It is interesting that Scott has admitted to having misgivings on SDN, and that he is trying to direct a ‘sequel’. Or should I call it - a reboot! As I read through the article, my reaction was disbelief and it got me pondering on the current state of networking. So I decided to spill it out - usually do not do this, but there is always a first! So, I am setting out on this contentious OpenFlow topic; for sure there will be a lot of dust raised….
So, how easy is it to realize the potential of OpenFlow?
The potential of OpenFlow is vast - there can be many use cases and applications that can be built around OpenFlow if you are able to invest work and creativity into it. Also in ‘theory’, OpenFlow can support ZTP (Zero Touch Provisioning) of a data center fabric.
Well, the entire premise of SDN and OpenFlow was of agility. If you are in business, agility is key. You need SDN solutions that you can roll out and leverage to your competitive advantage. Getting creative with OpenFlow use cases is probably not high on your list of business objectives. As far as the theory part goes - real world networks are just what they are - business enablers, not lab research projects.
Okay, so, somewhere around here is when I need to reach out for my fire resistant suit! But, let me go on with more on OpenFlow while I put that suit on…
OpenFlow primarily targeted Layer2
So why is this a problem? The assumption here is that all data flows through routers and switches. Which means OpenFlow can manipulate data streams at a Layer2 level. Now, you may hear some other networking vendors say this is not entirely correct and that it is possible to manipulate traffic at least up to Layer4. But isn’t there a lot more to the network? Does that not leave load balancers, traffic optimizers, ADCs and so on out of the OpenFlow equation?
Translating between network and business speak it would simply mean that the performance of your infrastructure would remain sub-optimal when it comes to supporting your critical application flows and business processes with all these post traffic high level manipulators and bandages.
“Nobody covers MPLS” - Scott Shenker
Are you kidding me? Real world networks run MPLS, BGP, RSVP, STP and a lot more on them. I can see your eyes roll - yeah, yeah.
You are saying this because Juniper is a MPLS company; OpenFlow is about dumbing down all these protocols on the network. So, we have already seen that OpenFlow is quite limiting. Now, add to that Scott’s admission that it was “… unforgivable. We just ignored current systems,” The conclusion there is quite obvious!
And, yes, Juniper’s roots are strong in the service provider MPLS area. MPLS is a mature, proven, reliable and standards based transport mechanism. The fact is that MPLS is not isolated just to carrier networks, many enterprise networks also have MPLS. There’s a reason why there are so many MPLS networks deployed today.
There are some vendors who placed their bets on OpenFlow. But there is still a lot of ‘creativity’ required around OpenFlow - work on getting more from merchant silicon and lot of programing to manipulate the bits around. But such creativity also needs more money and resources. Wait! Wasn’t OpenFlow supposed to change all that?
Part2 of this blog - there is a lot more to SDN than just OpenFlow.