nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Samba Response Time from US to UK

System

Hello,
Our TeamSite server is hosted in the US, however we have users in the UK. The US users experience acceptable response time from the samba connection to the server in the US, but the UK users experience reponse times that far exceed any reasonable performance. The file looks like it is read several times and in very small packets before displaying to the user. We have adjusted the send and receive byte settings in socket options and on the server but it did not appear to help. Has anyone out there experienced such a problem with samba and how did you address it?
Desperate for advice.
Mike

Find more posts tagged with

Comments

clogan

We are experiencing the same problem between our New York and Montreal offices. We are looking into ways to improve the network in general between the two offices and hope samba will become workable too.

nipper

I am curious, can you check out "normal" network performance between the 2 offices ? Maybe a Unix
server that can NFS mount the TS server or a windows server close to the TS server that you can mount
to Europe ?

Compare the performance of Samba to NFS. Cut the number of variables. For all we know, you have a 56 KB
dialup line to Europe. That would really suck.

Andy

Migrateduser

Andy,
We have tried tests from our NFS mounts and the tests come back with slight response issues, but the files are read only once and the packets are transferred very efficient. Any other ideas.
Mike

skip11

You probably want to check the wire speed within your lan (probably 10/100mbps) and the wan
connection speed. Unless you have a dedicated line like an ATM (a purchased lease line, or other)
you're going over the public net hopefully with a VPN. Many factors can influence latentcy in a network;
firewalls, poor routing (unecessary routing flaps), highly loaded gateways, etc.

direct access over a long haul line is going to be problematic if there are bandwidth issues, IMHO.

R.Barger
Credit Suisse Group
Zurich, Switzerland

skip11

Hi,

BTW, Samba connections can run over UDP protocol stack, and can use
DNS to resolve hostnames. Are you using different DNS hosts between the
two offices ? On solaris check /etc/resolv.conf for the order and correctness
of DNS servers, and make sure /etc/nsswitch.conf is appropriately configured.

SMB uses UDP on ports 137,138, and TCP on port 139. Use a hacker tool like tcpdump
to check the port traffic on the SMB server for the interface the packets are outbound on.
A version is available for free for all Solaris versions here: http://sunfreeware.com

example:
tcpump -i eth1 host <ip or hostname, ip better> and \(port 137 or port 138 or port 139\)
tcpdump -i eth1 192.168.0.1 port 137

check the man pages for tcpdump for all switches. I perfer to dump to a file with -w file
and read back in with -r file. -v (verbose) is good to see the TTL for each packet in the transmission.
For an mountain of output data use -vv and -vvv, each a more verbose output.

You can also use snoop as well if your sysadmins don't want robust packet capturing programs
on thier systems. snoop comes out of the box with Solaris.

If your SMB server is windows, use ethereal, but you will need administrator rights to access
the network device to put ethereal in promiscuous mode.

Anyway, IMHO, if there are inordinate amounts of DNS UDP lookups, performance for the transmission of
the entire packet frame could be, ah, problematic if the TTL for the whole transmission is overrun. My
experience with network performance issues has often led me to the use of these tools as a basis for
pointing to network config problems, or to eliminating them.

R.Barger
Credit Suisse Group
Zurich, Switzerland