I have installed NFS but get bad performance.
Copy 5MB from NFS Server (OpenVMS) to NFS Client (Linux)
real 0m19.284s
Copy 5MB from NFS Client to NFS Server
real 0m2.193s
Doing the same with ftp:
ftp> get
5483813 bytes received in 0.69 seconds (7.8e+03 Kbytes/s)
ftp> put
5483813 bytes sent in 0.67 seconds (8e+03 Kbytes/s)
Putting the nodenames in each local hosttable did not improve.
Tried another VMS-server with the same result.
First server is OpenVMS 8.3 and FTP V5.6 - ECO 3
Second server is OpenVMS 7.3-2 and FTP V5.4 - ECO 4
Linux client is Redhat
Occassionally I have had the same speed on NFS as on FTP,
but only on the first attempt. The consequent attempts were slow.
Now it is permanently slow.
I don't think this is a tuning issue while copying one 5 Mb file.
Any ideas?
Regards /Jan
in SYS$ETC: file: SYSCONFIGTAB.DAT
ovms_xqp_plus_enabled=0x00000007
And in
SYS$STARTUP: TCPIP$SYSTARTUP.COM
add:
$DEFINE/SYSTEM/EXEC TCPIP$CFS_SHOW_VERSION "OFF"
$define/system/exec TCPIP$CFS_TRANSFERSIZE 16384
$define/system/exec TCPIP$CFS_ODS_CACHE_SIZE 1024
$!
The above MAY help. No garantees.
Unfortunatly it did not make any change.
Thank for the try anyway.
> Unfortunatly it did not make any change.
> Thank for the try anyway.
You need to restart the TCPIP stack for those changes to take effect.
@SYS$STARTUP:TCPIP$SHUTDOWN
followed
@SYS$STARTUP:TCPIP$STARTUP
Done the restart and checked settings after that. No change.
Have resetted everything back to scratch.
[sys_jan@imlab01 ~]$ mkdir pdf
[sys_jan@imlab01 ~]$ time cp /mnt/janne/CONFIG_MANAGE.PDF ./pdf/
real 0m0.034s
[sys_jan@imlab01 ~]$ time cp /mnt/janne/CONFIG_MANAGE.PDF ./pdf/
real 0m0.036s
Now copying in the other direction:
[sys_jan@imlab01 ~]$ time cp ./pdf/CONFIG_MANAGE.PDF /mnt/janne/
real 0m4.610s
[sys_jan@imlab01 ~]$ time cp ./pdf/CONFIG_MANAGE.PDF /mnt/janne/
real 0m4.557s
After that copying to the Linuxclient is slow:
[sys_jan@imlab01 ~]$ time cp /mnt/janne/CONFIG_MANAGE.PDF ./pdf/
real 0m19.658s
I know this is an old thread.
The parameters I had given to the OP had made a difference to me, but
now, I am realising that the difference was in pulling up a list of
files in a directory. (aka: when on the Mac,s Finder, you click on a
folder, how long it takes for the filenames to appear in it).
Right now, I am copying entire user directory structure and it is
painfully slow. (aka: unix/Mac readuing files stored on the VMS system.)
Perhaps it is the cp -R command which is using nfs resources
inefficiently. (perhaps it does byte by byte read/writes ?)
For bulk transfer between fileservers/filesystems I'd use "rsync".
With proper switches it reduces the number of bytes transferred
by compression and also acts like "make" by inspecting file dates
and only transmits "newer" files.