Controllerless Networks

In a MESH topology, there are two MESH Portals and four MESH points. The Portals are on the same subnet/network so that the Points only connect to one of them at any given time. The connections stats (RSSI and whatnot) are very good.

My questions is this, when one Portal goes down and the MESH Points then transition to the other Portal, what is that transition time? Is it in the Seconds? Is it in the Milliseconds? All APs in the cluster should be fully aware of both Portals and have routes for them so I would think it would be milliseconds but I have not been able to find any documentation supporting/denying this.

Any MESH Gurus out there have an idea?

Cheers!

.

It could take awhile based on how long it takes for the point to scan for the possible channel that the other portal is on and the quality of the mesh signal.    It should not be relied on for "hitless redundancy".

Thanks CJ, that's what I figured but just needed another opinion.

Mesh back-haul fail-over in the industrial world has some pretty steep requirements in regards to delay and latency. Looking like Aruba is not going to fit the bill on this one. Can't win 'em all I guess :-)

Aruba's AP mesh solution is not marketed or designed to have fast failover between mesh components.  It is primarily for fixed point to point networks.  Fast failover would have to be managed in a different manner or solution.  For example the HPE 501 bridge supports fast roaming between access points, so that is specifically designed and marketed for fast roaming.  https://support.hpe.com/hpesc/public/docDisplay?docId=emr_na-c04037651

I am curious though, in a small deployment of say two mesh portals and four mesh points, could you use a single channel architecture for faster fail-over? Would the portals keep track of the routing or would it cause a loop?

Well, that is if you can, I have never tried to manually set mesh channels, I usually just let them do their thing.

.

The datapath route-cache keeps track of what device can be located on what tunnel.  I don't know if the failover is faster when you set the same channel, but it should not be counted on for fast failover, really.

---

@cjoseph wrote:

The datapath route-cache keeps track of what device can be located on what tunnel.  I don't know if the failover is faster when you set the same channel, but it should not be counted on for fast failover, really.

---

Understood as far as counting on it, just more curious now than anything. I am not finding anything out in the universe that really does do this fast so it appears I may be chasing a Unicorn... again. But, if using an SCA can take fail-over from say 5 minutes to say 3 minutes, that's a significant improvement.

Aruba and a few others are my comfort zones so I don't like meandering away from them if possible. So if that would prove the be the quicker (on average) solution, it could still be applicable. make sense?

So don't put portals on the same channel, failover is not meant to be common so better to have cleaner channels than CCI within the mesh cluster.

Mesh point failover from one portal to a backup can take UP TO 40sec, it depends on scanning, channels in the reg domain, etc. 

If the need is a more mobile mesh solution, aka points on vehicles moving around, that is not supported. We have a few other options (501, instant with mobile mesh, etc) but in terms of ArubaOS, it's not supported.

If you want you can share your use-case, but generally Aruba's mesh is meant to be for fixed infrastructure.

Thanks for the input JH.

I am just trying to figure what the limits really are for these radios. I constantly have opportunities to get creative to solve various industrial connectivity issues by using wireless. The guys who live in the OT space always want as much redundancy as possible. IT doesn't have to be perfect, but it just has to work. I have found that the IAP270 series by far fails over to alternate portals faster than other manufacturers of similar mesh products (like Fluidmesh for example, whose engineer told me it can take up to five minutes). I am betting they will work great for the deployment that spurred this whole thread.

I am a fan of Aruba outdoor radios as they have proven to be dependable and durable. 

.

It's all a matter of what their requirements and tolerances are. 40 sec is fine for an infrastructure failover, but some customers require much less time. So long as the customer is aware and understands the mechanics, then it's fine.