Performance debugging for distributed systems of black boxes
SOSP '03 Proceedings of the nineteenth ACM symposium on Operating systems principles
VLDB '06 Proceedings of the 32nd international conference on Very large data bases
Hi-index | 0.00 |
Dependency among various machines, applications or different components of an application is an important issue for planning, operating and managing performance of data-centers and IT-infrastructure of an enterprise. Though this information is very critical, real-life experience with many data center administrators and users shows that it is often not available and it is gathered manually when the need arises. In this paper we present a method and tool for tracking the inter-dependence among IT infrastructure components in a data center using a combination of non intrusive and intrusive agent-based monitoring approach. If legal or operational constraints prevent use of the intrusive agent module, the tool still offers a lot of utility with its non-intrusive monitoring module and has zero impact on production environment. A useful feature of the tool is its ability to quantify the strength of dependency as per user-defined criteria. For web and database servers, the tool can carry out workload profiling on per-client, per-database, per-table basis without adding any overload on the servers. This significantly increases utility of the tool for additional business objectives like capacity planning and performance management. Efficacy of the solution has been proved in real-life case study carried out with large retail chain customer.