So last night while working on a Scalable Compute and storage design for a client, this post popped up in my twitter stream from @ErinatHP;
New HP blog post “In the light of day the Cisco UCS hype doesnt match the promise” ; UCS not all its marketed to be http://bit.ly/dKj88W
So in my normal do not let a stupid dig by a lame duck player go unmatched I responded Oh I can’t wait to read this FUD (you can check me out on twitter @joshobrien77)
All the twitter marketing and pissing matches aside I meant what I said and I did look forward to reading the HP Spin on where their market is vanishing to. And here are my responses, while they might not be the most technical they are not un-informed from the basis of the Cisco UCS platform or the HP C7000 with FLEX-10 Platform. And remember at the end of the day I represent me not Cisco not my employer, just little old me.
Also just so if this gets nasty I want to make sure that I am crediting this correctly:
All of the HP Writes: Are direct Quotes from Duncan Campbel with HP on his blog which you can find here: http://h30507.www3.hp.com/t5/Converged-Infrastructure/In-the-light-of-day-the-Cisco-UCS-hype-doesn-t-match-the-promise/ba-p/83537
PLEASE READ ALL of Duncans Post BEFORE you READ Mine. I DO NOT PRETEND to REPRESENT HIS SIDE WELL AT ALL!
An incomplete solution
The article notes, Plans call for 90% of Cisco’s total IT load to be serviced by UCS within 12 to 18 months. So Cisco, by its own admission, has a 90 percent solution. What about the other 10 percent? Cisco is acknowledging it does not have a server platform for every workload in the data center. For instance, they do not have an integrated mission-critical solution, nor a Private Cloud solution that can be up and running in 30-days.
An incomplete evaluation of UCS:
I read the same Network World article. I was in a room full of server guys who are running HP C7000 series and IBM BladeCenters and I heard their responses as they each read it. So lets start with in 12 to 18 months Cisco will have 90% of their entire IT load on UCS. Holy crap! That is 51 Data Centers around the word going from 100% 3rd party to 10% 3rd party servers in under 2 freaking years.
So lets break down what those 10% 3rd party servers could be. Well for one like it or not Cisco has outsourced 100% of their pre-UCS servers to HP and IBM. So ACS, Call Manager, NAC and on and on. So not all of these systems have been certified to run Virtual. Those alone could account for some real part of that 10% But lets say its not. Last time I checked Cisco is a pretty big software dev shop. For years I called BS on Cisco stating they were not a hardware company but a software one. Well those days are done because that is pretty much what they do. Dev environments can require all sorts of oddities including OSX, Solaris, UNIX, LINUX not even counting their products like EOS. Some of those systems simply do not run on standard X86 hardware. Sucks but its the facts. Then on top of that they say total IT load to be serviced by UCS. That could also account for appliances doing systems management and services such as DNS and IPAM. Who knows but I hardly think the remaining 10% shows a lack of capability of the UCS platform.
So not lets deal with the Integrated mission-critical solution, nor a Private Cloud solution crap. First off from first hand experience you can spin up a Pair of UCS fabrics and a single loaded chassis in 8 hours from scratch. From there you simply plug in power and Ethernet to each subsequent chassis and your off and running. I have seen on multiple occasions it take the better part of 2 days for an experienced team to provision a C7000 new or after a significant profile change that led to having to re-build the unit. And let me just say I am not sure where HP has been or what they have been doing for the past 18 to 24 months but UCS is built from the ground up as the perfect platform for Private Cloud or should I say Enterprise Virtualized Data Centers. And as long as the platform required is x86 architecture, which most current and emerging systems are then Mission Critical is not really a challenge. Utilizing a test A/B Fabric design from our old friend Fibre Channel Storage UCS can take a beating and keep on going.
More of the same
The article says the current anchor site for Ciscos grand IT plan is a relatively new facility in the Dallas area that is already outfitted with 1,400 UCS blades, 1,200 of which are in production, and 800 legacy HP blades.
Lets think about this. Cisco announced UCS in March 2009. Its now almost 2011, so this means Cisco hasnt been able to migrate many of its workloads to UCS in all the time that has passed since the UCS announcement. Its still relying heavily on the HP BladeSystem. (Well take that as a compliment.). But wheres the agility, simplicity, and time-savings?
More Blah Blah Blah from HP
Anyone paying attention prior to the March 2009 launch of UCS would have listened to the Cisco DC podcast and kept up with the info flowing out of Cisco IT, both of which were amazingly open and forthright about the challenges of running the Cisco IT infrastructure as well as adoption of new Cisco Tech and some really interesting statements that showed the roadmap of the Cisco DC Architecture some of which are still just now coming out officially.
What I took away from all that pre-launch info was;
Hey look we love our new tech. We cant even talk about it in detail yet but its going to be crazy new servers and storage. But we wont be running 100% of it day one we have a business to keep online here. Plus we have to find budget for all of this just like everyone else, yes it may be cheaper for Cisco Internal but there is still a cost.
Your darn right there is a cost. Cisco still has to produce and pay people like Intel to build UCS. They have to at least break even on each sale internally to move forward. Not to mention even if there was a zero sum cost for Cisco IT on hardware there are HUGE investments in Legacy Server systems like HP and IBM. As far as HP taking that as a compliment that there are still legacy BladeSystems inside Cisco good for them. It is great to know that they take pride in having their shipping platforms classified as Legacy and that as this article clearly shows they agree. Well so do I. Island based servers systems with disconnected onboard management systems with complex non-standard communications systems are the past. While we may be moving back to a MainFrame type world of centralized compute, we are done with single vendors and crap designs dictating the future of innovation. Bravo to Cisco for reaching past in stone memory limits and crap management platforms to make servers scalable and future ready.
