Opening Cross: Latency, Line-of-Sight, and the Legacy of Sandy

More than six months after most aspects of life on the US East Coast returned to normal following Hurricane Sandy, two recent stories in the mainstream press remind us of that tragedy: first, that many victims left homeless by Sandy’s devastation are now suffering from post-traumatic stress, and second, that many New Yorkers will suffer a different type of stress when parts of New York City’s R and G trains undergo lengthy closures to repair lingering damage caused by the storm.

In Sandy’s aftermath, US financial markets closed briefly, largely due to transportation and power issues. Yet the immense, secure datacenters where the markets house their infrastructure also fell victim to the storm, despite being designed to withstand—and continue operating through—all kinds of disasters.

“We don’t often think about losing datacenters… and Sandy did take out a number of datacenters through flooding,” said Mark Casey, president and chief executive of network provider CFN Services, on a panel discussion organized by datacenter provider Telx last week. And though “networks were pretty resilient,” during Sandy, according to Casey, others warn of Sandy’s impact beyond physical damage and repair costs, in terms of how support contracts might change to guard against commercial impacts.

“From a fiber-optic side, when you have to fix a network, the normal service-level agreement goes out the window, because you’re not the problem—the power company might be the problem,” which would be beyond a provider’s control, said Michael Sevret, chief strategy officer at network provider Cross River Fiber. “So clients are now paying a lot of attention to the ‘force majeure’ part of a contract… and as an operator, you need to pay attention to what you are agreeing to.”

Also, while datacenters typically keep reserves of fuel and supplies on hand to continue operating uninterrupted through disasters, network operators should also hold extra inventory in case they need to repair networks. “When you’re trying to fix a cable, so is everyone else, and so it can be hard to get supplies,” Sevret added.

And though radio frequencies (RF) may not be subject to the same disruption as damaged cables, antennae can easily be jostled loose—just a few degrees of movement can send a signal off-course—while even normal weather can impact on microwave data’s effectiveness.

Yet trading firms and connectivity providers such as CFN continue to build out new microwave networks to connect users and marketplaces in key datacenter clusters like New Jersey, and on prime trading routes, such as between New York and Chicago, and between London and Frankfurt because of the potential latency savings over fiber networks—estimated at between 30 and 50 percent.

“We launched our microwave network last October between Basildon and Frankfurt, and we sold out the first tranche of bandwidth in... 60 days,” said Andrew Kusminsky, chief operating officer of network provider Perseus Telecom.

CFN and Perseus aren’t the only ones excited about microwave connectivity. Datacenter provider Equinix is currently making space available for antennae on the rooftops of its hosting facilities—or adjacent to its buildings—to meet demand from high-frequency traders wanting to connect to their datacenters via RF.

But just as microwave’s limited bandwidth means firms must be more selective about their data consumption, there are also a more select number of firms that can invest in these latency-reducing options, since many are still under pressure to control data-related costs. One such firm is Wells Fargo, whose MDCOP self-service usage assessment tool has proved so successful in reducing costs for data consumers that the bank is now looking at other cost centers and expense types that could benefit from using it.

If you’re sensitive to cost, I advise looking at what Wells Fargo—under IMD Market Data Executive of the Year award-winner Steve Listhaus—has achieved. And if you’re sensitive to latency, I strongly advise avoiding the G train.