Latency Special: Tata Communications explains its low latency strategy
06 September 2012 | Richard Irving
Tata Communications explains how its global low latency network is about more than just speed.
There is one question that is pretty much guaranteed to unnerve any battle hardened provider slugging it out in the cut-throat low latency market: What happens when a rival inevitably slices a handful of all-important milliseconds off a key route and your network is no longer the fastest in town?
It can grind a slick sales pitch to a stuttering halt in less time than it takes for a buy or sell order to course its way from one high frequency trader to another. And in a market that many players say has deteriorated into a winner-takes-all arms race – the so-called "race to zero" – it is a scenario that few executives will even contemplate publicly.
Not so John Hoffman, head of Ethernet product management at Tata Communications and the man charged with rolling out a global low latency network for the Mumbai-based carrier. For Hoffman not only stands ready to face down the greatest fear of his competitors, but to positively embrace it. For him, there is no shame in coming second: "It’s not always about being the fastest – sometimes it can be just as important to be diverse. We think there is a very strong business to pursue in being the number two choice for the low latency community."
In other words, Hoffman argues, it is better to become the go-to provider for a diverse route – the "Plan B" alternative should a main link suddenly go down – than to go head-to-head with a slew of rivals on a highly competitive ultra-fast route and fail to land a deal. "If you’re on the fastest cable and you don’t win a customer, you’re out of the game – you can’t pitch again because you can’t offer a diverse alternative. In fact, you can’t offer anything. That can leave the market wide open."
Latency goes global
In a world where a single millisecond advantage in speed is reckoned to be worth upwards of $100 million a year in trading profits, the fastest network between any two financial markets is widely perceived to be the only network and there are no prizes for coming second.
But Tata sees the opportunities differently. In June, the company launched a global low latency network – the first, it claims, to be based on a pure multipoint Ethernet platform. The launch sets the carrier apart from rivals who typically offer low latency connectivity over MPLS or relatively expensive Layer 1 technologies. The network already connects six of the world’s top 10 financial trading centres – London, New York, Chicago, Tokyo, Hong Kong and Singapore and plans are afoot to extend the offering to several emerging market hubs, including possibly Johannesburg and Mumbai.
The relative merits of a multipoint network could well come to the fore as the low latency market becomes increasingly commoditised. Tata’s own research, which it says is based on specific customer implementations, suggests that a firm looking to trade out of London, Tokyo, New York and Singapore, could slice as much as 35% off the cost of installation by adopting a multipoint model. As the chart opposite shows, a point-to-point network connecting four locations would require around 600Mb of capacity with 12 local loops while a multipoint architecture would require just 350Mb with four local loops. The savings also filter through at the operational level, Hoffman reckons, with some customers looking to halve their ongoing monthly opex costs.
Hoffman clearly has no intention of embroiling the carrier in any race to zero. The company shies away from marketing its network on speeds alone, even in specific cases where Hoffman is confident that Tata has an edge. However, Service Level Agreements (SLAs) necessarily include near real-time latency guarantees measured to within two decimal points of a millisecond from PoP to PoP. Data is analysed every five minutes each hour of the day and each day of the week.
An enterprise SLA option, meanwhile, gives guarantees for service uptime, packet delivery ratio, latency and jitter from business location to business location rather than PoP to PoP. The group, he concedes, is making a conscious effort not to sell the network on a millisecond-versus-millisecond comparison with rivals but rather to focus on the platform itself and its global reach.
Specifically, that means pressing home the benefits to customers of working with a single supplier instead of local point-to-point specialists in each of the main trading hubs: "To us, the low latency market is all about offering a package. We have been careful to create our network in such a way that we can shed old routes and bring on new lower latency capacity very quickly." And if the market suddenly changes, Tata has flexibility to react: "We can either jump to the new network, or stay where we are and aim to become the back up provider."
