I currently have a Solaris u5 machine with 2gig memory and 6 disks in a RAIDZ config. When i write across the network to a smb/nfs share i notice a pause in the writes every 5seconds.
Now from what i've read every 5seconds a TXG group goes in to the quiesced state and then gets sync'd to disk. What i would like to know is can/should another TXG group be opening at the same time and still allow writes while the sync is taking place in the other TXG and should this cause a write pause to my smb/nfs application? It appears i have 1gig of ARC cache so i believe 500meg of that can be used for writes. I've done a few writes tests across the network at various speeds to smb/nfs shares varying from 40MB/sec to 90MB/sec and i still get the pauses. If i was doing 40MB/sec when i hit the 5second sync time i would have about 200MB in cache and i thought it could happily open another TXG without causing a pause/throttle to an application. The RAIDZ pool i have can do around 140MB/sec write speed(from a re-write test when cache is full) and iostat of the pool show that every 5seconds it's not having any issues getting the TXG group out, which appears to take about 1-2seconds. So i guess i'm asking is this expected behaviour and am i missing something? Here is an iostat snip when writing to the pool across the network at around 35-40MB/sec. I was still getting 1second pauses every 5seconds. silvia 1.76T 987G 0 971 0 120M silvia 1.76T 986G 0 276 0 20.1M silvia 1.76T 986G 0 0 0 0 silvia 1.76T 986G 0 0 0 0 silvia 1.76T 986G 0 0 0 0 silvia 1.76T 986G 0 984 0 122M silvia 1.76T 986G 0 262 0 18.3M silvia 1.76T 986G 0 0 0 0 silvia 1.76T 986G 0 0 0 0 silvia 1.76T 986G 0 0 0 0 silvia 1.76T 986G 0 972 0 120M silvia 1.76T 986G 0 308 0 19.6M Thanks for your time :). This message posted from opensolaris.org _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
