Latency Special: Interview with Mike Persico, Anova Technologies
19 March 2013 | Richard Irving
Financial technology entrepreneur Mike Persico is challenging the low latency status quo through his highly specialist network operator Anova Technologies. Richard Irving talks to the chief executive about life in the fast lane.
Who holds the most spectrum licenses in the US? By some measures, it is a former marketing graduate turned serial financial technology entrepreneur with two successful exits to his name and a burning ambition to push the boundaries of electronic share trading to its physical limits. Or should that read the limits of physics? Meet Mike Persico, founder and chief executive of Anova Technologies and holder of more than 4,000 individual Federal Communications Commission (FCC) spectrum registrations. “Clearwire used to hold more but I think we recently took the lead”, he says with a tinge of pride.
Anova is one of a handful of highly specialist network operators that have broken out of stealth mode in the last 18 months to rewrite the textbook on building ultra-low latency networks.
Around the time Spread Networks was lighting up its $300 million state-of-the-art fibre link between Chicago and New York, Persico was starting to wonder whether the future of low latency was not under the ground at all – but rather in the air. He hired an intern who had travelled from Morocco to study at the Illinois Institute of Technology (which counts Nobel Prize winning physicist Jack Steinberger among its alumni), to look at some left-of-field alternatives. “I asked him what the fastest alternative was and he said milliwave but he insisted that it could never work. I told him to go out and design a network anyway and here we are today.” (The intern is now a highly respected engineer at Anova.)
In 2010, Anova stitched together a wireless link between New Jersey-based Direct Edge, America’s third largest stock exchange and Telx’s data centre at 111 8th Avenue in New York, which serves as one of the major gateways to the European financial markets. The route sold out in two days, so Persico rolled out a second wireless link between a new data centre at the Chicago Mercantile Exchange, the world’s largest futures and options market, and what was then the largest data centre in the world, located in a former printing works at 350 East Cermak, in Chicago. “By then we had got a lot better at optimising the signal so we had a lot more capacity. It sold out in 6 days. That’s when I knew the world of electronic trading had changed for good.”
This year, Anova, a privately held company that is backed by one of Wall Street’s top three firms, will generate more revenues from its wireless unit than from its fibre division for the first time. The backer, Persico is quick to point out, does not use Anova’s technology itself: “You tend to find that banks are not big players in the ultra-low latency market – they are a little reluctant to pay the premiums.”
To label Persico a contrarian, is to flatter him. The company motto, self disruption guarantees self preservation runs to the heart of everything that Persico does: “I’m not going to lay the door wide open for someone with second-mover advantage to attack us because we have grown soft and fat”, he vows.
Staying one step ahead of the competition
At every turn, Persico challenges the status quo, questions the consensus. As he talks on the phone, Persico picks up a strand of hollow core fibre, which developers hope will one day marry the advantages of super-fast wireless transmission with the reliability and capacity of fibre. “I have over 1km of this stuff in the office”, he says. “In 20 years, this will be the stuff of which metro networks will be built.
My job is to deploy next-generation technology such as this as soon as it is ready to roll out on a large scale. If I don’t do it, someone else will. You can bet your life Spread is looking at this very closely.”
Spend just a few minutes chatting through the vagaries of the race to zero with him and he will debunk just about every major assumption that Big Telecom holds near and dear on ultra-low latency.
For starters, he does not really see the bigger players as viable competitors: “The likes of Zayo, Level 3, Colt and euNetworks – these are not my competitors – it’s the small guys that were pioneering wireless connectivity when Anova first started up and have spectrum in some of the really busy hot spots that I worry about.”
That will sting over at Zayo, which is scaling up its low latency business after buying AboveNet last year, as well as at Colt, which would see itself as one of the leading service providers to the financial community. Others, like euNetworks would probably be quite happy to edge sideways away from wireless technologies anyway, to leverage their investment in fibre and target larger enterprise customers who might not be quite so latency-sensitive as to chase after every millisecond gain. Certainly, the barriers to entry in to the ultra-low latency world are rising ever higher.
US regulations require that a millimetre wavelength needs to have a “cylinder” of 17 metres of free air around it for every 1km that the signal travels. On a key route such as that between Direct Edge and BATS, the fourth largest stock exchange in the US – a distance of just 6km - the signal needs a 100m cushion.
The problem is most acute at the access points to key trading locations, where spectrum is in very short supply.
