Information générale
Titre:
Write buffering performance problem
I have a larger VFP table (200K Rows, 145MB) that I scan through from top to bottom and replace one field in each record with a calculated field from another table. The table is not in a DBC.
The REPLACE commands run very quickly. However, when I try to close the table it takes upwards of 5 minutes to apparently commit all the changes. VFP "locks up", and after a very long time (several minutes later) continues on it's way like nothing out of the ordinary happened.
I can also simulate the problem by using the FLUSH command after doing all my replacing.
The table is on a file sever on the same Ethernet switch as the workstation running the code in the VFP environment (non-compiled prg calling some utility classes -- VCXs)
If I move the large table to the local drive the problem disappears.
I used NetXRay to watch the Ethernet traffic during the "FLUSH" and found many very short intense bursts of bi-directional traffic between the workstation and file server. In between, there are periods of absolutely no traffic. This is repeatable across many workstations. I can also FLUSH after doing say 100 - 1000 REPLACES and get the same behavior (although the pause is shorter.)
Has anyone run across this behavior? Is there some way to change the buffering or another table property to work-around this?
Thanks!
Suivant
Répondre
Voir le fil de ce thread
Voir le fil de ce thread à partir de ce message seulement
Voir tous les messages de ce thread
Voir tous les messages de ce thread à partir de ce message seulement