I keep hearing people telling me: “I want to go SDN and orchestrate my network to get rid of this horrible vendor lock-in, so that we can buy any boxes we want in the future, no matter the supplier. To do that we bought Contrail.”
To me that’s vendor lock-in. Don’t get me wrong: Contrail and especially OpenContrail are excellent products, especially the latter as it’s open.
However this means you have to stick with it for the foreseeable future, unless you want to spend other time and capex to move to another platform, readapt all your northbound interfaces, etc.
If you think about it, the fact alone that most of the big players harvested the market for startups who made orchestrators means a lot. After having realised that the hardware lock-in was going to be less strong, they all rushed to acquire companies of the likes of Nuage (acquired by Nokia), Tail-f (now Cisco NSO) or Contrail (curiously enough, the amount Cisco and Juniper spent to acquire, respectively, Tail-f and Contrail is identical: $175M for Tail-f and $174M for Contrail).
This industry is at a crossroad these days, as it needs well established common interface standards, adopted by all the relevant vendors. The slow pace at which standardised YANG models have been promoted by IETF is yet another symptom of this old-fashioned behaviour that most of the big guys are having, despite all their claims about being ‘open’, etc.
We have really very interesting times ahead, I believe. YANG is going to be a big thing, in that modeling end-to-end services rather than micro-modeling the underlying protocols will be key to speed the delivery times of network services in the future.
Another big thing is that bigger companies will be able to protect themselves slightly better from this lock-in effect rather than smaller ones, as if you want to embark on the journey of adopting a fully-open solution (i.e. openstack+opendaylight, etc.) you’ll need some really skilled architects and engineers to survive and succeed.
Smaller companies will be forced to rely on big colossuses like Redhat/IBM/HPE providing them turnkey solutions, which they’ll have to stick with for a long while, hence the lock-in effect.
I think the network world is re-living what software developers lived in the last 5 years. For them, now, compute, storage, memory and network are a commodity, and they don’t need to mingle with IP addresses and static routes anymore.
The direction we’re heading towards is for network services to consider the underlying network a commodity. No need to configure VRFs, ACLs, HSRP, VLANs, BGP, etc. It’s all going to come for free.
Only problem is that no one seems to be able to really tell us how we’re going to get there…and when.
P.S. If you read this and have really little (or no) idea about what SDN is, I really encourage you to buy this: “Network Functions Virtualization (NFV) with a Touch of SDN”, it’s a brilliant book!