Opening Cross: Liberté, Egalité et Latency: Crackdown on Co-Lo?

As speed has become such a central component to market data distribution and the growth of algorithmic trading, co-location - the ability for trading firms to place their trading systems in or near to exchange datacenters to reduce physical distance between themselves and an exchange's data and matching engine - has become a crucial weapon in the low-latency arms race.

To me, co-location is not some evil enabler of front-running, but simply a replication of old exchange floors - where traders crowded around specialist stations - for a modern, electronic environment. When markets began to be traded electronically - and when "latency-reducing technologies" meant fewer keystrokes required for a human trader to place an order - firms moved traders off exchange floors onto large floors in their buildings, and connected to exchanges via networks. But once computerized algorithmic trading took hold, low latency became an advantage, and as exchanges moved the core of their systems into remote datacenters, firms began placing their servers in the same facilities, adjacent to exchanges' matching engines.

However, as politicians and regulators have turned their attention to high-frequency trading, latency-reducing technologies have been thrust into the spotlight, with the US Securities and Exchange Commission and the Commodity Futures Trading Commission both singling out latency, including co-location, for further inspection.

The CFTC closed the comment period for its proposals around co-location last week, which center around ensuring that all market participants have equal access to co-location facilities provided at "equitable" cost so as not to artificially discriminate against all but those with the largest budgets, and that transparent statistics should be available for latency between all components of the data transmission and trading process.

Responding to the proposals, the Managed Funds Association, which represents many large hedge funds, suggested that markets should list latency statistics by percentile rather than simply by longest, shortest and average latencies, and that "equitable" fees should include a provision that any fee incentives offered by a market or co-lo provider should be offered equally to "all similarly situated customers."

Meanwhile, the Futures Industry Association called for the publication of the number - and standard deviation - of data points used to compile the latency statistics, and also suggested a transparent policy for assigning space in co-lo facilities, starting with equal sized cages for all market participants, then moving to a "first come, first serve" basis once a center is full.

In theory - assuming reasonable pricing and virtually unlimited co-location space - this provides equal access for all to the fastest possible data. But the markets are constantly evolving - as are market data technologies - and it's important to remember that evolution is not a fair and egalitarian process: it's about survival of the fittest, and how mutations in a species (or in this case, a process) can provide an advantage and eventually become the norm.

The challenge of authorities like the CFTC and the SEC will be to maintain the principle of a level playing field without stifling innovation or "handicapping" those firms who can take best advantage of the technologies available. Because otherwise, we end up prescribing things like the exact length of cable between matching engine and co-lo cage. In fact, the FIA is already requesting exactly that: "To avoid discrepancies in latency within an exchange datacenter, as new datacenters are built or existing datacenters upgraded, steps should be taken to ensure the same distance from the rack space to the matching engine throughout the datacenter," says the FIA's response to the CFTC.