Download presentation

The aim of rapidly developing virtualization technologies is as an increase of security and avalilability as improving efficiency of computational resources utilization. In practice, it means putting the load on physical resources up to the moment of saturation by data (I/O-bound). Schedulers allocate a physical CPUs time slice to each task. Being based on statistical characteristic, schedulers are solving an optimization problem to make the resource usage more effective. However, while optimizing the overall system performance, they don’t always take into accound the efficiency of every separate process. Meanwhile, some of them (like encryption and compression) are essential in private cloud environments. In the current paper, it’s presented an investigation on how to improve a performance of a stand alone block data processing algorithm while providing performance isolation from others.

Author: Aleksandr Povaliaev

Aleksandr Povaliaev

Senior software engineer, EMC

Studied in SPbETU «LETI», department of Computer Science and Informatics. Took part in establishing of a programming club of the University. Graduated in 2001, degree with honors. Then continued studying. Post-graduate research was on modeling of physical effects (grant from Russian Foundation for Basic Research). My carrier began with developing of client-server applications, then the area of interests shifted to the high-loaded systems. Currently, I am on a position of senior software engineer at EMC company – the world leader in information storage systems. I am an author of several scientific papers and inventions, lectures at the university and supervises a research project.