Automatically Discovering the Relationship Between End-to-End Transactions and their Dependent Resources
Publication Date: 2010-Sep-10
The IP.com Prior Art Database
This invention provides a generic solution that allows the dependencies between many levels of enterprise system components to be automatically discovered using only simple and lightweight monitoring data.
Automatically Discovering the Relationship Between End -to-End Transactions and their Dependent Resources
Disclosed is an approach for the automatic discovery of enterprise system component dependencies using only simple and lightweight monitoring data. For example, given a system that includes a payroll application and an applications database system, this invention is able to implicitly discover relationships between the two, thus precluding the need to define explicit relationship definitions.
No known solutions provide a truly generic approach to identifying hierarchical relationships
in enterprise systems. Current solutions are typically confined to specific levels of abstraction, such as the computer technology level (eg, network packets) or the logical application level (eg, transactions). Each level of abstraction is monitored using domain or
product specific knowledge, and then presented to the user in that frame of reference.
A significant drawback to focusing on particular levels of abstraction is the inability to accurately diagnose faults originating outside of the frame of reference in which the fault occurred. There must be significant investment in explicitly defining the relationships between the various monitoring technologies. For example, in the payroll application it is difficult to integrate hard disk monitoring information with the payroll and database monitoring data. The name and location of the individual files accessed by the application need to be provided to the hard disk monitoring system. It is possible that the names of files that the payroll application uses is not known by the end user. Creation and maintenance of such knowledge bases describing the monitoring relationships requires a significant investment of time and investigation.
This invention uses two logical components: a simple monitoring infrastructure and an offline analysis system.
The monitoring infrastructure is lightweight and deployed within or alongside
components of interest. A few key pieces of information are reported during any activity of interest: context information to define the type and nature of the activity (its label) and the time at which the activity starts and stops.
The offline analysis system, which is the basis of this invention, uses the data produced
by the monitoring infrastructure to deduce dependencies between different activities. Activities that frequently overlap or occur concomitantly, or in general have an observable relationship in time, are identified as being inter-dependent. Moreover, the nature of the time relationship (ie, intersecting, following, subsuming) identifies the type and hierarchy of the dependencies.
Consider the payroll example application where the application, database, file system, and the hard disk have all been instrumented by the monitoring infrastructure. Monitoring data will include:
Application: data for business related data flow as it passes between the software compon...