But….as a week full of SDN will be starting shortly, here are some other thoughts relating to the topic of SDN. Many of which could be controversial =).
Understand default gateways and where and when they exist. With technologies like Fabric Path, TRILL, STT, and VXLAN, it is actually rare for product documentation to offer up *good* details on the L3 gateway for some reason. Why? If there is lack of support, that is okay, but be upfront about what the technology offers, and what the limitations may be. After all, if the focus is east-west traffic, L3 gateways may not matter at all for some clients.
Understand the proliferation of networks or overlays that will arise if you go down the path of network virtualization through the use of SDN. With server virtualization, it became extremely quick to provision new servers resulting in even more server sprawl than before, but because of tools such as Vcenter, it was manageable for IT teams to do more with less. Expect the same from networking. Years ago, every department and application owner wanted their own switch, but due to the capex of physical switches and the rise of VLANs, each tenant received a VLAN instead, and the business “suffered” because they had to share hardware. Looking into the future, instead, this will be accomplished in software with overlays or virtual switches and will end up being natively more secure than VLANs. Virtual Networks will be able to be provisioned just as quickly as servers are today. You could easily go from managing 100 physical switches to 500 virtual or from 100 physical to 50 virtual based on the actual requirements of the end users. So, be ready for it and make sure you really evaluate the provisioning tools for the overlays. Hope there is integration to the hypervisor mgmt tools to further simplify things too. As stated just above, if you’re deploying SDN based L2 virtual switches, make sure you know how Layer 3 works and how the heck your traffic gets out of that virtual LAN segment.
Understand the future of any SDN company or at least try to. Are they already working on other products or are they focused on making the ONE they have work? Since it’s still early in the world of SDN, look at the track record of their leadership team. Where have they worked prior? What about their funding? How much have they had? What stage are they in and what VCs are they working with? What are the VCs track records?
Cisco dominates. Everyone knows it (right?). The Cisco competitors are making a living by having a slice, maybe two, while Cisco enjoys the other six. With regards to SDN, much of the messaging needs to be improved. I think many companies are so focused on bringing down Cisco rather than focusing on the benefits of software defined networking. Remember, it takes time to develop an extremely hefty appetite. You don’t get fat over night.
This leads me to another point regarding controller based networking, i.e. one path to SDN (or is it the most popular path so far?). After much thought and internal debate, I don’t think “controller-only” companies will work out in the long run for the Enterprise. SDN apps-only companies will (assuming the controllers they integrate with do). Compare these to Metreos and IP Celerate in the world of Cisco Voice. But companies that develop a controller (with or without apps) without any switching products – may not succeed. Maybe that model will for the hyper scale web companies – those select few who are on the board of the ONF and have a large in house support staff, but not the Enterprise. Cisco and VMWare dominate the Enterprise market today. I suppose some could debate that too, but I’m pretty sure their market share is easily over 50% each being conservative to make my point. OpenFlow is the most common protocol used for SDN today, but which Cisco or VMWare switches support it? None. But once they do support it, or develop an equivalent protocol, they may be even harder to compete with. You’ll have a horizontal oriented system, i.e. switch+ controller (different companies) vs. a vertical oriented system (single vendor who provides both)? Who will win? In the Enterprise, I really must say I think it’ll be the vertically oriented system. It’s interesting given the hype of “the ability to be horizontal.” I even wrote about the vendor lock-in of SDN here. Take a read if you haven’t already. It’s an interesting topic in and of itself. So, to any end users going down the path of SDN, evaluate the options available – that’s really all you can do. Just remember, you have your own requirements, not the requirements a vendor gives you. If that means sticking with Cisco or another vendor even though they don’t always have the feature first, but they end up always having it down the road, then great. But, if your challenges and requirements are immediate and need to be addressed ASAP, you may need to evaluate new hardware/software companies other than the incumbent.
Above and beyond the networking industry being dominated by Cisco, I’ll also go on the record by saying this is made possible by their large global channel. They have 1000s of companies worldwide going to bat for them, which is a great thing, for Cisco. Their sales force isn’t 12,000 (which it was when I was there a few years ago), but a number far greater than that including their Channel. These Integrators/VARS are often overlooked in customer product evaluations and when making decisions on new products and technologies. So, another point to end users though – realize who will be available besides said new company TAC when issues arise or questions come up. Customers can request new partnerships be made between a new vendor and existing partner, but the net of it is really to understand who will be there to support you should things go haywire at 3am on Saturday evening.
Market education. Something needs to be done here because very few, I mean, very few are aware of OpenFlow and SDN, and that is not a good thing, if it is really going to be the future of networking. Earlier in this post, I stated, “You could be facing similar challenges and not even realize these challenges are challenges, or that there are solutions to these unknown problems. You may think it is just the norm,” referring to end users consuming network technology. I think that could be the case for SDN overall and hopefully that gradually shifts, but most are okay with managing complicated networks because that’s the way it’s always been done. So, if there are other ways, users must know.
Another point on market education. Not saying it’s a good or bad thing, but taking an independent view, I also think Cisco and their channel play a huge part in this. The newer SDN companies and even the smaller traditional network companies just don’t have the reach Cisco has, so whenever Cisco does commit to a new market, it actually helps the industry and their competition too. The smaller niche players know this, and I’m guessing they want Cisco to be more public on their SDN strategy too.
With all of that said, I really am looking forward to attending the Open Networking Summit in Santa Clara this week. It should be a great time and look forward to learning much more and meeting some really smart people.