Question on figure 3-4 of SNASw Design and Implementation guide

[a.mountouris]

Hello,

on figure 3-4 on page 3-7 of the SNASw Design and Implementation guide there is an example code of SNASw port and links for Upstream Hosts.

There is one port statement and two link statements, one for each host.

However, assume I have one host with two OSA interfaces, can I use the same coding (two links, one for each OSA), as a redundancy mechanism of the OSAs? How can I choose the primary link?

Or if I do this configuration I will achieve also load-balancing? (both links active)

Regards, Apostolos.

[romney]

The figure you are pointing to is specific to a traditional hpr/appn (not hpr-ip) configuration, so the two links for redundancy are important.

hpr-ip is different in that the redandancy is in the ip network itself. On the snaswitch router we recommend that you use a loopback interface for configuring the hpr-ip port rather than tying it directly to the interface. Generally there will only be one OSA per host that is used for hpr-ip and only one snasw link to each host. However, with more current releases of z/OS you can have more than one group (and hence OSA) specified in your hpr-ip xca major node. If you do this you can achieve load balancing across the OSAs, and you could have each snaswitch router have multiple links to each host (but still only using the single snasw hpr-ip port based on the loopback interface).

- Ray

[a.mountouris]

OK about hpr/ip but what about traditional hpr/appn? In case of two OSAs in the same host, can I still use two links under the same port, to point at each OSA using traditional hpr/appn?

[romney · Accepted solution]

Usually when customers talk about OSA it is OSA Express which is purely IP traffic. EE works in that environment because it is SNA/HPR over IP. In order to have traditional appn links using OSA you have to run in native SNA mode which is much slower than OSA express and precludes IP traffic over those interfaces, and that is why very few customers do this. If those conditions are ok for you then yes, you can still use two snaswitch links under the same port. The snasw port interface must be on the same LAN as both OSA adapters, and you would specify a different mac address for each OSA.

- Ray

[a.mountouris]

Thanks Ray, this answer resolves the issue. However,

I would like to hear more about the mechanism in this case (2 links under 1 port).

Are both links used to forward traffic to both OSAs? If this is the case, based on which mechanism is this happening? (round robin? something else?)

Is there a way to define primary (active) and secondary (standby) link?

[romney · Accepted solution]

HPR is built on top of APPN with respect to path determination. In APPN the session path is determined by the Topology and Routing Services (TRS) component in the Network Node serving the device originating the session. TRS calculates this path by looking at the characteristics associated with each possible Transmission Group (TG a.k.a. link) that can be used for the session. In the case you are talking about where you have parallel TGs (two or more links) between two nodes, and unless you do some customizing of the TG profiles, they will have equal weight (desireability). This means that over time (and with enough sessions) you would see the load balanced over the two links. If you customize the TG profile (on snaswitch you use the tgp keyword), you can achieve the result of all sessions on a primary link and no sessions on a backup link until the primary link fails. - Ray