Project
This study focuses on designing new Mesh of Trees based packet switch architecture and is inspired by buffered crossbar switch. Unlike conventional crossbar with or without crosspoint buffers which require complex hardware arbiters to resolve output contentions, simple arbitration processes are distributed over the buffers in the tree and this makes MOTS(N) scalable. Rather, it has simple and automatic fixed routing for cells from input ports to output ports. As variations of MOTS(N), improved version of MOTS(N) denoted as IMOTS(N) and IMOTS(N) with cut-through denoted as IMOTS-CT(N) are also proposed and studied.
This study focuses on designing new packet switch architecture which can tolerate output contentions automatically with less hardware components than ones in the conventional crossbar switches. CTC(N) switch can tolerate output contentions by a pipelining mechanism with buffers in the input ports. These buffers are used to decouple the scheduling task into N independent parts in such a way that N schedulers are located in N input ports, which operate independently and in parallel. As a result, CTC(N) switch is simpler and more scalable than existing crossbar switches.
The focus of this study is to build models to predict fault-prone modules using static and dynamic metrics rather than just static metrics as many other studies do. To make the scope of this study as broad as possible, five different tools (Logiscope, McCabe, Understand 2.0, CodeSurfer and cSuds) have been used to collect 85 static and 5 dynamic metrics for the target programs. Principal component analysis and other statistical techniques are used for the model construction. Such models are then validated by using defects collected for open source software.