Hello, I'm trying to optimize a NFS server used by diskless workstations (Linux, FreeBSD, OpenVMS). This server actually runs NetBSD 10.99.14 but I have built a -11 beta kernel last night. legendre# df Filesystem 1K-blocks Used Avail %Cap Mounted on /dev/raid0a 32529068 15004030 15898586 48% / /dev/raid0e 65058298 31682850 30122534 51% /usr /dev/raid0f 32529068 18580496 12322120 60% /var /dev/raid0g 264277976 137092064 113972016 54% /usr/src /dev/raid0h 548684628 394784644 126465756 75% /srv /dev/dk0 3876580176 2400009320 1282741848 65% /home kernfs 1 1 0 100% /kern ptyfs 1 1 0 100% /dev/pts procfs 4 4 0 100% /proc tmpfs 4162676 1052 4161624 0% /var/shm /dev/dk6 11335898764 9407711308 1361392520 87% /opt/bacula /dev/dk7 11343502476 3864697556 6911629800 35% /opt/video raid0 : raid 1 volume raid1 : raid 5 volume /srv contains all root filesystems for clients. /home contains users data. swaps volumes are exported to clients through iscsi (istgt over a ccd device). NFS itself runs lika a charm. Throughput is very close to maximal throughput of LAN. Today, bufq.strategies built in kernel are: legendre# sysctl -a | grep bufq kern.bufq.strategies = priocscan fcfs disksort I have seen I have removed from kernel BUFQ_READPRIO. I don't remember why I have removed this strategy... and I have tried to find some informations about these strategies. fcfs seems to be a trivial FIFO queue. A have found few informations about priocscan strategy. But what is the difference between disksort and priocscan ? And between priocscan and readprio ? Does prioscan works in read and write and readprio only in read mode ? Best regards, JB
Attachment:
signature.asc
Description: OpenPGP digital signature