SAP Power LPAR's SUSE Linux 11 SP4 network performance slow

Hi Team ,

We have SUSE LInux LPAR’s which all are running in the Power 822 ,These LPAR’s are hosted in the AIX VIO servers . As of now the " niping "values from SAP level is not good values , It shows around 360 msec. They want to reduce less than 100msec. Please let us know any kernel parameter or network parameters can be tuned to reduce “niping” values . Any parameter available to increase network parameters

bond1 Link encap:Ethernet HWaddr F2:30:9F:20:29:1F
inet addr:10.1.52.46 Bcast:10.1.55.255 Mask:255.255.252.0
inet6 addr: fe80::f030:9fff:fe20:291f/64 Scope:Link
UP BROADCAST RUNNING MASTER MULTICAST MTU:9000 Metric:1
RX packets:106570387 errors:0 dropped:51466640 overruns:0 frame:0
TX packets:9300417 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:11421148836 (10892.0 Mb) TX bytes:1625692948 (1550.3 Mb)

ifcfg-bond1 --configurations :

STARTMODE=‘auto’

or ‘onboot’

BOOTPROTO=‘static’
IPADDR=‘10.1.52.46/22’
BONDING_MASTER=‘yes’
BONDING_SLAVE_0=‘eth1’
BONDING_SLAVE_1=‘eth4’
GATEWAY=‘10.1.52.30’
#BONDING_MODULE_OPTS=‘mode=active-backup miimon=100’
BONDING_MODULE_OPTS=‘mode=active-backup arp_interval=2500 arp_ip_target=10.1.52.30 primary=eth1 arp_validate=all’
BROADCAST=’’
ETHTOOL_OPTIONS=’’
MTU=‘9000’
NAME=’’
NETWORK=’’
REMOTE_IPADDR=’’
USERCONTROL=‘no’
IPADDR_1=‘10.1.52.3’
NETMASK_1=‘255.255.252.0’
BROADCAST_1=‘10.1.55.255’

Current niping values .
qw0adm> more QW005406092018-162824.txt
av2;0.374;ms
av2;0.370;ms
av2;0.361;ms
av2;0.360;ms
av2;0.365;ms
av2;0.387;ms
av2;0.365;ms
av2;0.369;ms
av2;0.394;ms
av2;0.389;ms

Regards
Sebastian
91 9884430211

Hi Sebastian,

unfortunately, you’ve not posted that many details of your test. Where did you run niping, how does it connect (network- and “SAP Router”-wise) to the target instance? The LPARs you’re trying to optimize, are those running SAP applications or are these SAP routers?

As a first step, have you checked the IP-layer response time of the (client to SAP server) connection (or all individual connection legs, if crossing one or more SAP routers), via a traditional ICMP echo (AKA “ping”)? If you face high network-added delay because of i.e. small-bandwidth or high latency network connections, that’s what you should try to improve first.

Regards,
J