path: root/docs/docbook/projdoc/Speed.xml

<chapter id="speed">

<chapterinfo>
	<author>
		<firstname>Paul</firstname><surname>Cochrane</surname>
		<affiliation>
			<orgname>Dundee Limb Fitting Centre</orgname>
			<address><email>paulc@dth.scot.nhs.uk</email></address>
		</affiliation>
	</author>
	&author.jelmer;
	&author.jht;
</chapterinfo>

<title>Samba Performance Tuning</title>

<sect1>
<title>Comparisons</title>

<para>
The Samba server uses TCP to talk to the client. Thus if you are
trying to see if it performs well you should really compare it to
programs that use the same protocol. The most readily available
programs for file transfer that use TCP are ftp or another TCP based
SMB server.
</para>

<para>
If you want to test against something like a NT or WfWg server then
you will have to disable all but TCP on either the client or
server. Otherwise you may well be using a totally different protocol
(such as Netbeui) and comparisons may not be valid.
</para>

<para>
Generally you should find that Samba performs similarly to ftp at raw
transfer speed. It should perform quite a bit faster than NFS,
although this very much depends on your system.
</para>

<para>
Several people have done comparisons between Samba and Novell, NFS or
WinNT. In some cases Samba performed the best, in others the worst. I
suspect the biggest factor is not Samba vs some other system but the
hardware and drivers used on the various systems. Given similar
hardware Samba should certainly be competitive in speed with other
systems.
</para>

</sect1>

<sect1>
<title>Socket options</title>

<para>
There are a number of socket options that can greatly affect the
performance of a TCP based server like Samba.
</para>

<para>
The socket options that Samba uses are settable both on the command
line with the <option>-O</option> option, or in the &smb.conf; file.
</para>

<para>
The <parameter>socket options</parameter> section of the &smb.conf; manual page describes how
to set these and gives recommendations.
</para>

<para>
Getting the socket options right can make a big difference to your
performance, but getting them wrong can degrade it by just as
much. The correct settings are very dependent on your local network.
</para>

<para>
The socket option TCP_NODELAY is the one that seems to make the
biggest single difference for most networks. Many people report that
adding <parameter>socket options = TCP_NODELAY</parameter> doubles the read 
performance of a Samba drive. The best explanation I have seen for this is
that the Microsoft TCP/IP stack is slow in sending tcp ACKs.
</para>

</sect1>

<sect1>
<title>Read size</title>

<para>
The option <parameter>read size</parameter> affects the overlap of disk
reads/writes with network reads/writes. If the amount of data being
transferred in several of the SMB commands (currently SMBwrite, SMBwriteX and
SMBreadbraw) is larger than this value then the server begins writing
the data before it has received the whole packet from the network, or
in the case of SMBreadbraw, it begins writing to the network before
all the data has been read from disk.
</para>

<para>
This overlapping works best when the speeds of disk and network access
are similar, having very little effect when the speed of one is much
greater than the other.
</para>

<para>
The default value is 16384, but very little experimentation has been
done yet to determine the optimal value, and it is likely that the best
value will vary greatly between systems anyway. A value over 65536 is
pointless and will cause you to allocate memory unnecessarily.
</para>

</sect1>

<sect1>
<title>Max xmit</title>

<para>
At startup the client and server negotiate a <parameter>maximum transmit</parameter> size,
which limits the size of nearly all SMB commands. You can set the
maximum size that Samba will negotiate using the <parameter>max xmit = </parameter> option
in &smb.conf;. Note that this is the maximum size of SMB requests that 
Samba will accept, but not the maximum size that the *client* will accept.
The client maximum receive size is sent to Samba by the client and Samba
honours this limit.
</para>

<para>
It defaults to 65536 bytes (the maximum), but it is possible that some
clients may perform better with a smaller transmit unit. Trying values
of less than 2048 is likely to cause severe problems.
</para>

<para>
In most cases the default is the best option.
</para>

</sect1>

<sect1>
<title>Log level</title>

<para>
If you set the log level (also known as <parameter>debug level</parameter>) higher than 2
then you may suffer a large drop in performance. This is because the
server flushes the log file after each operation, which can be very
expensive. 
</para>
</sect1>

<sect1>
<title>Read raw</title>

<para>
The <parameter>read raw</parameter> operation is designed to be an optimised, low-latency
file read operation. A server may choose to not support it,
however. and Samba makes support for <parameter>read raw</parameter> optional, with it
being enabled by default.
</para>

<para>
In some cases clients don't handle <parameter>read raw</parameter> very well and actually
get lower performance using it than they get using the conventional
read operations. 
</para>

<para>
So you might like to try <parameter>read raw = no</parameter> and see what happens on your
network. It might lower, raise or not affect your performance. Only
testing can really tell.
</para>

</sect1>

<sect1>
<title>Write raw</title>

<para>
The <parameter>write raw</parameter> operation is designed to be an optimised, low-latency
file write operation. A server may choose to not support it,
however. and Samba makes support for <parameter>write raw</parameter> optional, with it
being enabled by default.
</para>

<para>
Some machines may find <parameter>write raw</parameter> slower than normal write, in which
case you may wish to change this option.
</para>

</sect1>

<sect1>
<title>Slow Logins</title>

<para>
Slow logins are almost always due to the password checking time. Using
the lowest practical <parameter>password level</parameter> will improve things. 
</para>

</sect1>

<sect1>
<title>Client tuning</title>

<para>
Often a speed problem can be traced to the client. The client (for
example Windows for Workgroups) can often be tuned for better TCP
performance. Check the sections on the various clients in 
<link linkend="Other-Clients">Samba and Other Clients</link>.
</para>

</sect1>

<sect1>
<title>Samba performance problem due changing kernel</title>

<para>
Hi everyone. I am running Gentoo on my server and samba 2.2.8a. Recently
I changed kernel version from linux-2.4.19-gentoo-r10 to
linux-2.4.20-wolk4.0s. And now I have performance issue with samba. Ok
many of you will probably say that move to vanilla sources...well I ried
it too and it didn't work. I have 100mb LAN and two computers (linux +
Windows2000). Linux server shares directory with DivX files, client
(windows2000) plays them via LAN. Before when I was running 2.4.19 kernel
everything was fine, but now movies freezes and stops...I tried moving
files between server and Windows and it's trerribly slow.
</para>

<para>
Grab mii-tool and check the duplex settings on the NIC.
My guess is that it is a link layer issue, not an application
layer problem.  Also run ifconfig and verify that the framing
error, collisions, etc... look normal for ethernet.
</para>

</sect1>

<sect1>
<title>Corrupt tdb Files</title>

<para>
Well today it happend, our first major problem using samba.
Our samba PDC server has been hosting 3 TB of data to our 500+ users
[Windows NT/XP]  for the last 3 years using samba, no problem.
But today all shares went SLOW; very slow. Also the main smbd kept
spawning new processes so we had 1600+ running smbd's (normally we avg. 250).
It crashed the SUN E3500 cluster twice. After alot of searching I
decided to <command>rm /var/locks/*.tbl</command>. Happy again.
</para>

<para>
Q1) Is there any method of keeping the *.tbl files in top condition or
how to early detect corruption?
</para>

<para>
A1) Yes, run <command>tdbbackup</command> each time after stoping nmbd and before starting nmbd.
</para>

<para>
Q2) What I also would like to mention is that the service latency seems
alot lower then before the locks cleanup, any ideas on keeping it top notch?
</para>

<para>
A2) Yes! Samba answer as for Q1!
</para>

</sect1>

</chapter>