Pomegra Wiki

Colocation

The colocation of trading servers at a stock exchange’s data center is a strategy used by high-frequency traders and institutions to minimize network latency. By locating their computers within the exchange’s facility, traders receive market data and can execute orders microseconds faster than traders connecting remotely. Colocation is expensive and creates a tiered system where wealthier firms have a speed advantage.

This entry is about infrastructure for trading speed. For latency more broadly, see latency tier; for the data accessed, see direct market data feed.

How colocation works

A trader arranges with an exchange to place a server (housing their trading algorithms and order-routing software) in the exchange’s data center, meters or feet away from the matching engine.

The colocated server connects to:

  • Market data feed: Direct connection to the exchange’s data publication system.
  • Order entry system: Direct connection to the exchange’s order acceptance.
  • Power and cooling: Provided by the data center.

Result: The trader’s algorithm receives market data and places orders with latency measured in microseconds, rather than milliseconds.

Latency advantage math

Remote trader:

  • Network latency to exchange: 5 milliseconds (data).
  • Server processing: 1 millisecond.
  • Network latency back: 5 milliseconds (order).
  • Total: ~11 milliseconds.

Colocated trader:

  • Server latency (within data center): 10 microseconds.
  • Server processing: 1 millisecond.
  • Exchange processing: 1 microsecond.
  • Total: ~1.01 milliseconds.

The colocated trader is ~10 milliseconds faster. In high-frequency trading, this is an eternity; thousandfold more orders can be placed and executed.

Costs

Co-location rack fees: $5,000–$20,000+ per month for server space, power, and cooling.

Network connections: $1,000–$5,000 per month for dedicated connections.

Systems and staff: Maintaining a colocated trading system requires dedicated engineers: $100,000–$500,000+ in annual salaries.

Total annual cost: $50,000–$200,000+ per exchange, easily $500,000+ for firms co-locating at multiple major exchanges.

This cost barrier means only well-capitalized firms can collocate.

Who collocates

High-frequency traders. The primary users. Millisecond advantages can generate millions in profits; the colocation cost is easily justified.

Large asset managers. Firms with substantial assets under management may collocate for algorithmic trading.

Broker-dealers. Investment banks often co-locate to provide the best execution.

Market makers. Designated market makers at exchanges typically co-locate to maintain fair quotes.

Retail investors never co-locate; the infrastructure is far too expensive.

Exchange policies

Exchanges offer co-location services and profit from the fees. However, exchanges have adopted policies to prevent unfair advantages:

  • Equal access. All qualified market participants can access co-location (though at a cost).
  • Fairness rules. Orders must be processed in a fair, orderly manner, regardless of co-location.
  • Dispute resolution. Exchanges have processes to review complaints of unfair advantages from co-location.

Despite policies, co-location inherently creates information asymmetries: colocated traders see prices microseconds before remote traders.

Controversies

Unfair advantage. Colocated traders have an unfair edge over retail investors with remote access.

Flash crash contribution. During the 2010 flash crash, colocated high-frequency traders’ rapid response to cascading prices helped accelerate the decline.

Wealth transfer. Some argue that co-location-enabled high-frequency trading is wealth transfer from retail investors to well-capitalized HFTs.

Systemic risk. The concentration of colocated systems at exchanges creates operational risk; if a data center fails, many traders are affected.

Global colocation

Major exchanges globally offer co-location:

  • US: NYSE, NASDAQ, CBOE, and others.
  • EU: Euronext, LSE, Deutsche Börse.
  • Asia: Tokyo Stock Exchange, Shanghai Stock Exchange, others.

This globalization of co-location has enabled high-frequency trading worldwide.

Regulators and proposals

Some propose limiting or eliminating co-location advantages:

  • Speed bumps. Introduce artificial delays so all traders experience uniform latency.
  • Auction systems. Instead of continuous order matching, use periodic auctions where speed confers less advantage.
  • Mandatory slow-down. Slow down all systems to eliminate microsecond advantages.

However, no major regulatory changes have been implemented. Exchanges resist, as co-location is profitable for them.

See also

Wider context