[DRBD-user] drbd tranfer limitations (random and sequential acccess) using raw

Thomas Dingermann td at trobisch.de
Thu Sep 23 14:47:02 CEST 2004

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.

crsurf schrieb:

>Hello list
>I´m using drbd in a Gigabit network (crossover cable) and when I synchronize data (sequential read/write) the tranfer hit to 30~40 Mb/s, but when I make inserts into database (random read/write) that access drbd device via raw, the transfer rate hit only 5~6 Mb/s. Exists some way to improve this performance?
>I increased the sndbuf-size to 256K, but no such effect. I increase interface MTU to 9000 and no such effect too.
>One database process that run in 4 min. without replication, run in 30 min. with replication.
>Now I´m trying using filesystem jfs instead raw to view if the performance will be improved.
>Maybe sync-nice can help us?
I think drbd is not the right thing for databases with small block-sizes (my INFORMIX uses 2048 bytes, see separate thread here). My SybaseSQL-Anywhere is running very fast on drbd-replicated-ext3-filesystem. I dont know if you use ASE, but you should try using larger block-sizes (if supported).

I got an test-prog from an informix-consultant wich runs on local disk in 7 seconds. It takes 8 minutes on drbd-replicated-ext3-filesystem. I am looking forward for an solution...

Best regards 


More information about the drbd-user mailing list