The technology that gives me a “nerd hard-on” this month is SDN WAN. Here is why.
SDN is Not a Revolution
Software Defined Networking in the Data Centre is so last year. The reality I am seeing is that a larger part of the Enterprise doesn’t want SDN that replaces the data centre network because of the risk and cost associated with that. Cloud Providers are a different market but they don’t need or want physical networking either. Cisco ACI has just a couple of hundred customers after 12 months of marketing and pushing (and thats a lot of marketing dollars, sales people, and not much results).
- You can start today without replacing anything.
- You can start small and incrementally grow over time.
- It only impacts a small part of the data centre so working with Operations & Security on changes is easier.
- Making small changes doesn’t need the CIO or Executives to find big budget chunk to get it done.
It is rare for the Enterprise to spend big chunks of cash on new stuff. The day-to-day is about small, incremental progress. Thats the SDN that is winning in the data centre.
2015 is all about SDN WAN
But hey, SDN Data Centre is so 2014. This year will be about SDN in the WAN where the budget justification is going to be much easier. The amount of money spent on service providers for bandwidth is just an enormously huge chunk of IT budget and even small amount can result in huge savings. WAN Optimisation technology lives by this reality.
Using overlay networking in the WAN gives you the same benefits as overlay networks in the data centre. There are five key things that I think are important:
- Increase utilisation by directing traffic flows over unused links. Think about those backup WAN links that do nothing for 350 days of the year. Routing protocols like BGP & OSPF can only calculate a single path through the network, at best you can perform equal cost routing. At worst you can use hop-by-hop routing policy to override the default condition and then try to keep it operational. What about those regional links that get used at about 10% and you keep them clear.
- Gain independence from carrier / provider. Overlays means that you can use many different providers in different locations and still build a single network. You don’t have to use MPLS or fancy routing to weld a single IP routing space together. Use links from five or ten different providers and then create a single overlay network across all of them.
- End to End Visibility including QoS & Analytics. Lets face it, SNMP is crap. Syslog is crap. NetFlow is pointless except at Core WAN sites and requires enormous amounts of processing to be useful. The data sources for network monitoring are just awful and we get low quality data from devices, at best. SDN is about defining new data models as APIs that can be dynamically programmed.
And ‘analytics’ is about using big data software to analyse the data for patterns and use pretty graphs for something more useful than sales charts by region.
This technology is still emerging but expect to see this in 2016 not 2015.
- Incremental deployment. SDN WAN doesn’t have to replace your existing routers. You can deploy SDN WAN devices like WAN optimisation boxes in the places that matter. You can start small and grow over time, prove the solution, modify the operations and security plans and build over time.
- Speed of Deployment / Technology Independence. I read somewhere that 75% of corporate WANs have been outsourced because there is just so much pain associated with managing large numbers of routers littered around the place. But what using 4G LTE to connect a new office while you wait for the tail circuit to be installed ? Why not use a WiFI provider ? Why not use them instead of a Leased Line ?
The heart of SDN is about having programmable network devices that can be modified easily and reliably. I don’t know where SDN WAN will end up but the transition to SDN will be through a migration using overlays so that you can keep your existing 5 to 10 year WAN outsourcing contracts until they expire and add a few sites at the edge to begin a continuous migration over a couple of years.
No one gets rewarded for recommending radical change. It took a decade for WANs to connect every branch, for VMware ESX to become widely used, for Windows to lose its attraction ….
It will take a decade to migrate to SDN in the WAN but I am sure that you will be able to see how SDN WAN works in 2015.
Human Infrastructure Newsletter
This article is taken from the first issue of my newsletter. You can subscribe to the newsletter here and get my personal take on being a human in the world of IT Infrastructure.