Hewlett-Packard has tweaked the Comware operating system at the heart of its switches to make them more amenable to the clouds and to implement what is being called software-defined networking (SDN). Communication between computers is now more fluid than ever as servers and storage are increasingly virtualised and made more …
Apples** and Oranges all over again
Brilliant bit of marketing, with the network test showing "significantly improved throughput". Of course, the comparison didn't address other, similar configurations (such as Cisco's VSS or Stackwise), instead focusing on the path-limiting features also known as STP and VRRP. It's really disingenuous to make such a comparison, given that STP intends to eliminate loops (i.e. single, shortest network path) and VRRP has no relavance in a VMotion scenario (operates at L2, not L3). I'm sure this demonstrates the hard work at HP to innovate, which is certainly the nicest thing I can say about it without using bollocks, dingus, asshole, fucktard, and many other assorted dirty words.
**Thankfully, "Apples" has not been copyrighted, trademark or (spiritual figment of your imagination forbid) patented by a certain fruity company in that oh-so fruity place called California.
I don't use either HP or Cisco myself but a couple Q's on the Cisco stuff you mention. Stackwise - from what I can see is just basic stacking for rackmount switches and does not apply to chassis systems? The main distinction is the implied ability to have a lot more bandwidth between a pair of chassis switches than a stack of 1U switches over long range ethernet.
My own stackable 10GbE switches support 160Gbps (2x40Gbps full duplex) stacking for close range, though if I wanted long range (up to 40km) it drops to 2x10Gbps if I wanted to use stacking, otherwise could use other protocols with more links/higher bandwidth.
Also from what I see the VSS technology from Cisco seems to apply to the now incredibly ancient Cisco 6500 ? I don't see mention of support on Nexus or other newer tech from them? Also VSS seems limited to two switches whereas the HP can go to 8 according to the article.
But I do agree yeah I haven't built a network that used STP in literally almost a decade, and would not use it for anything!
No STP, eh?
Pray tell, how is it you design large scale, highly available and redundant ethernet infrastructures without STP these days?
I agree STP is the bane of any networkers existence, but a thorough understanding of it has literally built my career. Can't wait to start using some sort of SDN or TRILL or something else, though. (too bad my clients are all cheap bastards)
Re: No STP, eh?
Re: Hey Cap'n
I don't mention Nexus because it doesn't offer anything similar. VSS and Stackwise are the only options in Cisco's camp at this point. Neither offer the extensibility of IRF, but it's just dishonest to leave them out of the comparison. Almost every vendor has a similar stacking or clustering technology, but none of them made it into the comparison.
If you're not using STP, I assume you're going pure L3? That's fine until some server goob just HAS to have VMotion or Exchange clustering, etc. There's enough applications that require L2 adjacency that it's really hard to isolate L2 to a single switch in enterprise or service provider environments.
would be the Cisco method of connecting distributed datacentres.
Also i thought that vMotion was a bit latency sensitive with <~10ms being the limit.
OTV is one way in Cisco's world, but that's really just a way to create a geocluster, where two physical sites (or three or four, etc) look like one logical site to the rest of the world. I don't believe Cisco would recommend it to interconnect sites between the Americas and India, for example. And especially for the reason you mention. Applications, like VMotion, that require LAN connectivity (sub-10ms, high bandwidth) cannot really operate in a widely geographically-distributed environment. The limit is generally within a single local, at 40km or less I would expect (stupid speed of light limitation and all).