Persico was quick off the mark to spot the looming constraints: “Back in 2010, we went out and registered all the shortest routes between all the key trading locations in New York and New Jersey with the FCC.” It was a considerable undertaking, he says, but well worth the trouble. “There’s very little spectrum available now.”
Persico is sceptical that bigger operators can now shake free enough spectrum to muscle in to the ultra-low latency market. But he saves more scorn, for the slavish way in which Big Telecom unquestioningly pursues reliability.
It is fundamentally wrong, he argues, to assume that electronic trading firms need to place so-called ‘five-nine’s’ reliability on networks at the expense of all else. “Here’s the thing”’ he says. “If I can be 100% faster, 80% of the time, then electronic traders will buy my service.” The point, he adds, is that most computer-driven trading programmes only work if they get to the trade first. If they are not there first, then they are out of the market and if they are out of the market, then there is no money at risk in the first place.
“Speed is everything: resilience, reliability, redundancy – all these things come a very poor second,” says Persico. And if speed trumps all, then the first casualty is bandwidth capacity. “There has been a subtle yet fundamental shift in the way electronic trading firms consume bandwidth over radio frequencies”, he says.
The sophisticated firms have gone right back to the drawing board to see what they actually need out of a network – how much data a system is required to carry and what sort of availability it needs to offer. “And that’s exactly why I don’t view Level 3 and Colt as competitors – that shift has been totally lost on them.”
Right now, he says, the bigger players simply fail to grasp that the focus is not on how much data you can reliably carry but how fast you can deliver it.
Which is not to say that Anova is not throwing R&D dollars at ways to expand millimeter wave capacity. “The real Holy Grail is to get circuits to 10 Gigabytes per second (G) over the air”, says Persico. That’s around ten times greater than what is currently possible and would either allow traders to beef up their data consumption again, or clear the way for Anova to sell more circuits on the same path. “The big challenge with 10G, is spectrum”, concedes Persico. There’s just not enough to go round at the aggregation points – especially on the rooftops of the big exchanges.”
Another key driver in the ultra-low latency market that Big Telecom struggles to grasp, Persico says, is the concept of time to market. In essence, ultra fast links undergo something of a rapid evolution over a very short time scale: The networks start off being a huge value-added for which a handful of early adopters will readily pay a big premium.
But as the technology becomes more widely dispersed, those customers that need to be competitive on speed end up having to buy the connectivity whether they like it or not. And at this point, the latecomers begrudgingly view the upgrade as a tax on their trading profits.
The time to market is essentially the lag between the point at which a new technology first gets rolled out, and the point at which it becomes a must-have commodity. It therefore follows that the quicker a provider can deploy the very latest technology, the greater the lead time that it can offer customers and the bigger the premium that it will be able to charge.
“If you can give an electronic trading firm a head start on a new link of anything from 6 months to a year then you’ve got a good route on your hands.” Big Telecom, Persico maintains, rarely finds itself in that position.
Persico reckons to count about 100 specialist electronic trading firms as customers - he eschews the term “high frequency traders”, noting that it now carries the faint whiff of scandal. The strategy, where firms use computers to fire off thousands of buy and sell orders when they spot tiny aberrations in the shares of related companies, has attracted mounting regulatory interest amid concerns that it played a part in the flash crash of 2010.
“We do not live or die by the actions of a few high frequency traders. The ultra low latency market is about so much more than that. We are a cottage industry built on the foundations of electronic trading and electronic trading is not going to go away – we are never going back to a world where you had guys in brightly coloured jackets wildly waving their arms about and shouting at one another.”
Fast forwarding to the future
So what does the future hold? Persico, in typical fashion, turns the question on its head. “The real question is what advances customers are willing to pay for. If you’re asking me whether I think wireless will always be terrestrially bound, then my answer is ‘no way’. But who’s going to finance the next paradigm shift in how we use it?”
Take for example, the crucial transatlantic corridor linking London and New York. Buy and sell orders currently take a shade under 60 milliseconds (mS) to ping their way there and back on the route: “I’m pretty sure that someone out there would be willing to pay handsomely if you could halve that to around 30mS through the use of wireless technology.”
But the problem, he explains, is that if someone did indeed stump up the financing for such a project, they would probably demand exclusive rights to it: “If you end up building out a route for just one customer, the return on investment is always going to be pretty low”, he says.