Please use this identifier to cite or link to this item: https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/461
Title: File based fast Process State Recovery in an Automated Trading System
Authors: Ruksan, M.M.M.
Issue Date: 22-Oct-2013
Abstract: An Automated Trading System is a high performance system which typically handles a very high message rate as high as 30000 to 50000 per second and still demanding a very low latency. The system consists of a large number of processes which process messages received from various other processes. The system should guarantee as much as 99% availability and to fulfil this requirement, the system is operated in a fully fault tolerant mode where each process is mirrored with a backup process. Once a process fails the mirror process takes over and the crashed process is warm started. The purpose of the warm start is to recover the business state of the process to the state before the failure. Currently warm starting of the process is done by replaying all the messages received by the crashed process until the time of failure. This method is trivial but the time taken may be very long when the number of messages to process is large. This project was to come up with a possible alternative mechanism to warm start a process which would give a better performance. This approach preserves the state of the process on disk and recovers its state from disk. A C++ frame work library called „Hibernate‟ is developed which can be used to develop processes with the capability to be swapped to disk and recover the state from disk. A sample process is also developed using the frame work and a comparison is done to measure the performance of the new methodology with the trivial approach used in the current system. The results obtained in the evaluation section are very impressive as the process recovery from disk is much faster than the conventional replay approach. Thus this developed framework can be used in the Millennium Exchange product with some additional architectural changes as discussed in the final chapter.
URI: http://hdl.handle.net/123456789/461
Appears in Collections:Master of Computer Science - 2012

Files in This Item:
File Description SizeFormat 
Project_Report_2009_MCS_053_M_M_M_Ruksan.pdf
  Restricted Access
1.46 MBAdobe PDFView/Open Request a copy


Items in UCSC Digital Library are protected by copyright, with all rights reserved, unless otherwise indicated.