Thursday, April 15, 2004
Noon - 1 pm
Wean Hall 8220
The Case for Preemptive Lock Scheduling in OLTP Workloads
OLTP workloads are increasingly common in computer systems, such as e-commerce and inventory management. It is valuable to provide priority scheduling in these systems, to reduce the response time for the most important clients, e.g. the "big spenders." Two-phase locking, commonly used in DBMS, makes prioritization difficult, as transactions wait for locks held by others regardless of priority. Existing lock scheduling solutions, including non-preemptive priority inheritance and preemptive abort, have performance drawbacks for TPC-C type workloads.
We evaluate several existing preemptive and non-preemptive lock prioritization policies for TPC-C type workloads. We conclude that high-priority transactions benefit greatly under preemptive abort as compared with non-preemptive policies. The low-priority transactions, however, suffer excessively.We conduct a detailed statistical analysis, explaining the performance results observed.We propose and implement POW, a lock prioritization policy, that provides all of all the benefits of preemption, without its costs.
David is a 3rd year PhD student working under Mor Harchol-Balter and with Anastassia Ailamaki on database and schuling problems, particularly those hard ones involving concurrency control and data consistency. In his spare time he studies coffee and chocolate.
Seminar Info Contact:
or visit http://www.pdl.cmu.edu/SDI/