Ahmed Atef wrote:> I'm sorry for bothering you again but i have small question
> regarding the performance of SAMBA and its scalability ...
Don't apologize, you ask good questions (;-))
I'm ccing this to the list as people will want
to know, and the subject hasn't come up for
over a year, when an academic site worried about
classes all logging in at once.
> Total number of user login to SAMBA at the same time is 40 users
> and all will login at the same time also all will log-out at the
> same time ... everyone expected to have roaming profile around
> 30MB so long time will be taken to upload and download these
> huge profiles ... how is SAMBA expected to behave in this case ???
Ok, you can simulate this pretty easily with smbclient
and a script. And it's better than you'd expect...
Some experiments for you and the mailing-list to look at:
First experiment: 40 clients log in, transfer a size-zero file from
the server's local disk to /dev/null and log out,
---
$ ls -l ~/foo
-rw-r--r-- 1 davecb staff 0 Nov 13 09:54 foo
make_login_load 40 //elsbeth/davecb get foo /dev/null
Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.0-alpha0]
smb: \> smb: \> elsbeth>
---
and it ended at approximately Mon Nov 13 10:07:45 EST 2000, for
17 seconds, with a maximum number of SMBDs of 3. So far so good.
Then I improved the script (attached) and did 1, 10, 20, 30
and 40 clients each doing a get of a 30MB file, measuring high
and low times, giving us:
Clients Fastest Slowest MB/S
1 5.9 5.9 5.08
10 13.5 29.6 10.14
20 31.2 47.4 12.66
30 41.5 64 14.06
40 43.1 90 13.33
This is an upper bound (worst case) on the
performance of my small server:
it includes the time to disconnect again!