Sponsored Content
Operating Systems AIX PowerVM SEA etherchannel performance ? Post 303018506 by vilius on Friday 8th of June 2018 07:03:58 AM
Old 06-08-2018
PowerVM SEA etherchannel performance ?

Hello,

POWER7 machine.
4 x 1Gbit port ethernet adapter is dedicated to vios. 8023ad etherchannel is created using those 4 ports. Etherchannel adapter is shared to lpars using SEA.
If I test network performance directly from vios partition (using iperf) I'm geting nice 4Gbit throughput. But if I test from lpar partition (using iperf) I'm getting only about 1.5Gbit throughput (one time I saw 2Gbit, but never above 2Gbit).

Seems like SEA problem but where ?

Thanks,
Vilius M.
 

10 More Discussions You Might Find Interesting

1. AIX

Problem with Etherchannel

I'm trying to configure Etherchannel on AIX 5.3 and I'm having some problems. I am aggregating both of the onboard NIC's (ent0 and ent1) into a new adapter ent2 for failover reasons obviously. At first I was experiencing a problem where about 50% of all my packets would be dropped. Then after... (4 Replies)
Discussion started by: acascianelli
4 Replies

2. AIX

AIX / Etherchannel / VLANs

I have 1 AIX server, 4 dual ported fiber attached ethernet cards and 4 VLANS coming in. Is it possible to present those 8 ports as 1 IP address using etherchannel? Thanks. (5 Replies)
Discussion started by: jwholey
5 Replies

3. AIX

Can EtherChannel be a redundancy config?

I have one p5 server, no lpar, 2 physical network adapter, bundle as one EtherChannel. The two 2 physical network adapter connect to 2 network switches. How to make this as a redundancy config? Current is if one physical network adapter down, the seond one will go through the network. but if... (2 Replies)
Discussion started by: rainbow_bean
2 Replies

4. AIX

EtherChannel interface ping pelicularity

I have few AIX servers in my data center, which have their EtherChannel configured to have a fat pipe (active/active) and a nagios server to monitor the server uptimes. Nagios server pings every server in the data center and we have pretty much every known unix flavor in this data center. When... (3 Replies)
Discussion started by: thenomad
3 Replies

5. AIX

Aggregated Etherchannel

Hi, I have a AIX5.3 P6-520 with 8x1Gb Ethernet ports with possible connections to 2 CISCO 2960 switches. I can create 8023ad Etherchannels which work correctly with the switches (I have set up for LACP passive and a valid channels) But can create 2x 8023ad 4Gb Etherchannels with no backups.... (2 Replies)
Discussion started by: spitzie
2 Replies

6. AIX

PowerVM HA questions

hi guys, i need to develop following setup for a customer: high availability oracle database on aix7 and linux in 2 different LPAR using dual ps700 blande a ds3400 and HMC. the question is, it is possible to have automatic vioc failover(aix,linux) when for example 1 vios goes down? in... (6 Replies)
Discussion started by: gullio23
6 Replies

7. AIX

Need IBM PowerVM

Hi all please give me a link for IBM PowerVM (4 Replies)
Discussion started by: babinlonston
4 Replies

8. AIX

Etherchannel failover testing

Hi Guys, I am trying to test an etherchannel failover and switchback. I did failover from primary to backup adapter, but now I am not able to switch it back to primary. phcxxxxx:/ > entstat -d ent4 | grep Active Active channel: backup adapter ... (1 Reply)
Discussion started by: snchaudhari2
1 Replies

9. AIX

PowerVM/lpar creation problems

Hi, I'm a newbie to PowerVM and LPARs ( I work with VmWare and Oracle VM Server for Sparc) .recently we purchased an 8202-e4b server, with two enclosures for add'l disk capacity. it is licensed for PowerVM standard,came preloaded as such, AIX 7.1. I've set up HMC v8.2, and run it through a VM on... (30 Replies)
Discussion started by: computron
30 Replies

10. AIX

Powervm ivm vios

Hello, After installing on P6 which is POWERVM IVM VIOS enabled (VET CODE D21C77ACD9229817CA1F00002C10004164 ) i get this message "I/O hosting requires a hosting partition - boot not permitted". HMC was connected to the machine then HMC was removed through ASMI Searching on... (2 Replies)
Discussion started by: filosophizer
2 Replies
alt(7)							 Miscellaneous Information Manual						    alt(7)

NAME
alt - DEGPA Gigabit Ethernet interface SYNOPSIS
config_driver alt DESCRIPTION
The alt interface provides access to Gigabit Ethernet (1000Mbs) through the DEGPA device. The interface supports full-duplex operation in a switched or point-to-point configuration, and provides the following features: The interface has Link Autonegotiation enabled by default. Some switches do not support Link Autonegotiation. To turn Link Autonegotiation off, use the following command: # lan_config -ialt0 -a0 Note that you may add this command to the /etc/inet.local file to preserve the set- ting of Link Autonegotiation across system restarts. JUMBO packets are disabled by default. JUMBO packets provide a non-standard larger packet size. This enables the interface to carry more data with less CPU overhead. To enable JUMBO packets, use the following command: # ifconfig alt0 ipmtu 9000 Note that there are several interoperability issues with using JUMBO packets (for example, if your switch goes from 1000Mbps to a 100Mbps client, JUMBO packets will not work on a 100Mbps LAN). In order to use JUMBO frames, you will need a switch that supports JUMBO frames or a point-to-point configuration with a partner that supports JUMBO frames. Receive flow control is enabled. There is currently no way to turn this off. Gigabit Ethernet performance with TCP/IP depends on several factors. Some of the influencing factors are as follows: The speed at which data can be delivered to the interface influences throughput. If your CPU(s) are busy doing several tasks, the task using Gigabit Ethernet may not get enough run time to deliver packets. In general, faster CPUs will deliver better throughput. Fast access to the PCI bus is critical for high throughput. Using a 64-bit PCI slot will give you better performance and use less PCI resources than a 32-bit PCI slot. Putting the interface on the same PCI bus as other peripherals will degrade throughput. Each system type may also have different PCI-to- host speed considerations (the speed at which the PCI-to-host hardware allows the device to operate). The standard TCP/IP applications (for example, ftp and rcp) are not designed to run at Gigabit speeds. TCP applications that expect performance should use a message size of 65000 bytes and a window size of 128000 bytes. Even when an application is modified to use these settings, high throughput may not be attainable. This is particularly true when an application is waiting for data to send (data from a disk, for example). ERRORS
The following diagnostic and error messages contain relevant information provided by the alt interface, and are displayed to the console. Each message begins with the adapter identification, including the number of the adapter. The alt interface could not find adequate I/O addressing on this system to operate. This is a fatal error, and the DEGPA-SA will not operate on this system. There was a memory alloca- tion problem or the device initialization has failed. This is indicative of a hardware problem. Indicates that the Gigabit Ethernet link is up. The Autonegotiated keyword indicates that the link was autonegotiated. Note, this will only be on if autonegotiation is enabled). The ReceiveFlowControl keyword indicates that Receive Flow control is enabled on the link. Indicates that the link is no longer estab- lished. No communication will occur over the link while it is down. RELATED INFORMATION
Commands: ifconfig(8), lan_config(8) Files: inet.local(4) Network information: arp(7), inet(7), netintro(7) delim off alt(7)
All times are GMT -4. The time now is 04:58 PM.
Unix & Linux Forums Content Copyright 1993-2022. All Rights Reserved.
Privacy Policy