Although RRVM does not impose any significant network load on top of modern interconnection networks (e.g. FDDI), we would like to investigate whether RRVM imposes any significant computation load on the server workstation. For this reason, we repeated the previous experiments, with the difference that all RRVM servers were running on a single workstation. This experiment puts pressure not only on the network, but on the single server workstation as well. The performance of RRVM as a function of the number of participating clients is shown in figures 7 and 8 for I/O block sizes 32 bytes and 2 Kbytes respectively. Both figures suggest that there is some performance decrease as the number of participating workstations increases, esp. for small block sizes. On the other hand, the total number of transactions per second sustained stays particularly high in all cases ranging from 712 to 1076 transactions per second for FDDI-based systems, and from 352 to 544 transactions per second or Ethernet-based systems.
The performance of RVM is much lower, reaching at most 128 total transactions per second for the four RVM systems executing.
Figure 7: Server Load: Performance of RVM as a function of the number of the participating clients - sequential accesses - all servers run on a single workstation - I/O block size = 32 bytes.
Figure 8: Server Load: performance of RVM as a function of the number of the participating clients - sequential accesses - all servers run on a single workstation - I/O block size = 2 Kbytes.