Understanding the basics of bare-metal switches
A comprehensive collection of articles, videos and more, hand-picked by our editors
In this Q&A with Big Switch Networks CEO Doug Murray, SearchSDN learns how the company's new bare-metal switching strategy is progressing.
In the early days of the SDN gold rush, startup Big Switch hitched its wagon to OpenFlow and encountered some major obstacles -- specifically a cartel of switch vendors who were slow to offer quality OpenFlow support on their hardware. Big Switch responded by embracing the overlay network model that rival Nicira proselytized (and converted into a $1.2 billion acquisition by VMware). Big Switch hired a bunch of people from companies such as VMware to advance the overlay strategy.
But Big Switch discovered last year that the overlay approach wasn't going to work, so the company hit the reset button in September. It embraced an SDN strategy centered on bare-metal switching, which had been pioneered by the Open Compute Project and software vendors such as Pica8 and Cumulus Networks. The company brought on Juniper Networks executive Doug Murray as the new CEO. Murray, who had been leading Juniper's Asia-Pacific operation, now leads a company whose fortunes rely heavily on Asia. Japan is a strong market for OpenFlow, and most of the companies that manufacture bare-metal switches are based in China and Taiwan. In this Q&A, find out what Big Switch has in store for 2014 and beyond.
What have you focused on since taking over Big Switch Networks?
Doug Murray: I took a step back and started to [look at] everything, from our open source strategy to our product strategy. The first step was that we can't be all things to all people because we're a small company. Now, our primary market is looking at the data center and how to simplify networking in the data center using SDN and fabrics.
The second thing we looked at is how we take our approach to products and the development of the products themselves. [Our] Big Tap application is a … great entry point because it's not inline. It sits out of band and helps stream traffic and packets off to NPBs [network packet brokers] and to monitoring tools.
But I wanted to make sure we didn't overwhelm ourselves with too many applications up front. When I came to the company, we had a little over ten potential SDN use cases up on the whiteboard. I wanted make sure we were prescriptive and focused in our approach. So we focused on Big Tap as our initial application.
And we have our cloud fabric, which is in large part the foundation of the company. That will come online in 2014; [it's]a cloud-switching fabric, which is focused on the 'unified P+V' [unified physical and virtual], the concept of the centralized controller that manages the physical [bare-metal] switches and virtual switches in hypervisors.
Also in my first 90 days, I was really looking at the culture and leadership of the company. I made some additions to the team and changes there. Just before I joined, the board brought in Prashant Gandhi [from Cisco], who is leading product management and strategy. After I joined, we brought in Jeffrey Wang [from Brocade] as vice president of engineering. He's built fabrics in several different companies. And Joe Palazola [from Juniper] joined us just a few weeks ago. He is leading operations and support. As we continue to ramp up our products -- as Big Tap gets more traction and we start shipping the cloud-switching fabric later this year -- having a solid support infrastructure was important.
Did those changes in leadership trickle down to the rest of the company? Did you have a lot of turnover among engineers?
Murray: The changes in engineering and sales ranks started to [happen] when the company made the shift from overlay to P+V because some of the people who had come on board came to build overlay products. We did have some turnover in 2013. Some of it was natural and healthy for when you go through a change like that. We brought in some new engineers. Over the last 60 days in particular, we've brought in more engineers and are investing on the product side. At this point I would call it stabilized and in a position where we are actually adding engineers.
The focus now is on rolling out an SDN cloud fabric. What does that involve? When I hear the word fabric, I think of hardware.
Murray: Clearly the emphasis for us is on software when we talk about the fabric. Big Switch 1.0 was where [our technology] was an overlay [with] a controller that used OpenFlow to effectively manage other people's devices. We're now using bare-metal switches from ODMs [original design manufacturers] in Taiwan and China, and we now have the ability to take our Switch Light OS and use that as the starting point to work with the fabric. We're trying to take what companies like Google are doing -- purchasing bare-metal switches -- and bring that capability to other data center companies. Google and Facebook have hired networking departments where they can buy bare-metal and then build their own software. ... That is really enhanced and tailored to their specific applications and their specific data centers. They can do that because of who they are. If you are Google, you can hire dozens, if not hundreds of people in Mountain View.
