Every Microsecond Counts Today brings modern technology and with it, an increasing number of Internet-based applications that require an end-to-end latencies, requiring the need latencies to be transmitted at a rate of milliseconds or even microseconds. Many applications though still demand that the latency remain stable, otherwise having little to none ‘jitter’.  These new applications range from the popular multimedia services to things like voice-over-IP, multi-player gaming and even video conferencing. Since these applications are growing at such a successful rate, customers are placing just as increasing demands on operators to help provide and manage the networks needed to meet these new stringent demands. Unfortunately many of the tools that are currently available for use are not able to create accurate measures of the latencies in the magnitudes that are being demanded, nor can they even detect or localize any loss of spikes. This leads to a loss at extremely small time scales like tens of a microsecond.

Here is an example, a trading network that connects a stock exchange to the number of data centers where automatic trading applications run. In order to stop any unfair opportunities, network operations personnel must be able to ensure that the latencies between the exchange and each of the data centers are within 100 microseconds of each other.

This problem can be solved, just like many others by using current routers that typically support two distinct accounting mechanisms: SNMP and NetFlow. Each router cannot do the work all on its own, it needs the other to help pick up what it cannot. SMNP can only provide cumulative counters, which can be very useful when estimating load, however it cannot provide the latency estimates.  On the other hand NetFlow can give samples and timestamps to help with all the receiving packets and also allows the company to get a calculating of the latency requirements from multiple routers.

Tags: high frequency trading, low latency, ultra low latency

This entry was posted on Friday, January 8th, 2010 at 1:37 pm and is filed under Low Latency Resources, White Papers. You can follow any responses to this entry through the RSS 2.0 feed. You can trackback from your own site.