Contents   Index

Evaluation

$\mathcal{ERRARE\ HUMANUM\ EST}$

These are the conclusion that can be drawn from the running times:

• If the cached files are used then the gain (compared to no caching) is approximately 95% (e.g., 2d (50) compared to 2c (982)) when weakly connected, and approximately 50% when fully connected (e.g., 1c (2) compared to 1d (1)).
• Disconnected operation has no overhead (1f, 3e).
• The overhead when weakly connected is negligible (2c,4c, and 4d).
• The overhead when fully connected is acceptable (1c, 3c, and 3g)--although in percentage rather high, then it must be remembered that the overhead only occurs on open and close, and shared files should have a lifetime that make this overhead negligible (i.e., short-lived temporary files should not use $\mathcal{P}$$e$$\mathcal{S}$$t$$\mathcal{O}$).
• Locking requires virtually no overhead (1d compared to 1h, and 2d compared to 2f).

All in all, I find these results to be acceptable. The overhead when fully connected may need some improvement.