This section presents performance measurements of gms_net and sequential file access using the GMS/Trapeze prototype in Digital Unix 4.0. We measure all the RPC variants presented in order to illustrate the costs and benefits of the gms_net and Trapeze mechanisms discussed in the previous sections. The file access tests are intended to show the rate at which a GMS/Trapeze client can source and sink data to network storage servers using these communication mechanisms for payload transfer and asynchronous read-ahead. A secondary goal is to show the effect of the operating system kernel interface chosen to read or write file data at these speeds, which are close to the limits of the hardware.
The systems used for these measurements are Alcor and Miata, two DEC Alpha platforms based on the 21164 CPU. Our Alcors (AlphaStation 500 and 600) are clocked at 266 MHz, and use a CIA host-PCI bridge (ASIC pass 2). Miata (PWS 500au) is a newer 500 MHz machine with a Pyxis bridge (ASIC pass 257). All systems are equipped with M2F-PCI32 Myrinet LAN adapters connected through an 8-port switch (M2F-SW8). Both the CIA and Pyxis I/O bridges deliver almost the full bandwidth of the 32-bit 33MHz PCI standard (132 MB/s) in one direction; however, Alcor receives at half-bandwidth and Miata sends at half-bandwidth. Most of the experiments in this section involve a one-way high-volume data transfer: to circumvent the bridge bottlenecks Alcor is always the sender and Miata is always the receiver. Release of the improved Miata-II is imminent, but it is not yet available at the time of this writing.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
sec)