Publication

Systems Group Master's Thesis, no. ETH Zürich; Department of Computer Science, April 2009
Supervised by: Prof. Nesime Tatbul
Financial Market Data has a trend of ever increasing growth in volumes. Increased volume of data put a burden on processing capabilities of market data processing systems. Financial companies and other market participants need low latency access to recent market data to be very responsive and high throughput to have more processing capability. The only way to cope with increasing volume of data and processing complexity is through improving their systems in the direction of high-throughput and low-latency. In this thesis, we worked on such a real-world production level market data processing system. We thoroughly analyzed and benchmarked this system by setting up a replica of the real-world scenario in our lab. We investigated the problems and bottlenecks in this system in-depth and came with solutions to improve the overall performance of such a system. We experimentally evaluated our improvement idea and compared the results with the current solution. We have shown up to %26 improvement in the overall latency of the system.
@mastersthesis{abc,
	abstract = {Financial Market Data has a trend of ever increasing growth in volumes.
Increased volume of data put a burden on processing capabilities of market data
processing systems. Financial companies and other market participants need low
latency access to recent market data to be very responsive and high throughput
to have more processing capability. The only way to cope with increasing volume
of data and processing complexity is through improving their systems in the
direction of high-throughput and low-latency. In this thesis, we worked on such
a real-world production level market data processing system. We thoroughly
analyzed and benchmarked this system by setting up a replica of the real-world
scenario in our lab. We investigated the problems and bottlenecks in this system
in-depth and came with solutions to improve the overall performance of such a
system. We experimentally evaluated our improvement idea and compared the
results with the current solution. We have shown up to \%26 improvement in
the overall latency of the system.},
	author = {Cagri Balkesen},
	school = {ETH Z{\"u}rich},
	title = {Analysis, Benchmarking and Performance Improvement of a FAST Implementation},
	year = {2009}
}