While that may expose Tata to more aggressive pricing pressures, it can be a cost worth paying: "It is certainly true that you might not be able to charge the premium for your route that you can demand if you are unequivocally the fastest." However, there is a lot less fluidity among customers when they are looking for a diverse network offering. "You have a much better chance of winning customers and a much better chance of keeping them for longer."
Picking its battles
By its own admission, Tata weighed its options very carefully before pitching its tent on the low latency battlefield. In fact, the carrier has been considering a move for more than a year, but elected to wait until engineers completed a final link in a 9,280km cable connecting India to Europe via the Middle East, before flicking the switch. In total, Tata has spent more than $800 million building out a wholly-owned cable network ring spanning the entire world. The global network incorporates an advanced submarine cable network and a Tier 1 IP network which links to more than 400 PoPs across 200 countries. It also connects to nearly one million square feet of data centre and co-location space worldwide.
The carrier’s low latency offering piggy backs off some of that cable – but not all. Around 50-75% of the initial build out incorporates third-party network assets, which marks something of a mind shift for Tata. But the fragmented nature of the low latency market particularly lends itself to partnership arrangements: for one thing, the pool of customers is relatively small making it hard for a service provider to justify the expense of building its own network. For another, regional specialists typically boast a competitive advantage in local financial centres, where they have often been a significant player for many years: "Revenue sharing agreements through partnerships tend to help everyone’s cost structures – especially in markets that are relatively small, such as low latency", says Hoffman. The ideal partner is one who shares little customer overlap: "The flashpoint in any partnership comes when both sides try to sell to the same customer. We therefore wouldn’t look to strike up a deal with someone who is trying to roll out a global network – we are looking to expand using very geographically-centred partners."
If ever the low latency market is going to prove its relevance to a wider range of businesses other than just speed-addled high frequency traders, then such cost metrics will be significant. Already, Tata has had strong interest in the network from both an IT developer and a big-name pharmaceuticals giant. Rivals would doubtless give much to know what sort of application a pharma firm is developing that is so obviously mission-critical to the business that it must run real-time across the company’s entire network. Hoffman, not surprisingly, is giving little away. But he is more than a little pleased that Tata has succeeded where so many rivals have failed in reinventing is low latency offering to non-financial firms: "We are definitely seeing the customer base evolve and Tata is certainly well positioned to respond because of the unique way in which our low latency network is configured."
Perhaps the most striking thing about Hoffman’s low latency network is the role it is expected to fill within the carrier itself. Tata has long been an aggressive early mover in Ethernet and currently offers one of the largest multipoint-to-multipoint service networks in the world. Could this initiative say more about Tata’s views on the wider future of Ethernet than the opportunities prevailing among financial services firms?
Clearly low latency is a fundamental requirement to the financial sector these days and if, as a network provider, you are unable to meet the bread and butter requirements of your customer base, then you are going to struggle to be seen as anything other than a fringe provider, says Hoffman. "But I am wondering whether the activity around low latency is indicative of a new move to develop purpose built networks."
It’s getting harder and harder to differentiate your product in the Ethernet space, he explains. But if carriers start to create purpose-built networks, the game might change: "I wonder whether low latency is the start of something bigger. Most of our customers operate on our standard network. But the whole point of rolling out a global low latency offering is to provide an additional layer of service – to send out a message to customers that we have a network to meet their specific needs." It is feasible that someone in the future might develop a tailor-made diverse network whose aim is to specifically avoid all those routes that cross pain points such as Egypt and the fault lines off Asia. For now, Tata’s Ethernet chief is putting the finishing touches to Phase 2 of the low latency network.
As expansion plans evolve, the focus is likely to shift back to Tata’s own assets rather than to those of partners: "Phase 2 will be much more about looking at where our existing cable is and exploring where there might be demand for a low latency service within that", says Hoffman. A focus on emerging market centres is likely, given Tata’s on-going strengths in such regions: "We’ll stick to our emerging markets forte rather than look to centres where all the competition is", he says.