By John P. Gray, HP Networking
Having worked in the service provider space for many years, I’ve always had a great deal of respect for Juniper's technical networking chops.
Juniper's recent "Stratus" announcement initially reinforced those feelings as I read about its latest endeavors around "Qfabric" which is designed to help customers build a flatter, lower latency, highly scalable data center network.
The point of this post is NOT to debate or refute Juniper scaling or performance claims. And I will only briefly mention the fact that HP also currently designs and builds very high performing/scaling (flat) two-tier data center networks with switch products like our core A12500, our top-of-rack A5820 10GbE switch and Virtual Connect Flex Fabric Module for BladeSystem customers.
At HP, our network products are only one aspect of our Converged Infrastructure (CI) vision for the data center. To address real change in the enterprise, you need to understand networking, compute, storage, management and power/cooling – and have a deep understanding of applications, how they are being used and how best to integrate all of these moving parts. HP does.
Network scaling and performance are table-stake requirements. But at what cost?
What stuck me was a public comment made by the Juniper CTO who was trying to make the case that because Juniper "saved the Internet" back in the 90s (debatable) due to massive bandwidth demands, so now Juniper Stratus/Qfabric can also "save the data center" from imploding upon itself for the same reasons.
No argument here on the bandwidth front. The more the better … 10, 40, 100 GbE. Bring it on.
But the reason the Internet is so wildly successful isn't just because it just got FASTER …it's because it's OPEN.
Juniper has seemingly taken a page out of the Cisco playbook by introducing a closed and propriety network architecture. (Not to mentioned very expensive one). Stratus/Qfabric was not designed to interwork with a company's current data center deployment (or even Juniper’s current switches for that matter).
Moreover, Juniper was mum on any details surrounding how it might deal with the increasingly complex virtualized server-to-network edge (in terms of visibility, control and management) – an area where HP is leading the charge from an industry-standards perspective. Check out this podcast with HP’s Chief Technologist Paul Congdon or this document to learn more about VEPA and other important technology and standards in this space.
The point is this: Customers have grown to expect and demand increasing levels of openness and industry-standards (especially in the data center with x86 servers, storage, virtualization and networking). Not less.
I'm a bit surprised by how a vendor (that is not even a networking or data center incumbent) would think customers would be willing to risk embracing a closed (network only) point solution from a vendor with limited data center expertise?
What’s your opinion? I’m very open to continuing this discussion.