[openib-general][PATCH 1 of 3] repost: Client Reregister support for kernel space

Eitan Zahavi eitan at mellanox.co.il
Wed May 31 10:03:57 PDT 2006


Hi Sean,

Well, this is a very old topic well discussed years ago. All credits to
Ashok Raj which you know better then me. The argument against the idea
for the SM to be the keeper of these registrations goes as follows:

Our problem has to do with data that is not persistently stored in the
fabric (which is: Multicast MGID->MLID mapping, Service-Records and
InformInfo records). 

Without the use of client re-registration the only way to keep this data
consistent between the master SM and its standby slaves (or across SM
restarts) is by using a transaction safe model. 

Transaction safe models are well known and require distributed handshake
or journaling systems (in our case distributed ones). Anyway what this
means is that every transaction (like creation of new Service Record or
registering to a multicast group) will have to be first committed and
approved by all standby SMs before it is responded.

This strict transaction safe model is not fitting very well with the
requirement for scalability of the fabric - which is hard to make even
without that complication. 

The Client-Re-Registration concept resolves this need as the clients
need to track their registrations and repeat them with the new master
SM. 


Eitan Zahavi
Senior Engineering Director, Software Architect
Mellanox Technologies LTD
Tel:+972-4-9097208
Fax:+972-4-9593245
P.O. Box 586 Yokneam 20692 ISRAEL


> -----Original Message-----
> From: openib-general-bounces at openib.org [mailto:openib-general-
> bounces at openib.org] On Behalf Of Sean Hefty
> Sent: Wednesday, May 31, 2006 7:35 PM
> To: Leonid Arsh
> Cc: Roland Dreier; openib-general at openib.org
> Subject: Re: [openib-general][PATCH 1 of 3] repost: Client Reregister
support for
> kernel space
> 
> Leonid Arsh wrote:
> >  the most urgent and critical case is
> >  the SM failure/restart when the SM is not connected to the host
directly.
> 
> Why not patch the SM to handle this sort of case and rebuild its
database
> without every client in the fabric needing to send it MADs?  Why can't
the SM
> save/restore the configuration itself?
> 
> - Sean
> _______________________________________________
> openib-general mailing list
> openib-general at openib.org
> http://openib.org/mailman/listinfo/openib-general
> 
> To unsubscribe, please visit
http://openib.org/mailman/listinfo/openib-general



More information about the general mailing list