Using AOP for detailed runtime monitoring instrumentation

Abstract

Although AOP has long been used for monitoring purposes, the level of detail that virtually all AOP frameworks support is not enough to support broad classes of monitoring applications. Often the method or function call is the basic unit of code weaving, and if some data access weaving is supported it is usually very limited, e.g., to object field access. In this paper we demonstrate the need for AOP to be extended if it is to support broad runtime monitoring needs, and then present two new joinpoint types for AspectJ, the basic block and loop back edge types. These allow much more fine-grained weaving of advice, which in turn supports a much larger category of runtime monitoring applications. We demonstrate the feasibility of these extensions with example prototype implementations using the AspectBench Compiler (abc).

Dedicated Static Sampling Pointcut Designators

“…That location could be a decision structure or a loop back-edge or any executable statement. In previous work, we defined the point to be three possibilities [10]: Code: In this case, the weaving is based on a particular point or set of points that exist in code. Time: Using time as a point means that sampling is controlled by some kind of timing theme.…”

“…This program converts a JPEG image into textual HTML-enhanced "ASCII art". This test used the basicblock designator [10] combined with one of the sampling designators; e.g., the pointcut expression is like "before: basicblock() && weaveProbability(.3)". The advice is a simple basic block profiler that counts how many times each block has executed Fig.…”

The denotational semantics of a pre-woven language

“…Then using some rules, these aspects are applied to an underlaying program code. This gives developers a tool to factor out the crosscutting functionality in order to be replicated as many times as need in the program source A basic AOP model defines some specific fundamental pointcut designators (PCD's), 1 which are features in the program execution where the advice of an aspect can be weaved public aspect AllMethodExecution { private int exeCount; private pointcut allExecutions(): execution ( * * . * (..)); before(): allExecutions(){ print("Starting " + thisJoinPoint); System.out.println(++exeCount); } } } Fig.…”

“…The weaving could also be dynamic. Meaning there is a place into where the weaved advice could be inserted at runtime [1].…”

Monitoring of Client-Cloud Interaction

“…Techniques for automated instrumentation are code rewriting [131], bytecode instrumentation [12,29,132], and insertion of advice statements for aspectoriented programming [134,135,149]. Instrumentation techniques can be furthermore distinguished into static and dynamic, and instrumentation typically affects the performance of the software.…”