On 8/19/2011 8:49 PM, [email protected] wrote:
On 8/19/11 8:04 PM, Patrick M. Landry wrote:------------------------------------------------------------------------ *From: *"Alexander J. Maidak" <[email protected]> *To: *"Patrick M. Landry" <[email protected]> *Cc: *[email protected] *Sent: *Wednesday, August 17, 2011 9:07:12 PM *Subject: *Re: [lopsa-tech] Poor NFS performace from CentOS5 client to Oracle (Sun) Unified Storage NAS Is the filer more then 80% full? Is it slow for writes and reads or just writes? How slow is slow and what do you expect "fast" to be? It is not more than 80% full. The speed problem is most noticeable when asking for a listing in the directory with the 7000 subdirectories. It is not uncommon for that command to take 30 seconds to a minute or more to complete from the CentOS5 hosts. It doesn't take more that 5 seconds on the Solaris hosts.
Ah, I think I know exactly what this (above) is. Linux (CentOS) does not fully implement the READDIRPLUS NFS RPC call, but Solaris does. When the client asks for READDIRPLUS, the server fills in the packet with a large number of replies and sends them in bulk. When the client asks for READDIR, every single entry is enumerated as a single RPC which means 7000 RTTs. If you are using ls -l or ls -color, automatically double that to 14000 RTTs because it will be doing getxattr and/or stat calls for every entry. ouch!
The real crime is that CentOS even has it there in the kernel, but doesn't use it most of the time, only on an initial ls, usually.
Can you tell that I've seen this exact behavior?! ;) _______________________________________________ Tech mailing list [email protected] https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
