SDN Thoughts and Interesting Interactions

The other day there was a comment from Derick Winkworth (DW) on the Plexxi blog.  It is worth reading in total so I providing the text in addition to the link.  The parts in bold are my additions.

————————————————————

“I was looking at the SEBoK wiki the other day and something occurred to me. From an application perspective, its all infrastructure. Network, servers, storage, and appliances. And these things are part of the delivery path for the application. They are all part of the application. Looking at this way, the infrastructure is a system of systems. Today routers and switches operate as true autonomous entities. So this makes them individual systems part of a larger overall system, “the network,” and yet again from the application perspective, “infrastructure.”

In their current state as autonomous entities all of their state and data was maintained locally and difficult to access and keep abreast of. Additionally that intelligence limited our ability to control and manage the nodes themselves. They weren’t meant to do what we need them to do now. They were meant to be independent operators forming a network for packets to pass. They weren’t built for applications that are really distributed across multiple virtual-elements through the infrastructure.

At a technical level, we have reduced the overall number of intelligent agents in the network by combining them into a single entity with central intelligence. We don’t need, in the data center for instance, many autonomous entities. What we need our data center to do can only be done by centralizing the intelligence and using modern software to control the nodes at the forwarding level.

We are still, overall, a system of systems, we have just taken a portion of the overall system, some portion network nodes, and reduced the overall number of sub-systems (or intelligent agents). What would have required enormous effort to engineer across many intelligent systems (and expense since they would require larger CPUs, more RAM, and more complicated software on all those nodes), is now possible with simple nodes awaiting instruction from a single intelligent agent. The problem (delivering distributed applications) is easier to solve with a single agent.

The interesting thing is that we didn’t know what we needed the network to do and now that we are starting to wrap our heads around it, we are seeing all these DC fabrics and other solutions coming out of the woodwork. If you consider the infrastructure as an extension of the application, then the application is really spread out across the various subsystems (appliances, servers, storage, etc). Now that all this can be virtualized, SDN allows us to stitch the application “components” together across the physical infrastructure in a way that simply passing packets/forwarding frames could not.”

————————————————————

Last week I summarized a lot of my SDN thoughts in a single post, but in fairness I think DW did a far better summary than I.  In case you missed the presentation, Cisco has been out talking about their SDN strategy.  Here is a link to the presentation.   When I think about how change occurs, it occurs in the social fabric of a community.  I am using the term community to refer to all groups: industry groups, social clubs, nation-states, corporations, etc.  I once wrote a 17k word missive on this subject.  My point is that change starts with the allegiance of the intellectuals – it does not start with the status quo.  Change starts when leading influencers start to stitch together relationships, in the same way we want to stitch application components across the physical infrastructure.  That was point of this post.  When I read DW’s comments on the Plexxi blog, I want to ask: what are you seeking?  I see this every week in my job.  I sit across of people who are so full of curiosity about the network.  I want to reach across the table and tell them to admit it, ask the questions.  If you are blogging, writing, thinking, questioning, seeking, then I would say you are leading.  The questions leaders are asking can be grouped into these broad categories:  Where Do We Come From? What Are We? Where Are We Going?

Where Do We Come From?

From an application perspective, its all infrastructure. Network, servers, storage, and appliances. And these things are part of the delivery path for the application. They are all part of the application. Looking at this way, the infrastructure is a system of systems. Today routers and switches operate as true autonomous entities. So this makes them individual systems part of a larger overall system, “the network,” and yet again from the application perspective, “infrastructure.”

What Are We?

We are still, overall, a system of systems, we have just taken a portion of the overall system, some portion network nodes, and reduced the overall number of sub-systems (or intelligent agents).

Where Are We Going?

Now that all this can be virtualized, SDN allows us to stitch the application “components” together across the physical infrastructure in a way that simply passing packets/forwarding frames could not .

Email has been around as long as networking.  In a conversation with colleague, it occurred to me that email might be a leading indicator for networking from a behavioral (i.e. use) and deployment (i.e architecture) perspective.  Allow me to indulge without all the CCIEs rolling their eyes.  The first email system I used and administrated was cc:Mail.  I just dated myself with that fact as we were in the Reagan to Bush era.  Back in those days, I remember traveling in Europe and how we used email.  I would land in LHR, CDG, MUC or AMS, get to the hotel and dial into the mail server in the US.  The modems would connect at ~1200bps and start syncing.  With the process started, it was dinner time.  By the time I got back to the hotel room there would be an inbox full of messages to read.

Along came the internet and email gateways, which led to a lot more interaction on email.  Email went from being mainly an internal system to a B2B and eventually a B2C system.  Email really was the death of systems like Lotus Notes.  Yes, all the LN fans will send me hate mail, but if you ever used LN and then used email you know what I mean.  It was just easier to email documents, presentations and files to your work colleagues.  That was the moment that email went from messages to a file transfer system and eventually a file storage system.  It was easier to use email and people tend to favor what is easiest to do.  When email morphed into a corporate file transfer platform, that was when IT managers started managing your inbox size.  Do you remember these quaint days when you would get messages warning that your inbox or email folders where too large?  I once worked in a company where the amount of storage allocated to your email account was calibrated to your seniority in the firm.  Back in these days when you spoke to the CIO or the IT managers in a firm there would always be a discussion about the email usage, the cost of keeping up with email storage, blah, blah.  Again, it all seems so quaint.

Change happens much faster than people are willing to acknowledge.  In 2005, I was presenting at the Telesoft Partners VC conference.  At that conference I watched Paul Chamberlain of Morgan-Stanley give a presentation titled “Google and Beyond.”  One slide stood out to me in his presentation and that was slide 12.  The archive of presentations can be found here.  The commentary around slide 12 was that Paul was on Nantucket or MV (I do not remember which one) when he was on the team taking Netscape public in August 1995.  He was at his mother’s house or something like that and the only phone he could use for the IPO conference calls was a black rotary dial phone.  If you look at the slide you now understand the background.  Six months after Netscape went public, Mary Meeker and Chris DePuy in their famed Internet Report, declared that “at a minimum, e-mail should become pervasive.  So should Internet/Web access: Email is the “killer application of the Internet today, and browsing through information services the “killer app” of tomorrow,” see, the Internet Report, Mary Meeker and Chris DuPuy, February 1996, page 1-2]. 

When I think of email today, I think how pervasive it has become.  I have four (4) email addresses that I use on a regular basis (@gmail, @koss1, @plexxi, @me) and few others I have not checked in years or do not use (@yahoo, @verizon.net, @att.net).  I never think about inbox size or storage anymore.  Email just works, it just happens and there is nothing to manage.  I access all my regular email accounts from my iPhone, iPad and MacBook.  All my email addresses and associated content are stitched across the infrastructure of the network (hat tip to DW).

When I think of the network and how SDN will change how networks will be deployed, I think about applications.  All the angst you hear from network engineers will fade away, just like inbox size limitation has gone away.  The network will become dynamic, rather than fixed.  That is what happened to email.  Email went from the world of rigid rules, autonomous systems that were fixed by management to just working.  The applications and the network will go the same way, it will just work.  Over the past forty years people designed a lot of networks to provide connectivity to all places.  What will be doing in the SDN world is designing how applications will flow across the network.  A lot of people cannot see it today, but that is because they are standing in the forest with their nose about an inch from a tree trying to figure out how to provide network for every type of tree in the forest.

/wrk