ARTICLES PCB Design for New I/O Interface "InfiniBand" Shared Bus Architecture In a bussed architecture, all communication shares the same bandwidth. The more ports added to the bus, the less bandwidth available to each peripheral. They also have severe electrical, mechanical, and power issues. On a parallel bus, there are many pins necessary for each connection (64 bit PCI requires ~90 pins), making layout of a board very tricky and consuming precious printed circuit board (PCB) space. At high bus frequencies, the distance of each signal is limited to short traces on the PCB board. In a slot-based system with multiple card slots, termination is uncontrolled and can cause problems if not designed properly. Figure 1: Conventional Shared Bus Architecture There is a load limit to a bus design that only allows a few devices per bus. Adding a bridge device to provide another bus with a new load limit behind the bridge overcomes this limitation. Although this allows for more devices to be connected to the system, data still flows through the central bus when accessing devices on other parts of the system. Latency and congestion increases with each bridge added to the system. A bus must be designed to operate under fully loaded conditions assuming the worst case number of devices allowed by the specification, which fundamentally limits the bus frequency. One of the major issues with a bus is that it can�ft support �gout of the box�h system interconnects. To get systems to talk together, a separate interconnect is required, such as Ethernet (server-to-server communication) or Fibre Channel (storage networking). Switched Fabric Architecture A switched fabric is a point-to-point switch-based interconnect designed for fault tolerance and scalability. A point-to-point switch fabric means that every link has exactly one device connected at each end of the link. Thus the loading and termination characteristics are well controlled and (unlike the bus architecture), with only one device allowed, the worst case is the same as the typical case and therefore I/O performance can be much greater with a fabric. Figure 2: Switched Fabric The switched fabric architecture provides scalability which can be accomplished by adding switches to the fabric and connecting more endnodes. Unlike a shared bus architecture, the aggregate bandwidth of a system increases as additional switches are added to the network. Multiple paths between devices keep the aggregate bandwidth high and provide fail-safe, redundant connections. There are three major system elements to an InfiniBand subnet. They are Host Channel Adapters (HCA), Target Channel Adapters (TCA) and switches. Hosts reside in the server and control the RDMA access. Targets are either native InfiniBand storage devices or routers that provide InfiniBand to Ethernet or Fibre Channel communications. To ARTICLE Top page <<Former / Next>> To DESIGN WAVE MAGAZINE's Homepage (English) To CQ Publishing's Homepage (Japanese) Copyright(C) 1996-2002 CQ Publishing Co., Ltd.