Extensible, Scalable Monitoring for Clusters of Computers
LISA '97 Proceedings of the 11th USENIX conference on System administration
BorderPatrol: isolating events for black-box tracing
Proceedings of the 3rd ACM SIGOPS/EuroSys European Conference on Computer Systems 2008
Diagnosing distributed systems with self-propelled instrumentation
Proceedings of the 9th ACM/IFIP/USENIX International Conference on Middleware
X-trace: a pervasive network tracing framework
NSDI'07 Proceedings of the 4th USENIX conference on Networked systems design & implementation
| Hi-index | 0.01 |
Performance analysis for a multi-tiered enterprise application is always challenging. Performance problems in one tier may have a "domino effect" on all other tiers, resulting in cross-tier waiting in all other tiers and the whole system under utilized. Traditional technologies that focus solely on internal causes of one tier are not sufficient in this case, while others that require distributed tracing each machine across all tiers are not feasible in many production systems. This poster presents an approach that focuses on the middle tier, typically the application server tier, to identify real bottleneck in a multi-tiered application with the support of a performance tool named SLICE. By selectively tracking method invocations that cross-tier boundaries, and extracting contextual information associated with these invocations, tier bottlenecks can be quickly identified. Experiments on DayTrader, a three-tiered application, have shown that performance bottlenecks caused by client machines or database servers can be quickly identified.