High frequency trading is a computer based trading system centred on stock markets. These systems analyse market data and take decisions and execute them within a very short time, even in nano second scale due to the very high competition. Software based algorithms have been increasingly improved to achieve this very low latency during the past years. These traditional systems connect with stock exchange through a standard NIC and then the OS kernel handles resource management, while the application stack performs the needed computations. Now focus has been given to hardware based improvements instead of this traditional system because, the standard NIC are not optimised to handle trade exchange protocols and interrupt based kernel OS causes long delays.
FGPA (Field programmable gate array) are used to custom build hardware that can concurrently use resources to give results in latency. The FGPA can be programmed to undertake both communication and trading strategies. This speeds up the round trip latency compared to software based systems. The memory used with these FGPA are a critical choice because of its effect on the performance. Most FGPAs use types of DRAMs but they are slow and are prone to errors. A better replacement is to use a Synchronous Static RAM which is faster than the former by a factor of about 24. Among SRAMs QDR family of RAMs have the highest performance.
Relation
Importance of processing speed improvement even if it is by a very small amount in the nanosecond scale, when particular applications like stock exchange is considered.
The need for the memory to be as of comparable speed as the speed of processor because if it were slow then the higher speed in the processor becomes ineffective.
The properties of DRAMS and SRAMS, their differences and suitability for different applications.
Message
The scientific community is in a very competitive race towards making processors faster and faster.