An environment for automated performance evaluation of J2EE and ASP.NET thin-client architectures

Abstract

Assessing the likely run-time performance of applications using thin-client architectures during their design is very difficult. We describe SoftArch/Thin, a thinclient test-bed generator that synthesises performance testbed thin-client and server code from high-level software architecture models. This generated code is performance tested using a third-party tool and the results summarised. Architecture models can be evolved and tests repeated during application development to inform software engineers of realistic performance characteristics of their designs. Our environment currently supports J2EE and ASP.NET-based thin-client code generation and performance testing.

MDABench: Customized benchmark generation using MDA

“…In such applications, measurements in the form of benchmarking [6] and prototyping [14] are used to obtain valuable information for architects. Comprehensive prototyping can however be expensive.…”

“…The semantic links between our generated benchmark suite code and performance analysis models then exists since they derive from the same application design model. Some pioneering work has been done on generating benchmark and prototyping applications using models, as in [13,14] [16,20]. However, these have several limitations:…”

Benchmark Generation Using Domain Specific Modeling

“…Similar to DSM, MDA supports code generation from architecture designs directly. Some pioneering work has been done on generating benchmark and prototyping applications using models, as in [17][18][19]. In particular, [17] introduces SoftArch/MTE, a tool to generate deployable test-beds for distributed systems from high level specifications, automatically deploy and send testing results back to those models.…”

Model driven benchmark generation for web services

“…Some pioneering work has been done on generating benchmark and prototyping applications using models, as in [14,15] for component-based distributed applications. However, it has several limitations in terms of standards compatibility, leveraging existing code generation "cartridges" and load test modeling [27].…”

“…For complicated commercial systems, measurements in the form of benchmarking [16] and prototyping [15] are used to obtain valuable information for architects. However, comprehensive benchmarking and prototyping can be very time consuming.…”