In the article they also take a shot at Cisco based on the fact that Within five months, HP announced that all six of its internal next-generation data centers were running company operations worldwide on HP Networking solutions, and were Cisco-free for core WAN routing and switching Well again I say good for HP. You just did what any half decent Data Center Network team could have done. Back in the day I was part of a 5 person team that deployed a state wide WAN supporting 500 Edge sites, 200 Servers and 2000 LAN users in less than about 8 months from hardware landing to done. But god help me if those 200 servers we were supporting did not take the better part of 5 years to cycle out. Migrating network hardware and migrating servers and their legacy non-virtualized work loads are simply not apples to apples. Instead it is more like Apples and a pissed of pack of T-REX. So hype away but but be honest with yourself and the rest of the world.
A networking story that doesnt add up
The article notes that each UCS chassis in a cluster is connected to a top-of-rack access switch. From that switch, storage traffic is sent over a 16Gbps connection to a Cisco MDS SAN switch, while network traffic is forwarded via a 40Gbps LAN connection to a Cisco Nexus 7000 switch. In the future, it will be possible to use FCoE to carry integrated storage/LAN traffic to the Nexus and just hang the storage off of that device. The fact is HP Networking is taking bigger bites out of Ciscos business creating a semi-panicky position for them.
Wait a minute. Hasnt Cisco been telling customers that all the standards and products were ready to support end-to-end FCoE today? What gives?
HPs networking story has never added up, so why not say the same about Cisco?
Really the mother of all lame network stories wants to step up and say that Cisco cant get the networking part of this straight. This from the company that I took a risk on 4 years ago only to find that their free support model and switches that have a better OS so it requires less ram actually equaled cant call us except between 8 and 5 EST M-F and yeah you cant enable 75% of the switches features while doing multiple VLANs cause we didnt use enough ram. Wow that is rich. But I digress .
Today Cisco is leading the charge with FCoE. It was about time we did something. We were really starting to bottleneck at storage we all know it. HP should know it better than anyone considering their plans for the EVAs were just throw another full frame at it if you need more IO and bandwidth. Funny how that has played out for HP zealots and their rooms full of EVA frames. So go ahead HP keep telling the scalability and we have Cisco on the ropes story.
Oh yeah lets talk about The fact is HP Networking is taking bigger bites out of Ciscos business creating a semi-panicky position for them. So ProCurve died in the womb. Actually your own Services group killed it because they liked selling Cisco better than your own in house product. But thats water (filled with ProCurve Blood and chunks) under the bridge because now you have H3C/Pro?/3COM or whatever the whole mess is. Here is what I know. Take Cisco out of the picture and just look at Juniper, Arista and Force10 and you have players that 90% or more of the DC Engineers will trust before they put in a product branded in any way with HP on it. And simply put your Flex-10 products just drive home that issue with ongoing issues with Host based driver failover issues and proprietary Rate Limiting nonsense.
Around end-to-end FCoE there are smarter people than me to talk about this. But yeah yeah the Nexus 7000 did not ship with FCoE or FEX support or a keggerator but guess what it did ship and out sell HP DC switching. And while I am not thrilled with the whole storage protocol war I think FC is crossing through its prime and we will see iSCSI, NFS, FCoE and who knows what else step up and replace it. What Cisco has not done in any way shape for form is isolate legacy FC installations. Instead they have the only real platform to allow customers to start testing the waters of FCoE or dive straight in and convert.
The final thing I want to comment on and if your still reading this your muttering thank god, is the statement Duncan makes concerning his blog.
However, I never consider my blogs as a marketing platform, rather a candid information exchange.
I have no idea who Duncan is and I am sure he is a great guy but he is either saying this tongue in cheek or is really mis-guided about what his job is and where he is posting this. This entire post is a hack job based on an article and some wild assumptions about material in that article that is twisted to reflect not industry issues but why HP is so much better than Cisco. But keep in mind I never consider my blogs as a marketing platform. That and the lead out of And if youd like, I invite you to dig deeper into the HP BladeSystem, the BladeSystem Virtual Connect portfolio, or the complete Converged Infrastructure story and architectural framework.
Well Duncan your blog is a Marketing platform. You do work for a Legacy Server and Storage Provider. That does not bother me at all. Just own up to the fact of what you are doing. At the end of the day our word should be our bond no matter if that is what we write on a blog, say over the phone or tell to a customers face.
To HP good luck and good hunting. Because myself and others like me are coming hard and we are not going to hesitate to toss your legacy crap right out of the Data Centers we are working in and replace it with scalable future technology. If and when you catch up things could get really exciting.
One of the most amusing parts of the HP stance was about the “90% solution”. As you point out, refreshing even 90% of a huge enterprise’s compute in 2 years is a huge accomplishment.
Even funnier than that, is that I have personally installed hundreds of thousands of dollars worth of IBM (AIX, OS/400) server equipment into a *brand new* HP data center to support one of their legacy business applications. Why? They couldn’t migrate it to one of their platforms and had to continue running it on IBM equipment.
Guess what? No vendor has a 100% solution. There will always be niche cases where you go to a specialty provider. Sure, HP and IBM *claim* to have 100% solutions across any workload, but go into their datacenters and you’ll find competitors gear all over the place. I don’t see that as a problem, or a weakness, just recognition of the fact that sometimes you buy a specialty product to meet a business need – even if you compete in most spaces with that supplier. It only gets amusing when the same vendor claims 100% solutions.
Let’s drop the FUD and just concentrate on the things you CAN do, alright? :)