a private cloud for their client internally, and in many cases they're also managing the factory; the factory data and the factory machinery is being ran by the same IT group. So they have internal infrastructure processes that they support as well as private cloud type services.

TT: To what extent does building out for web-scale IT require a certain amount of agility to get started? It's like you need a certain amount of agility to get to that solution to get the agility. Or do you disagree?

BC: No. I absolutely agree with that. You have to be willing to commit to the journey and put some resources, head them down the path, and put them on the journey to this transformation. And figure out what's working and what's not and how to fix it.

TT: In getting that initial agility started to get that agility necessary for web-scale IT, what are the missing ingredients?

BC: Obviously, open-source software is a key ingredient and open-source hardware is a key ingredient. And I think that there is another ingredient, and that is having an agile solution provider.

And by the way, there are many examples of solution providers that are embracing this open-source hardware and software and saying, "I'm gonna build solutions around that." And there are companies that sell or license software that runs on white boxes; white-box switch gear, white-box servers. And so they're a critical partnership as part of this transformation. They're going to help companies in this transformation initially, and maybe forever, but they don't want to take the burden of maintaining the software stack. There are companies that are already doing that, and recouping the cost of that through service agreements. So you're not stuck with the license; you're paying a third party to provide a service agreement for you.

— Joe Stanganelli, Contributing Writer, Telco Transformation