Project California: Unveiled !
By JimmyRay on Fri, 03/27/09 - 1:04pm.
Besides the awesome bell bottom britches one of the coolest things about the United States Navy was all the countries I got to visit. Now, I am from a small town in Tennessee and just going to San Diego was a major league culture shock for me. But another country, well that was something all together different. I learned that just because something is cool or acceptable in the States it can get your butt kicked overseas. For example, when I first went over to the Middle East I would give folks a rousing good ole’ American thumbs up sign. Hey if it is good enough for the Fonz and pilots fixin’ to cat launch off a Carrier then this MUST be a universally accepted form of coolness. After noticing the shocked look on the locals faces or the jeers in Arabic, a resident I was scuba diving with asked me why I did not like that Dude, Lady or Kid. "What are you talking about?" I said. He replied; "Thumbs up over here means F.U"… oh man… A little knowledge goes a loooooong way.
There has been a ton of press and statements from folks around the Cisco Unified Computing System (UCS). Now I expect silly goober statements to come from analyst who think an IP address is where they go to the bathroom at. But from competitors that really should know the technology well enough to NOT make a whanker statement, I did not expect that. Statements like:
- Why is Cisco getting into the server market?
- We have been doing this for years and believe servers in a switch is a great idea
- Our solution will inter-op even with Cisco products
Certainly I do not expect a competitor to say, "Wow this is cool! Our position is to recommend our entire installed base to purchase Cisco UCS" not at all. I think understanding the technology first is really the key here before exposing yourself as a knob that did not take the time to read the material and truthfully hurting your sales force answering to their customers about this. This is a MAJOR issue for me. As a former field pre sales engineer (SE), I just hated it when marketing would gave the field some half baked response to a sales rep and then the SE would spend hours researching and trying to clean it up.
So what is UCS? I had a chance to work with the UCF team to build out an entire TechWiseTV show on this groundbreaking architecture. I was skeptical and truthfully if any other team approached us on this, I would have fought it hard to wait until version 2.0 comes out. The UCS team is unlike any team I have worked with at Cisco. When other Cisco groups are sipping wine and discussing the complexities of Kurt Vonnegut, the DC folks are eating chicken wings, drinking nickel beer and doing impressions of their favorite Family Guy lines. They are raw and they are not looking for minor changes or faster devices, they are looking to radically change the entire data center methodology and process. They are the single brightest group of folks I have worked with at Cisco. ANYTHING coming from this group is as different as a Ferrari is to a bicycle. Products like the Nexus series come from this group and there is still nothing even close to those devices.
When it comes to UCS it is ALL about the architecture and not a single component or piece. Part of the misunderstanding with UCS is a procedural norm in releasing gear. Vendors release a switch or a router or a piece of software or a VOIP device so it makes sense to look at UCS as a server looking thru those glasses. You can not look at UCS with the same glasses you look thru to eval other products because this is a true honest to goodness brand new data center architecture from stem to stern and port to starboard.
Here is a tiny bit of engineering milestones that are being overlooked on UCS that make it unlike any other product on the market today:
- The chassis that holds the blades is not designed as an intelligent manager, it is designed to remove heat and increase airflow. Traditional blade centers are like expensive pizza ovens. The UCS chassis alone has 50-70% increase in backplane airflow over other blade systems. Heat is the killer of many data centers in cost and reliability.
- This architecture was designed from the ground up for virtualization. Not just the traditional server virtualization, but it's architected in a three tier virtualized model. UCS virtualizes the compute, the network and the storage elements. Our design options here are amazing. Plus it is designed as an open model giving your data center maximum flexibility. In the spirit of keeping the data center open and flexible Cisco and VMWare submitted VN-Link to the IEEE as a proposed standard. In the system the UCS uses standardized protocols and APIs.
- VN-Link. This is like Elvis at a pork chop fry! VN-Link allows me to have consistent security, QoS and traffic policies across the entire system. A Cisco switch is actually in the hypervisor which increases performance increases my visibility and troubleshooting tools. A true first and incredibly valuable.
