[opensuse] Slow SMB/CIFS transactions (compared to WinOS)
![](https://seccdn.libravatar.org/avatar/776ae9647c4c4b35b3958d71e7072114.jpg?s=120&d=mm&r=g)
I have to use CIFS shares a lot at work. I have found (with Ubuntu and openSUSE) that the shares takes much longer to navigate than on Windows (all versions). My machine is fast, CPU and RAM is no issue, so what else is Samba doing differently than Windows to navigate the shares? (does MS maybe cache a lot of the transactions? it seems so) Manne -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
![](https://seccdn.libravatar.org/avatar/aea1d8248292e6482742234c5cb514de.jpg?s=120&d=mm&r=g)
Manne Merak wrote:
I have to use CIFS shares a lot at work. I have found (with Ubuntu and openSUSE) that the shares takes much longer to navigate than on Windows (all versions). My machine is fast, CPU and RAM is no issue, so what else is Samba doing differently than Windows to navigate the shares? (does MS maybe cache a lot of the transactions? it seems so)
You could use a utility like wireshark to see what difference there are in traffic. I've noticed linix clients being notably slower with SMB & CIFS than windows machines as well. It's not an ideal solution, but if you have the option for those servers to export the files systems using NFS (i.e. -- if the servers are linux servers), you may get comparable performance. It seems NFS works better for linux clients, but I haven't been able to setup and benchmark clients of both types against servers of both types on identical hardware to know the exact numbers. If you come up with better numbers, (more accurate, substantial, or reasons), maybe you could help point toward some solutions. Some of it may be the how the machines are tuned at the different layers -- specifically -- TCP window sizes and such which windows usually has optimized for local-networks and low-delay only. *nix typically has better optimizations for wider-area networks and longer round-trip delays (though on a local network, linux machines will usually give superior performance when focusing on per-packet round trip times. Do you have any numbers of what is slower or not? Maybe linux is using higher values for 'timeouts' on something so it waits longer for things to be done? But that's pure guessing. Since my local servers usually run linux, I usually have it running both NFS and CIFS and use the 'appropriate' network protocol for the given client -- so I tend to notice "similar" ballpark performance numbers. If that's an option, you might try using NFS and see if that gives you similar performance...? linda -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (2)
-
Linda Walsh
-
Manne Merak