On 23 Aug 2018, at 15:48, Jeffrey D. Sessler wrote:
It’s great to hear Aruba is adding features such as “automated RF management” that Cisco has had for over a decade.
My understanding of the “automated RF management” part is directly related to the upgrade process (and not DCA/TPC, as you’re suggesting, which Aruba has had for some time).
It splits the APs automatically into groups based on their channel assignment (since, given similar approach as DCA, this gives a rough estimation on “APs that are overlapping each other” — could also be improved in the future using signal strength an AP sees other APs). It then moves clients off of one of those groups (making them join other, adjacent APs), reloading those clientless APs into the new software version, and then moves clients back when it moves onto the next “channel group”. Cleanse and repeat until all groups are done, giving you “zero downtime”.
This is at least how it was last time I read about it, and is by far superior to the way Cisco does it (where you manually have to fiddle with groups within Prime — and that’s without talking about Prime itself…).
The Cisco-solution also requires a separate controller to do this, whilst Aruba uses it’s redundant controller by automatically handling “splitting” the HA-pair (by upgrading one of them, moving the APs according to the “channel groups”, and then finally upgrading the last controller).
The “equivalent” with Cisco would be to split your HA pair manually, move all APs to one of them, upgrade the other, move them using the rolling-AP-group-thingie in Prime, then upgrade the last, and finally join them back as a HA, causing significantly more downtime than a normal Cisco upgrade process. Or you could buy a completely separate WLC to achieve this, but that’s somewhat a waste of money if you already do HA/SSO (and buy WLCs in pairs).
In all seriousness,. if you’re talking specifically about AP updates, cisco has had AP code pre-download for years, resulting in between 2 to 4 minutes downtime when rebooting a multi-thousand AP controller. Not hitless, but low impact for sure.
I’ve never managed to do less than ~400 seconds on HA/SSO-enabled 8540s with 3k+ APs. That’s “a lot of time” many places (maybe not edu, but for sure in healthcare or other mission-critical businesses), which would be reduced to whatever time it takes for a client to re-associate after being “kicked” off the network (so time depends on the client, but would probably be sub-1s in many cases).
-- Joachim ********** Participation and subscription information for this EDUCAUSE Constituent Group discussion list can be found at http://www.educause.edu/discuss.
