|
Hello Brian,
Sorry, I knew I should have sent more.
I am using LVS-DR all on the some network director has one NIC.
the director:
Slackware 8
kernel 2.4.19(RML's preempt kernel/lock break patches)
300Mhz K6-2
128MB RAM
the client was pretty much the same as the director
The Realservers
Slackware 8
kernel 2.4.19(RML's preempt kernel/lock break patches)
233Mhz K6
128MB RAM
Do not use a preemtible kernel for TCP related work. Preemptible kernels are for
workstations, rendering and analog synthesizing à la Rebirth 338 or TB-808 :).
Maybe we could come up with like a template for people to fill out when
sending to the list so they will send enough information for people to
work with. That would help for cases like mine and for people asking for
help. Just an idea. Let me know if you have any other questions.
Well, I sort of gave a template in my last posting:
If you do tests, you need to give the community more information about:
o exact kernel (vanilla + all patches, .config)
o machine: CPU, mainboard, RAM, PCI-slot speed
o network topology and involved HW
o /proc/sys/net/ipv4/* settings
o /proc/sys/net/ipv4/vs/* settings
o how much free idle and CPU time the machines have under load test
o cat /proc/slabinfo
o vmstat
o ratio of 'TX/RX packet' rate ---> lost packets in %
Also interesting would be:
o TCP/UDP packet size
o timing: how many bytes and packets have been sent over what timeframe
o recv/send buffer settings in /proc/...
o a gnuplot graph if possible
Joe has done some extensive testing [1], although he's not given all of the
above information. This was very helpful in deciding tcp defense strategies for
SYN flooding or QoS rate policies.
You could always come up with such a form, I'll will help you with input.
[1]
http://www.linux-vs.org/Joseph.Mack/performance/single_realserver_performance.html
--
echo '[q]sa[ln0=aln256%Pln256/snlbx]sb3135071790101768542287578439snlbxq' | dc
|
