Thursday, March 11, 2010

Blessay:Cisco UCS Really Is Just Blade Servers With Fancy NICs & A Response to HP Network Ambition

April 2, 2009 by Greg Ferro · 7 Comments 

I was review­ing some doc­u­ment­a­tion the Cisco web site for an archi­tec­tural white paper and came across this image in some mar­ket­ing material

What this image clearly shows is that the UCS strategy is really about get­ting a fancy NIC into a server.

ucs-its-a-nic-really-1.jpg

Each blade has a fab­ric con­nec­tion card to the UCS 6100 (fab­ric inter­con­nect), which then con­nects to a Nexus 5000 (the fab­ric master).  

Why does this matter ?

Simple. Lets con­sider HP and their C-​​Class chassis. HP has their “Virtual Connect” strategy for blade serv­ers with the NonStop fab­ric that con­nect all the serv­ers via the mezzan­ine cards to Interconnect mod­ules. Interconnect mod­ules sup­port Fibrechannel, Infiniband and Ethernet and so on. This means that HP can con­trol all the choices that the cus­tomer can make for the serv­ers because they “own” the design of the blade server chassis.

When bladeserv­ers were first released, IBM and HP had Cisco develop mod­ules that could be installed. More recently though, IBM has stopped even offer­ing Cisco switches for large parts of its port­fo­lio and HP is quite hos­tile — in the “HP Virtual Connect for Cisco Network Administrators” they refer to the Cisco switches as “tra­di­tional approach to admin­is­ter­ing the net­work”. Thats got to smart a bit.

Want to bet that HP didn’t want to play ?

Why wouldn’t HP want Cisco to provide the net­work­ing ? My guess is that HP has net­work­ing under devel­op­ment and they are freez­ing Cisco out of the data centre. HP has made sub­stan­tial inroads at the edge of the net­work with their ProCurve products. The Data Centre net­work is an entirely logical next step. Its incre­mental, extends their server expert­ise /​ port­fo­lio and integ­ra­tion and can eas­ily be sold to exist­ing cus­tom­ers through the server teams.

So HP might have shut Cisco out?

So Cisco does servers ?

I think HP did so. HP used the Tandem acquis­i­tion to put the NonStop server fab­ric into their blade serv­ers, then built net­work­ing mod­ules that posi­tioned them as com­pet­it­ors to Cisco. This was prob­ably some time in 2007.

Cisco response has been to build a net­work fab­ric, func­tion­ally sim­ilar to the NonStop server fab­ric, but with much deeper integ­ra­tion into the net­work offer­ing much greater vir­tu­al­isa­tion of the stor­age and net­work lay­ers. This is con­sist­ent with their cur­rent mar­ket­ing mes­sage around their net­work his­tory (although, with video, voice and home products who knows what Cisco is really about these days).

And its all about the message

ucs-its-a-nic-really-2.jpg

This gives Cisco a “mes­sage” that their product is “dif­fer­ent” from HP or IBM. A dif­fer­en­ti­ator like this can be used to ini­ti­ate new oppor­tun­it­ies with cus­tom­ers — to make that into English, you can approach the CIO /​ CTO and try and flog your new strategy. They had to make some new tech­no­lo­gies such as FCoE and all its sup­port­ing tech­no­logy, but that’s just a few broken eggs.

ucs-its-a-nic-really-4.jpg

And they had to build the blade serv­ers. Obviously. If Cisco didn’t make their own serv­ers no-​​one else was going to put Cisco net­work­ing in. Easy decision from that per­spect­ive, and John Chambers is an ambi­tious man. Growing the busi­ness in the data centre by extend­ing into the server farm makes logical sense. Remember, HP is likely to extend from the server into the net­work, so meet­ing them in the mar­ket is nat­ural. Especially with the hype about Cloud Computing devel­op­ing over the last two years.

But the actual serv­ers are not dif­fer­ent from any­thing else in the mar­ket. They are a little bit ahead of the curve, but HP/​IBM will there with sim­ilar products in a few months.

Its all about the network.

So, where is the Special Sauce ?

The HP Virtual Connect sys­tem does Ethernet, Fibrechannel and Infiniband as sep­ar­ate ele­ments or mezzan­ine cards.

Cisco uses a single Ethernet fab­ric to do all of this. By cre­at­ing FCoE from noth­ing, Cisco can com­bine all stor­age and net­work access over a single Layer 1 con­nec­tion. To tick off the marketecture:

  • no fibre­chan­nel = less com­pon­ents = less power & cost
  • uni­fied fab­ric = less con­fig­ur­a­tion = less OpEx
  • uni­fied fab­ric = less equip­ment = less CapEx
  • uni­fied fab­ric = more flex­ib­il­ity = ser­vice auto­ma­tion = cloud com­put­ing (BINGO!)

You get the idea, I am sure.

HP and Cisco — Frenemies ?

I wrote some­thing about this here

Conclusion

As a designer, this little graphic made UCS much more com­pre­hens­ible. The mar­ket­ing mes­sage that UCS is some sort of revolu­tion has obscured my think­ing. I was look­ing for rocket sci­ence, and all I got was mar­ket­ing smoke. And some bendy mirrors.

But I wish they would use nice smelling smoke next time. This one still smells a bit fishy.

