The challenges of continuous integration in web-scale services and operations can require an open, software-defined approach, says Ian Hood, chief architect for global service providers at Red Hat.
Previously, in part one of this Q&A series, Hood addressed the perpetual cycle of success that the SD-WAN deployment relies on as a matter of network effects while in part two, Hood delved into the issues surrounding SD-WAN's role in video enablement. (See Red Hat's Hood on Perpetuating SD-WAN Success, and Red Hat's Hood Talks Security for SD-WAN.)
Now, in the third part of this series, Hood talks about the intersection of virtualized technologies such as SDN (including SD-WAN) with web-scale IT -- as well as Red Hat's role in web-scale operations and services. In the final installment, look for Hood to talk more about how web-scale IT helps enable LTE networks and the Internet of Things (IoT).
Telco Transformation: Last time, we spoke very extensively about SD-WAN. Can you explain how software-defined networking and virtualization intersect with web scalability?
Ian Hood: An interesting thing people tend to forget about is you've got this big massive data center from Google, and these billions of transactions happen daily inside that data center or across those data centers. But if you look at what we do on a daily basis on the network to actually make that search on Google, 60% of those happen on mobile phones, which is a call setup/teardown just to get on that mobile network.
So it's really translating that web-scale-okay application orientation to "Can I apply that same thing to scale up to handle the fact that the world has gone mobile and the world will go to IoT, and to handle that same kind of scale for those types of enhancements to the networking technologies? How do I take advantage of those continuous-integration, continuous-deployment approaches in spinning up not just applications, which are static from a networking perspective to actually delivering a network service -- be it a mobile call with data on top of it, or be it an enterprise service with firewall and intrusion detection and things of that nature?"
So that's where they intersect. It's really a process to deliver scale. In this case now the other challenge for the operator is to take that to a highly distributed architecture. So that's where the virtualization comes in to take advantage of technology, but it becomes a distributed compute problem instead of a compute problem in less distributed sites.
You're invited to attend Light Reading's Virtualizing the Cable Architecture event – a free breakfast panel at SCTE/ISBE's Cable-Tec Expo on October 18 featuring Comcast's Rob Howald and Charter's John Dickinson.
TT: What role does open source have to play in web-scale IT?
IH: Well, if we sort of take a look at what web-scale is about and what most people think of it as, it's a combination of all the things that the large over-the-tops and all the social media sites do on a daily and regular basis; how they keep up with that in a continuous-integration, constant-update, always-on world. And so from that perspective, if we take a look at how we develop open source and what kind of process we do to try and develop solutions and innovations in that fashion, then that's the same goal that we're on; it is to try to accelerate the way that things can be developed and delivered to the end customer. So it just goes hand in hand with what we've done with the enterprise with open source and what we've done to help trigger the over-the-tops to go drive even higher scale in what they do. This is just something that goes hand in hand to make that happen.
TT: When it comes to web scalability, to what extent are we talking about scaling up versus scaling out?
IH: Let's just take the reality of economics here. If you have to keep adding more horsepower and having more backup stuff to do what you do as you scale with the number of users, the number of calls per day you do, then that defeats the purpose of trying to scale to massive numbers in terms of, whether it's the number of searches per day, the number of things people send across on a video perspective, all those things economically take a toll on the data center's architecture.
So in order to make that so you can actually scale, I'll call it "total service capability," scale out is the approach most of the web-scale data centers are being built, so that they can take advantage of things in a horizontal-parallel mode. What they're doing is that they're taking advantage of a common architecture horizontally -- and then if hardware fails or a hard drive goes away, there's another one that picks up on its behalf and no one really notices.
TT: Do Red Hat's partnerships presently include web-scale IT? To what extent?
IH: Yes, because the partnership is in the communities that we jointly support to help things like the Cloud Native Compute Foundation -- so that's one right there that talks to that.
To get to this point, there's sort of the "Okay, how do I architect the hardware? Then how do I architect the software challenge?" But the main challenge is: "How do I deal with the complexity of trying to make use of that hardware for many applications?" And that becomes a service and application orchestration challenge. So as a result of that, that's where we tend to do the partnerships such as "How do we make [it] scalable, consistent, and easy to -- regardless of infrastructure underneath -- actually spin these up and down, ensure elasticity, [and] move them around wherever the hardware's available when we need to?" So that's where the partnerships lie -- mainly in the community. As for what people think of as the web-scale folks, we provide them tools that actually help them do that. One of the biggest ones is the acquisition of a company called Ansible that allows us to help them with that, and they are huge users of that technology.
— Joe Stanganelli, Contributing Writer, Telco Transformation