Adaptive middleware in go - a software architecture-based approach

Abstract

Adaptive middleware is essential for developing distributed systems in several applications domains. The design and implementation of this kind of middleware, however, it is still a challenge due to general adaptation issues, such as When to adapt? Where to include the adaptation code? What to adapt?, and How to guarantee safe adaptations?. Current solutions commonly face these challenges at the implementation level and do not focus on the safety aspects of the adaptation. This paper proposes a holistic solution implemented in Go programming language for developing adaptive middleware centred on the adoption of software architecture principles combined with lightweight use of formalisms. Software architecture concepts work as an enabling approach for structuring and adapting the middleware. Meanwhile, the formalisation helps in providing some guarantees before and during the middleware execution. The proposed solution is evaluated by implementing an adaptive middleware and comparing its performance against existing middleware systems. As shown in the experimental evaluation, the proposed solution enables us to design and implement safe adaptive middleware systems without compromising their performance.

Multiple transport protocols in an adaptive RPC-based framework

“…Figure 2 depicts the quartiles and median of the results for f ibonacci(2), i.e., an application with low processing demand. According to these results 1 , the flavours have the following order from the shortest mean response time to the longest one: HTTP (0.9907 ms), HTTPS (1.0503 ms), RPC (1.0523 ms), UDP (1.1027 ms), TCP (1.1303 ms), TCP+TLS (1.1570 ms), HTTP2 (1.5408 ms) and QUIC (1.6329 ms). By using the T-Test with a confidence of 95%, all samples have significant differences between them.…”

“…Before presenting the proposed solution, it is worth describing the framework gMidArch [1] that is implemented in the Go programming language. gMidArch is an RPC-based framework that helps in the design, implementation, and safe execution of adaptive middleware systems.…”

“…This benefit, however, was not explored in the experiments due to the single sequential requests to f ibonacci(N ). Also, as mentioned in Section 3.8, the HTTP2 extension uses reflection and annotations to decouple the business 1 Available at https://github.com/gfads/midarch/tree/master/evaluation code from the middleware one, and this feature decreases the middleware processing time. As for the QUIC extension, a possible reason for the high mean response time is the use of a non-stable Go package implementation of the QUIC protocol.…”

“…While the middleware community has widely incorporated adaptive mechanisms into middleware solutions [1], these solutions usually lack support to communication protocols' heterogeneity. In practice, existing non-adaptive and adaptive middleware systems usually work atop a unique transport protocol, e.g., TCP or HTTP.…”

“…This paper presents an extension of an existing framework named gMidArch [1]. gMidArch is an architecture-based framework in which middleware developers reuse a library of architectural elements to implement adaptive middleware systems in the Go programming language.…”