Help with lb6m lacp for increase bandwidth

[jorgefreyre]

Need help.

i have a lb6m, i want actívate lacp on 2 groups of port (13, 15, 16, 17) for get 20GB bandwidth,

config:

(vm-admin) #show running-config
!Current Configuration:
!
!System Description "Quanta LB6M, 1.2.0.18, Linux 2.6.21.7"
!System Software Version "1.2.0.18"
!System Up Time "11 days 18 hrs 25 mins 14 secs"
!Additional Packages FASTPATH QOS
!Current SNTP Synchronized Time: Sep 5 02:56:43 202 UTC
!
set prompt "vm-admin"
network protocol none
network parms 192.168.220.49 255.255.255.0 192.168.220.1
serviceport protocol none
serviceport ip 192.168.200.49 255.255.255.0 192.168.200.1
vlan database
exit
ip ssh server enable
ip ssh protocol 2
sshcon timeout 160
configure
ip name server 8.8.8.8
username "admin" password 82f53ac17a898ed627ea26c2736df53e425f1d21a3a12810a251893865d2abb96d19beae497ca8b55f484e4a5250ed451ba6b331195bceaa41d8ebbf75bff69f level 15 encrypted
aaa authentication enable "enableList" enable
line console
exit
line telnet
exit
line ssh
exit
spanning-tree configuration name "2C-60-0C-D0-D2-E5"
port-channel name 1/1 r720-2
port-channel linktrap "1/1"
interface 0/6
no auto-negotiate
addport 1/1
exit
interface 0/7
no auto-negotiate
addport 1/1
exit
port-channel name 1/2 r720-1
port-channel linktrap "1/2"
interface 0/1
no auto-negotiate
addport 1/2
exit
interface 0/2
no auto-negotiate
addport 1/2
exit
port-channel name 1/3 r820-1
port-channel linktrap "1/3"
interface 0/9
no auto-negotiate
addport 1/3
exit
interface 0/11
no auto-negotiate
addport 1/3
exit
port-channel name 1/4 r610-3
port-channel linktrap "1/4"
interface 0/15
no auto-negotiate
addport 1/4
exit
interface 0/16
no auto-negotiate
addport 1/4
exit
port-channel name 1/5 r610-2
port-channel linktrap "1/5"
interface 0/13
no auto-negotiate
addport 1/5
exit
interface 0/17
no auto-negotiate
addport 1/5
exit
snmp-server sysname "vm-admin"
!
interface 0/13
lacp actor admin state passive
exit
interface 0/15
lacp actor admin state passive
exit
interface 0/16
lacp actor admin state passive
exit
interface 0/17
lacp actor admin state passive
exit
interface 1/1
port-channel load-balance 6
exit
interface 1/2
port-channel load-balance 6
exit
interface 1/3
port-channel load-balance 6
exit
interface 1/4
no port-channel static
port-channel load-balance 6
exit
interface 1/5
no port-channel static
port-channel load-balance 6
exit
router rip
exit
router ospf
exit
exit




port 13 and 17 ------ conected to ------ HOST esxi vm 1
port 15 and 16 ------ conected to ------ HOST esxi vm 2

status

vm 1:

root@pgbouncer:~# iperf -c 192.168.220.247
------------------------------------------------------------
Client connecting to 192.168.220.247, TCP port 5001
TCP window size: 85.0 KByte (default)
------------------------------------------------------------
[ 3] local 192.168.220.248 port 47070 connected with 192.168.220.247 port 5001
[ ID] Interval Transfer Bandwidth
[ 3] 0.0-10.0 sec 9.49 GBytes 8.15 Gbits/sec


vm 2:

root@database-backups:~# iperf -s
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 128 KByte (default)
------------------------------------------------------------
[ 4] local 192.168.220.247 port 5001 connected with 192.168.220.248 port 47042
[ ID] Interval Transfer Bandwidth
[ 4] 0.0-10.0 sec 9.52 GBytes 8.17 Gbits/sec
[ 4] local 192.168.220.247 port 5001 connected with 192.168.220.248 port 47050
[ 4] 0.0-10.0 sec 9.54 GBytes 8.18 Gbits/sec
[ 4] local 192.168.220.247 port 5001 connected with 192.168.220.248 port 47070
[ 4] 0.0-10.0 sec 9.49 GBytes 8.14 Gbits/sec

 

[Sean Ho]

Try with parallel streams, e.g., iperf -P 4. LACP doesn't magically double your bandwidth.

 

[NablaSquaredG]

Try with parallel streams, e.g., iperf -P 4. LACP doesn't magically double your bandwidth.

Even that likely won't work.

LACP is usually based on MAC (dest / src) or MAC (dest/src) + IP (dest/src). So if you increase the number of streams in iperf, you won't see more speed from LACP except if you have LACP configured for MAC+IP+Port hashing, which is non standard as far as I remember.

 

[Sean Ho]

Ah, great point; I assumed ESXi would default to a hash mode that includes source port.

 

[jorgefreyre]

on esxi:




on switch:






result on linux vm

 

[Sean Ho]

Looks like hashing on src/dst MAC only, hence same results on parallel iperf. What will be using the bandwidth? Consider multipath iSCSI perhaps?