Telstra last week made an announcement regarding software-defined networking, network-function virtualisation and an expansion to its Pacnet-enabled network (PEN) service.
Now, I like Telstra and they do some things very well, so don’t misunderstand me, but unfortunately this is mostly marketing spin, buzzwords and another Telstra product that won’t go anywhere.
Firstly, SDN doesn’t mean anything. SDN is a vague term for a collection of technologies with no standards, little adoption, no actual products that are marketable and no one who really wants it. SDN is a solution looking for a problem with the biggest issue being that there are very few network engineers who know it – and even fewer who care.
I am a network architect, engineer, programmer, Linux engineer and I couldn’t care less. SDN is to networking what Windows 8 was to… well, no one. We don’t want it, we don’t need it, and no one actually asked for it.
Telstra has been in the game of elastic fabrics (from an engineering perspective, not billing) since it purchased PEN for $850 million in December 2014. But as I've stated during a number of talks at industry events, carriers cannot build a successful elastic fabric that is used by anyone other than their own loyal customers in their closed eco-system.
The key issue is that the biggest value in a fabric is the marketplace. Telstra will never build something where it allows open connectivity to other carriers freely. This is the same issue with Equinix's Cloud Exchange. There is little on there to attract many people because it is only available in Equinix data centres and people on the Equinix ecosystem.
Megaport is so hot right now globally because it is an underlying ecosystem that does not dictate who is on the platform and what they can do. They provide little value-add other than facilitating rapid and cheap connectivity. That is why it so magical. It is in every popular data centre and anyone can join freely and offer anything to anyone, whether you are a carrier, ISP, value-add service provider, vendor (hardware and software), cloud player or indeed just end users connecting to each other.
The other defining feature of an elastic fabric is its low-cost and minimal-to-zero contract terms. Just like elastic compute from AWS, Azure, Google Compute, OrionVM and others. You pay for what you need, when you need it, and that is all.
How does Telstra's PEN product compare to that? It is expensive, in very limited data centres, with few people to connect to other than a couple of cloud providers. Sure, it has some automation (SDN, anyone?). But any organisation with a few programmers, a decent understanding of the network and its APIs as well as some good UX people can make a nice and scalable orchestration platform. The only thing PEN has going for it is the international connectivity, but it is extremely expensive and other players will address this soon (Megaport does Australia to New Zealand for $10 per Mb).
Telstra's move is also too little too late. AXONVX does 40Gb interfaces and it was designed to rapidly scale to do 100GbE when needed. Megaport has offered 40GbE and 100GbE since it started (but alas has not sold much of it due to the cost of the switching equipment – though that is also about to change with the Open Networking revolution).
If an elastic fabric is in a locked eco-system, it will not gain mass adoption. Not Telstra, not Equinix, not Vocus nor Optus if they try. NextDC will also have growth issues with AXONVX if it doesn’t expand beyond its own data centres and continues to treat it like as a 'NextDC product' as opposed to its own thing.
The best part of all of this for resellers, MSPs and systems integrators is that all elastic fabrics are essentially the same from their perspective. It is just old school VLANs being rapidly deployed from A to B (and sometimes C, D and E). If you know current networking, you are probably fine and you can join as many fabrics as you like (and can afford) and even link them together.
The analogy I use for Elastic Fabrics is that of a Westfield shopping centre. We all go there because all the shops are in one place with nearly everything we could need. A successful elastic fabric is exactly the same with the number one, most important factor being the community of providers (shops) that are there. The more providers, the more reason for businesses (and later consumers) to get on board. The next key reason is that once you are connected, bringing up new VLANs is almost instant. No waiting four to eight weeks for a carrier to deploy something for you. To bring up a Megaport VXC across Sydney is no more than about 60 seconds and you are on your way.
All that is great, but the key part of all of this is cost and terms. Megaport charges $100 per month or $200 per month (depending if the data centres are close) and even do $10 per day or $20 per day for short-term connectivity. The price is the same if you configure it as 2Mb or 5Gb… they don’t care. Prices obviously differ interstate, but are still attractive.
There is nothing engineeringly amazing in Megaport except their productisation and the drive to be a disruptor. Telstra and others could of course do this, but they won’t. They are too big, too monolithic and the elastic fabric concept eats straight into the core of very big revenue sources for them (for now).
Look at the pricing Telstra and VMware vCloud Air versus AWS, Azure and others and you see why there is minimal uptake.
Skeeve Stevens is the founder of eintellego Networks. Disclaimer: Skeeve is the architect of NextDC’s AXONVX platform and also a long-time friend and supporter of Megaport.