- Management. Data Centers have grown in more of a pod like fashion giving me a ton islands I have to manage. The UCS has ZERO management cards in the blade center itself the chassis just holds the blades and removes the heat. The management is embedded into the hardware itself. Consider this, policies like config automation, provisioning, auditing/compliance, release automation and change management all embedded. With traditional systems the more you add the more complexity this adds. With system wide embedded management we reduce parts, failure points, heat and complexity by instead of managing per server island, pod or rack, I am managing an entire system and assigning or moving resources as needed system wide.
- The adaptor. Most vendors consider an adaptor with a TCP Offloader a data center grade card. Just like bad tires can make a BMW ride like a log wagon the UCS adaptor is designed as an architecture enabler. It supports 128 virtual NICs and they can be a mixture of Fiber Channel or Ethernet. It has two 10GB ports designed in a cut thru architecture to burn those wires up! Plus the additional intelligence to optimize virtualized devices is really something else.
I could keep going here especially when it comes to the Fabric Extender but look folks, just like the old political statement; "It's all about the economy" UCS is all about the architecture. Not a single part or piece but an entire system. Hey, don’t take my word for it. I work for Cisco and they help pay my mortgage. I like to tell folks, I am an engineer first and employee second. From one engineer to another, take a look for yourself. Look at the entire architecture as a system. Wanna see it work close up and dig into the meat and tators of the system, then on 09April 2009 at 12CST tune into TechWiseTV for the episode "Dawn of the New Data Center" http://www.cisco.com/en/US/netsol/ns914/networking_solutions_program_home.html
It is unlike anything Cisco has ever released but it is built upon a foundation of everything we have ever done.
Jimmy Ray Purser
Trivia File Transfer Protocol
Indiana Jones and the Copper Scroll? According to the Dead Sea Scrolls there is a yet to be found copper scroll that is a detailed treasure map to secret tombs filled with gold, silver and spices. I'm heading out to go fedora shopping! OK it is for a copy of Red Hat...
- About Networking Geek to Geek
Jimmy Ray Purser is the technical co-host for Cisco's TechWise and BizWise TV. Jimmy Ray also conducts advanced training for engineers across North America and Europe and regularly speaks at industry conferences such as VON, CeBIT, N+I, and Networkers. As a field engineer, Jimmy Ray experiences networking first hand behind the console or in the rack. He is an active member in the IEEE and the Ethernet Alliance and has designed, installed and tested numerous networks for Fortune 500 companies, the United States military and other institutions worldwide. He holds 3 U.S. patents for Ethernet security algorithms with two others pending and one defensive publication, as well as numerous other vendor certifications in networking and security.
Purser holds a Bachelor of Science degree in electrical engineering from Southern Illinois University is currently pursuing a master of science degree in electrical engineering.
Sponsored Links
- FREE Encryption Tool - Encrypt and share your confidential files easily- Sophos
- Microsoft Open Specifications - Fostering innovation, interoperability and vibrant developer communities – Learn more- Microsoft
- Learn the advantages of web app acceleration with Akamai- Akamai
- Compuware can help you optimize application performance. Find out how. - Compuware
- Linux and Unix in Your Environment? Don’t Leave Security to Chance- Quest Software
- Securing Virtualized Data Centers - eBook - Trend Micro
- Efficiency goes up. Costs come down.- Microsoft
- SQL Server Experts: Virtually unstoppable- Microsoft
- Stay informed with custom newsletters from Tech Dispenser- Tech Dispenser
- FREE LCD Console and 16 node central management tool with purchase of any KN4 series IP KVM- ATEN
-
Network World, Inc
The Leader in Network Knowledge
It's wanker, not whanker
What a wanker
Awesome!
I have been misspelling this for years! Thank you!!!
Jimmy Ray
yeah but....
how many shops want complete and total "single-vendor lock-in"? That will be what makes or breaks UCF , and like the economy, noone can foresee that future...
(PS - i love it, i just think they should have partnered with HP or IBM on it, or at least let them get a slice of the "action" by allowing their blades to compete in the UCF chassis - Now they are stepping on the very hands that feed them - should be quite interesting! :)
You're the best
Jimmy Ray, you should review all products. Please tell me you are writing a book soon. Never stop the blogging! You are my favorite blogger. I always learn something new
Total Fluff
I'm sorry, this "explanation" is total fluff.