For a lot of other data center companies, that's a very difficult thing for them to do. Our vision is to take that Google-like model to everyone else.
You have Switch Light. You have ODMs on board. You have the controller and applications. What work do you have left to do before you roll out the SDN cloud fabric?
Murray: We do have all the nuggets today to make that happen. A lot of development focus is on a series of things.
We want make sure from an insertion perspective that we have architecture nailed, not just in the sense that Switch Light is sitting on bare metal and working [with our] controller, but really [that] the entire pod working effectively. We look at this in terms of a pod architecture, where you have a pair of [redundant] controllers interacting with hypervisor vSwitches and interacting with physical switches as well. The first step is … making sure we have secure multi-tenancy [and] a scale-out architecture, and that we do so with simplicity. We're base-lining using standards to make sure we have the resiliency and scale to do this. One of the challenges the company saw when pursuing the overlay model was that [it] usually didn't save that much in terms of cost. If anything, It could be arguably another network to manage. So we're very focused on cost. Also, [the overlay approach] didn't scale very well, so we're focusing on everything related to scale. How do you take the device out of the box and make sure the provisioning is simple? How do you make sure there is redundancy throughout the fabric and that the fabric scales?
We're spending a lot of time on the operating model itself. For example, we are optimizing the plug-ins for OpenStack and enhancing the GUI [graphical user interface] to get to the point where we can drive choice for customers. When they say, 'I want to use REST APIs [application programming interfaces] to hook into my existing systems; I want to use CLI [command line interface]; I want to use a GUI; I want to use OpenStack,' we want to make sure we have the ability to drive choice in terms of how we do that.
More on Big Switch Networks
Big Switch thinks Cisco has too much sway with OpenDaylight
Big Switch partners with everyone but Cisco and VMware
Fourteen questions to ask an SDN vendor before signing up
There is also a great deal of work to do with our HCL [hardware compatibility list]. We're working with Broadcom and hardware suppliers in Asia to make sure we have an effective support model in place for bare-metal switches and going through a lot of the elements there as we select which devices we're going to support initially. We want make sure they're tested and have gone through QA [quality assurance] processes on our end and not just on our ODM supplier's end.
At this point, we have five prospects that are live in trials with the fabric. A sixth will go live with a proof of concept in March. We're going take feedback from those customers in terms of where we are with everything from GUI management and OpenStack, bring that back in, do one final spin with code to make sure we have got it, test it all and then go out with the product. We're trying make sure we have this right.
On the hardware side, you're looking at Facebook's Open Compute Project for bare-metal switching?
Murray: Open Compute is a key part of our strategy as a company. We donated code to OCP [Open Compute Project] … [and] ONL [Open Network Linux]. If you're a company like Big Switch or one of the incumbent vendors, and you get one of these switches from the ODM suppliers, there are a whole host things you need to do before you actually start to test it with your software. If you have a Broadcom chip [Trident+ or Trident II], there are a bunch of baseline things you have to do with boot drivers and loaders. It's stuff everybody does every time you do this. As we were going through how we do Switch Light, we realized there are some nuggets here we can donate to the broader community to make it easier … to go to bare-metal. So we donated ONL to that community and the ODM suppliers have been very supportive, as have companies like Broadcom.
Big Switch recently announced version 3.0 of the Big Tap product. What's new in that version?
Murray: First, there is Ipv6 support. Japan is a key market for SDN. We are working with a carrier in Japan. We will go public [with that customer] soon. And so, having IPv6 and additional features is obviously important.
We enhanced the GUI. It's very clean and quite slick. You can actually look at the nodes within the fabric, what they are connected to. Hover over them and it tells you about traffic flow and provides a very good view in terms of how things work.
What's your financial situation right now? Are you going to seek another round of venture capital?
Murray: We're currently not in need of funding. We have an excellent board. We're privileged to have companies like Index Ventures, Khosla and Redpoint involved because they have a great background as well. They've been part of the first chapter of networking, so they've been instrumental in looking at what we are doing. Over the short run, from a cash perspective, we're doing well. When I joined the company, I looked at things such as cash flow. We are investing in engineering and will continue invest in go-to-market and marketing this year. Do I want do another round to ramp and propel the business forward? Absolutely, but it's not currently on radar.