I don't understand HPE: they bought Aruba, already had a strong wired switches division. They also bought Cray, which obviously has their own advanced networking expertise. And now Juniper?
It is not like they lack in-house talent, they have it!
bc569a80a344f9c · 5h ago
They’ve been trying to compete in the controller-less WiFi market for a while and it’s not going well. Tons of larger (higher ed, which have complex environments with very many end points) campuses have tried to move to Aruba Central, and subsequently moved back because it isn’t working well.
And there’s huge operational benefits to Mist’s architecture even if you leave out the AI part. My best guess for why Aruba hasn’t been able to replicate at least the architecture is that they’re carrying around too much legacy they can’t break with to make this feasible.
Since they’re nearly explicitly saying “we are buying Juniper for Mist”, and at least in my opinion aren’t doing that only for further developing that portfolio towards the data center, I guess they’ve just decided to buy someone who was able to execute on that architecture shift.
wmf · 5h ago
Aruba probably declined after HPE bought it (just like every acquisition) so now they need another one to juice revenue.
Cray networking is HPC specific and isn't used in the campus or general data center markets.
tw04 · 3h ago
> already had a strong wired switches division
Huh? They had basically 0 market share outside of SMB in the switching space.
Cisco, Arista, Juniper, Force10 (Dell).
Foundry was the bottom dollar alternative before being acquired by brocade, which Broadcom eventually ruined. And mellanox in the HPC space.
I’ve literally never run across HP switches in any enterprise datacenter, including shops that were “all in” with HP.
It’s interesting that this requires HPE to divest from Instant On, which is their SMB brand. It’s unclear how that alone maintains a competitive landscape for enterprise customers that largely would have chosen Aruba’s more traditional offering.
That said, it’ll be interesting to see what use cases for data centers they come up with. Mist is one of very few infrastructure providers I’ve seen where AI/ML features actually improve operations. Surprisingly often in ways that I’m puzzled competitors have not been able to duplicate: something as simple as sending all access point logs through outlier detection and automated triage to proactively send RMAs for broken hardware before the customer even complains.
It is not like they lack in-house talent, they have it!
And there’s huge operational benefits to Mist’s architecture even if you leave out the AI part. My best guess for why Aruba hasn’t been able to replicate at least the architecture is that they’re carrying around too much legacy they can’t break with to make this feasible.
Since they’re nearly explicitly saying “we are buying Juniper for Mist”, and at least in my opinion aren’t doing that only for further developing that portfolio towards the data center, I guess they’ve just decided to buy someone who was able to execute on that architecture shift.
Cray networking is HPC specific and isn't used in the campus or general data center markets.
Huh? They had basically 0 market share outside of SMB in the switching space.
Cisco, Arista, Juniper, Force10 (Dell).
Foundry was the bottom dollar alternative before being acquired by brocade, which Broadcom eventually ruined. And mellanox in the HPC space.
I’ve literally never run across HP switches in any enterprise datacenter, including shops that were “all in” with HP.
Press release.
It’s interesting that this requires HPE to divest from Instant On, which is their SMB brand. It’s unclear how that alone maintains a competitive landscape for enterprise customers that largely would have chosen Aruba’s more traditional offering.
That said, it’ll be interesting to see what use cases for data centers they come up with. Mist is one of very few infrastructure providers I’ve seen where AI/ML features actually improve operations. Surprisingly often in ways that I’m puzzled competitors have not been able to duplicate: something as simple as sending all access point logs through outlier detection and automated triage to proactively send RMAs for broken hardware before the customer even complains.