Many data-intensive services exist, which create value for different stakeholders. Examples include mobile data analytics services, sensors collecting information for energy management of residences, or network equipment collecting data at high speed for traffic analysis. Such services impose performance requirements for the platform, which is used for implementation of data processing functionalities.
This paper focuses on evaluating the impact of disk-based RAID and virtualization, when designing and implementing a platform for write-intensive applications. The results indicate that hard drive speed, and RAID configuration both have a significant impact on performance. However, their effect depends on size of data, and utilization of direct I/O for writing. Virtualization on Eucalyptus cloud platform had a significantly negative effect on write performance.
Pekka Pääkkönen and Daniel Pakkala (VTT): The implications of disk-based RAID and virtualization for building a platform for write-intensive services
Presented at SAC 2015, the 30th ACM/SIGAPP Symposium On Applied Computing, 13-17.2015, Salamanca