Shown in Table 5 are the Globus performances on the different individual clusters of the MecaGRID and the communication times relative to the nina values. The computational ratios shown in the tables that follow are based on the total computational times relative to the INRIA-nina cluster; Examining Table 5 one can imagine the difficulty in optimizing Grid computations due to the different processors speeds, RAM, cache, and LAN speeds on the different clusters. For the AGARD swept wing case the size of the executable is 236 MB and is less than the available RAM/processors on the different clusters. Thus one would expect that the INRIA-nina and the IUSTI clusters to show about the same performance as both have 2 GHz processors. Table 5 shows the IUSTI cluster to be slightly slower than the nina cluster. A possible explanation for this difference may the difference in LAN speeds, 1 Gbps for nina and 0.1 Gbps at the IUSTI. Likewise one might expect the CEMEF cluster to be approximately twice as slow as the nina cluster rather than a factor of 2.3 that again may be related to the different LAN speeds (1 versus 0.1 Gbps). One would also expect that the INRIA-pf cluster would be a factor of two slower than nina instead of a factor of 3.0. In general, it is difficult to evaluate the relative importance of the different cache sizes, LAN speeds, and the RAM available to a processor particularily for a small mesh.