Nils Knafla's research project
Nils Knafla's research project description:
Two industry trends in the performance/price ratio of hardware systems have implications for the efficient implementation of object-oriented database management systems (OODBMSs) in a client/server computing environment. Firstly, the continuing fall in price of multiprocessor workstations means that such machines are cost effective as client hosts in OODBMSs. Secondly, although the performance/price ratios of both processors and disks are improving, the rate of improvement is greater for processors. Hence, the disk subsystem is emerging as a bottleneck factor in some applications. Recent advances in high bandwidth devices (e.g. RAID, ATM networks) have had a large impact on file system throughput. Unfortunately, access latency still remains a problem due to the physical limitations of storage devices and network transfer latencies.
There is a two order of magnitude gap in access time between semiconductor memory and disks, the rate of improvement for memory is also higher. Prefetching is an optimisation technique which can close this gap if object access is reasonably predictable. We are currently exploiting complex object relationships under which conditions prefetching can be successful.
My general research interest is the performance optimisation of object-oriented databases. Physical disk management, main memory management, clustering and prefetching are important areas for performance optimisation.
Here you find a Poster describing my PhD work.
Back to the home page.
Nils Knafla