Please rate this post:

  Why Rate Posts?
1 Star - It\\\'s Crud2 Stars - It\\\'s Tosh3 Stars - Something\\\'s missing4 Stars - Needs works5 Stars - Good Enough6 Stars - Good7 Stars - Excellent8 Stars - Brilliant9 Stars - Astonishing10 Stars - Awesomely Godlike? (1 votes, average: 7.00 out of 10)
Loading ... Loading ...

Comments

7 Responses to “Blessay:Cisco UCS Really Is Just Blade Servers With Fancy NICs & A Response to HP Network Ambition”
  1. Tony says:

    You’re an idiot. Go back and read the design docs instead of mak­ing a mor­onic state­ment after read­ing a one page press sum­mary. Jags like you should keep on doing it the old way and when what little cred­ib­ilty you have is in the cr-​​pper we wel­come you to join the rest of the engin­eer­ing crowd

    • Greg Ferro says:

      Mmmm, doesn’t leave much room for dis­cus­sion. Indeed I have read the design docs, and reviewed much of the mar­ket­ing mater­ial avail­able. However, I am not a true believer in FCoE. In many aspects I am a Ciscotist, but the Data Centre strategy built around FCoE is tran­si­ent and prob­ably not a happy long term strategy.

      I can cer­tainly see the value of the Network Fabric concept espoused in UCS, but cur­rently think it should be based on IP, not Ethernet. Intelligent com­ment welcomed!

      • Nick says:

        I don’t agree with your thought that FCoE should be based on IP. I think Cisco have it cor­rect, make it as simple as pos­sible. IP is required for cross­ing net­works and get­ting from end to end. However, FCoE is tar­geted at SANs and there­fore will always be switched within a network.

        For your reg­u­lar net­work com­mu­nic­a­tion, the fab­ric will sup­port tra­di­tional TCP/​IP traffic. Also, there is noth­ing to indic­ate that FCoE is going to be tran­si­ent. It will become a stand­ard­ized pro­tocol and 10GBoC will proliferate.

        • Greg Ferro says:

          I think the FibreChannel is a pile of crud, and put­ting it over eth­er­net is double crud. The idea of intel­li­gent net­work fab­rics (which is what fibre­chan­nel does) that some­how morph or munge the data flow has never been suc­cess­ful and it’s unlikely to start now.

          The rise of iSCSI and is NAS imple­ment­a­tions clearly shows that FC/​FCoE does not have a long term future and acts only as a trans­ition mech­an­ism for leg­acy FC-​​attached storage

  2. Omar Sultan says:

    Greg:

    So, I think the other dif­fer­en­ti­ator for the UCS is the abil­ity to man­age it more hol­ist­ic­ally – more com­pre­hens­ively and in a way that is more vir­tu­al­iz­a­tion aware than cur­rently pos­sible with other solu­tions. I know there is not a whole lot of detail about UCSM right now, but I did a couple of recent blog posts on this and I prom­ise more detail is shortly forthcoming.

    On a totally dif­fer­ent topic, being the “offi­cial FCoE stand­ard bearer” :) , I know we have agreed to dis­agree on the FCoE front, but I am curi­ous – what would you need to see in the industry (bey­ond final FCoE and DCB stand­ards) or our approach to change your mind about FCoE hav­ing some legs.

    Regards,

    Omar Sultan
    Cisco

  3. Rodos says:

    Well as Omar has respon­ded not sure what point there is hav­ing little ole me writ­ing some­thing but what the heck.

    I think you have missed quite a bit of the details around what UCS does as well as how it com­pares to the com­pet­it­ors. To com­pare them both you are going to have to go a bit deeper than you have gone. However I under­stand that often in a blog post this is not the purpose.

    UCS is a lot more than a fancy NIC, its the far bey­ond that with abstrac­tion at a large scale. You can do some of this with Virtual Connect but you need extra sauce to do it out­side of chassis.

    UCS breaks the mold in quite a few other areas, I would sug­gest you have bit more of a dig. If you are keen, grab a copy of the California book.

    Having played with the kit for the last few weeks I can attest that its a lot more than smoke, and it has not caught on fire yet.

    Cheers

    Rodos

    • Greg Ferro says:

      Yep. Read it, reviewed it (some­where on this site), thought about it. Haven’t changed my mind.

      There is noth­ing revolu­tion­ary about UCS, more like a very small step in evol­u­tion for vir­tu­al­iz­a­tion. I do not miss that Cisco needed to do some­thing unusual to pen­et­rate an (oth­er­wise) closed mar­ket, spe­cific­ally, the mar­ket­ing needed to have a nar­rat­ive that could tran­scend the noise from IBM /​HP /​Dell, and this is the res­ult. But doing some fancy sil­icon tricks and closed form pack­aging is not a benefit.

      I retain an open mind. Today, the sys­tem is unus­able except for niche use cases, and the wider mar­ket has no interest in cloud computing.

      A a res­ult, UCS boils down to server with a fancy NIC. There is no change to Windows/​Linux, no change to Intel archi­tec­ture, no change to any­thing, but a bit of sil­icon in the gear box.

      YAWN.

Speak Your Mind

Tell us what you're thinking...
and oh, if you want a pic to show with your comment, go get a gravatar!