The reasons Alan lists are the reasons we went LACP. With Guest Lan, we were lighting up 4 osa ports. We were concerned about sending all the traffic over 1. And OSA failures (and they do fail) result in a pause to switch over to the backup vswitch port. We have an app that doesn't like tiny dramatic pauses (lots of folks get paged). Our next step is VSS in the cisco switch to eliminate switch failures from causing the pause and the switch to the 1 (backup) interface.
We explored using channel bonding briefly to have a guest's traffic over 2 interfaces. We are using channel bonding for an app that can't use vswitch, but all the other servers do use the lacp vswitch. But it won't solve your problem Pieter it sounds like. Sounds more like a capacity issue. Marcy "This message may contain confidential and/or privileged information. If you are not the addressee or authorized to receive this for the addressee, you must not use, copy, disclose, or take any action based on this message or any information herein. If you have received this message in error, please advise the sender immediately by reply e-mail and delete this message. Thank you for your cooperation." -----Original Message----- From: Linux on 390 Port [mailto:[email protected]] On Behalf Of Alan Altmark Sent: Thursday, January 22, 2009 3:52 PM To: [email protected] Subject: Re: [LINUX-390] bonding multiple qeth vnics to vswitch? On Thursday, 01/22/2009 at 05:05 EST, "Harder, Pieter" <[email protected]> wrote: > You just lost me. LACP is on the trunk side of the VSWITCH. I am > talking > bonding on the port side of it. What exactly did you think of doing > that was > solved by LACP? Talk about a confusing conversation! For those who left their scorecard at home today, LACP (Link Aggregation Control Protocol) support in the VSWITCH is directed only at the physical switch. The guest-facing side of the VSWITCH does not support LACP. (Let's avoid terms like "port side" and "trunk side". When a guest is using a virtual trunk, which side is the "trunk side"?) LACP is the thing that lets the host and switch coordinate traffic. The only use of LACP that I've seen/heard about so far is for better error toleration and recovery, better physical OSA management (dynamically delete from group so you can put on microcode fixes, then re-add), and to satisfy Management demand to "Stop wasting OSAs! Why aren't both LEDs flickering?!? I'm not paying for LEDs that don't flicker!" (FYI, IEEE 802.3ad spec lists 9 benefits of which increased bandwidth is only one.) While you can have multiple vNICs on an aggregating VSWITCH, there's nothing that will spray/deal queued frames across the vNICs since they each have a unique MAC address. But even if it did, you're doomed if the buffer fill rate is higher than the consumption rate. All you've done is delay slightly the point at which all guest buffers are full. Alan Altmark z/VM Development IBM Endicott ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
