Little Bursts, Big Money
The financial services industry uses multicast signals to send live, redundant streams from multiple locations to many recipients. CBOE uses multicast to ensure customers get accurate quote and trade data.

"Financial multicast data is particularly bursty traffic. It doesn’t rate limit itself or adapt to the available bandwidth like TCP does," Rittner said. "Sub-second bursts can be quite large. Simply looking at router or switch statistics can mask these micro-bursts that we can find through time interval analysis. In the financial industry, losing any data is a big deal. That's why we use Observer Expert and GigaStor to ensure we see everything.

"For example, a few years ago we had a problem whereby some of our customers were seeing a small amount of packet loss in the multicast market data they received from us. We confirmed that the data was leaving our network and being sent to their WAN service provider. We also looked at the data rates, which seemed to be fine. It was only after days of investigation when we exported the captured data to Excel that we discovered some micro-bursts were exceeding the bandwidth of the customer’s WAN circuit. Now with the GigaStors and Observer we can perform the same analysis in minutes."

Financial Success Strategy
Rittner's team is tasked with network design and analysis for the CBOE LAN/WAN and networks for the CBOE Futures Exchange (CFE), CBOE Stock Exchange (CBSX) and the OneChicago Exchange.

"Often we find that the problem is congestion of WAN links not under our control," Rittner said. "We need to look at the traffic flowing over a given path and determine the bandwidth. The ability to use GigaStor on that is a big time saver."

CBOE placed GigaStors and Expert Consoles strategically across the network, the GigaStors being placed on core links and external-facing interfaces like connections to member firms and market data vendors.

"Network Instruments simply has a better value proposition," Rittner said. "The ability to capture and store huge amounts of network data and be able to run it against things like Connection Dynamics and Time Interval Analysis, all at a better price point, is an easy choice."

Previous Page