You are here: DIME Home > Technology > Market Data Latency

Market Data Latency

Just a few years ago, market data latency was measured in milliseconds. Since that time the major world equity and derivatives exchanges have made significant investments in building latest data centers offering collocation services to firms looking to take advantage of low latency and as such, the standard measure of low latency is now microsecond.

Author: james davis
Date: May 18, 2012 - 3:46:16 AM

Printer friendly page

Just a few years ago, market data latency was measured in milliseconds.  Since that time the major world equity and derivatives exchanges have made significant investments in building latest data centers offering collocation services to firms looking to take advantage of low latency and as such, the standard measure of low latency is now microsecond.  As technology continues to improve, firms are now looking to measure latency in nanoseconds. Do firms really need ultra-low latency when it comes to market data reception?  It depends.  For firms who seek to profit from high frequency trading – a vague term to certain; how frequent is frequent? – Ultra low latency is an absolute must.  Such firms must ready to make the proper investment required to:

· Develop software capable of executing a high-frequency trading strategy

· Purchase hardware and underlying software capable of receiving this data, including operating systems, middleware and feed handlers

· Acquire low-latency market data feeds and collocate in near to the exchanges to receive market data directly, and as close to the source as possible

Whether executing index arbitrage, traditional arbitrage across regional exchanges, or complex quantitative algorithmic trading strategies, well-designed software applications are required to act on strategies that require the submission of market orders within microsecond.  Generally speaking, such applications will be written in C++ due to its inherent efficiencies. Although less common, some HFT applications have been written in Java.

HFT applications must run on high performance hardware and underlying software capable of executing the desired strategy.  These applications should be written for a Unix operating system such as Sun ecosystem's Solaris, CentOS (a Linux variant based on Red Hat), or enterprise Linux distributions such as Red Hat or SuSE.   "Out of the box" installations are insufficiently configured and must be finely tuned to take full advantage of HFT software applications.  Additionally, low latency feed handler and middleware must be capable of managing ever-growing market data rates. Vendors such as NYSE Technologies, Exegy, SR Labs and others have all introduced technology to meet these requirements. One of the most significant and important trends we see in this technology space is the move towards accessing data not through the network, but via direct connection, and communication between hardware and software via direct memory access.  Direct memory connections greatly improve the speed and efficiency of data transfer and significantly cut the strain placed on system hardware such as CPUs and network interfaces.

As equally important as choosing the right hardware and software is the selection of proper market data feeds.  HFT strategies will, in the majority of cases, must direct data feeds from each exchange as opposed to consolidated feeds such as those provided by Bloomberg, Thomson Reuters and others.  Recently, NYSE Technologies announced the release of their new “Arca Integrated Feed” aimed specifically at those firms that need to consume vast amounts of data from the NYSE Arca market while maintaining ultra-low latency.  Other major exchanges are likely to follow suit in the coming months.  It is important to note that even if a firm has need for level 1 quote and trade data, direct feeds for personal exchanges will give much lower latency than any other type of feed, including the consolidated tape as provided by Securities Industry Processors CTA (operated by NYSE Euronext covering tapes A and B) and UTP (operated by NASDAQ covering tape C).

Finally, ultra-low latency HFT strategies must the placement of hardware and software in near to where each exchange both publishes data and accepts order flow.  Even the most finely tuned and well-written hardware and software are subject to the laws of physics.  This type of colocation is often called Proximity Hosting and allows firms to find their trading applications, quite literally, at the source. In the coming years HFT will continue to play a major role in capital market dynamics and firms must be aware of what is required to effectively compete in this space.

Gravitas Technology providing the Gravitas Technology,Hedge Fund Cloud,Hedge fund colocation

View all articles by james davis

Link to this article:

Follow & Share Your Favorite Video Articles

Tell a Friend About This Site

Tweet about this site

Follow @dimeco

Recommended Associates

Do you write?

Would you like submit your articles and have them approved on a priority status? Find out more about how you can become a Priority author for pennies a day! Click here.