You say "[ ... ] there is nothing even close to [ ... ] products like the Nexus series." I'm not sure what you mean by that, but if you're talking about port density, for example, the Nexus 5000 series runs in the middle of the pack. Plenty of vendors are shipping 24-port 1RU 10GigE switches (Arista, BNT, Extreme, Force10, to name just a few -- check out http://10gbe.net/productsentry.html), and some are shipping even denser boxes (Extreme has a 32-port 1RU box, Arista has a couple of 48-port 1RU boxes). If you want over 26 ports on a N5K, you're buying a 2RU 5020, which, BTW, is a *gigantic* machine. (You guys should sell these by the pound!) Oh, and some of these guys ship L3 switches -- the N5K is L2-only. (Gotta sell some Nexus 7000's somehow, I guess -- too bad it's L3 engine is totally oversubscribed, with 80 Gbps of L3 forwarding performance to spread around 32 10GigE ports.)
You say "The chassis that holds the blades is [...] designed to remove heat and increase airflow." And then you somehow equate *removing* heat with *reducing* the heat produced by the box. Pardon the pun, but that does not compute.
You say "A Cisco switch is actually in the hypervisor which increases performance". The thing nobody ever mentions is that the Nexus 1000V is a *SOFTWARE* switch, i.e., every packet is forwarded by *SOFTWARE*, at *SOFTWARE* speeds!! Performance is going to *suck*. Even your $30 home 8-port 10/100/1000 switch has a switch ASIC inside. I would *love* to see someone plug an Ixia into a Nexus 1000V and watch the poor thing fall over dead. (The N1KV, that is, not the Ixia. :)
You say "The UCF has ZERO management cards in the blade center [...]. The management is embedded into the hardware itself." What does that even mean? Are you telling us that there is a "UCF management chip" in there somewhere? Of course not! There is not a dedicated management blade -- the management software must therefore run on each of the compute blades.
Talk about fluff
People like you make me sick. All he asked was for you to look at the entire solution. You obviously did not do that. My guess is your a angry competitor that is just puking out the claims of your marketing department. At least J.R. is calls out his angle.
The sad thing is narrow minded people like you become managers since they are too limited to stay as consultants. The Extreme switches you mentioned have been boat anchors for years. In the data center it is all about features. And by the way, of course the Nexus 1000 is software, I do not think VMWare makes a hypervisor appliance.
Reply to DC Guy
Excellent! First off, thank you DC Guy for the level of detail in your post. I am grateful you took the time to do this. On the switch level stuff that you mentioned in the first paragraph, My problem with the switches that you have mentioned in your listing is that they are good if we want to keep building data centers like we always have. Personally, energy cost, real estate, staffing are showing us we can continue to do this. The above mentioned switches are good competitor for the 6500 series of DC design. A Nexus enabled DC is a new type of DC. First off, I have a brand new code base. From entry code line 00x00 to 00x... the NX-OS was design for data center operations. It is so tweaked out that I can run a full debug on the switch and still keep cooking without my end users noticing a slow down. As for speed, in my testing using the procedures in RFC 2544 and 2899 for device benchmarking in the Ixia test suites, Wombat LBM Options Price Reporting Authority (OPRA) and Middleware Agnostic Messaging API (MAMA) benchmarks, and TCP and User Datagram Protocol (UDP) synthetic benchmarks. I achieved line rate/zero loss at all packet sizes up to 9216 jumbos AND line rate/zero loss on multicast as well. Plus DC features like single stage fabric, virtual output queuing, delayed drop and Nexus series is also peaked on the FCoE which I personally believe is the future of the DC is what makes this line of switches second to none.
History has shown us that we can not keep running separate networks. For example:
Bridges---->VLANs
PSTN------->VOIP
Power------>PoE and that standard is going even higher in wattage
Building Controls---->IP Enabled
Leased lines---->Last Mile Ethernet
I believe FC days are numbered which frames my belief in a UCS type of system. That statements always causes some folks to want to throw down, but I telling ya' FCoE is the new Sheriff in town, and the Nexus is the new station house.
As for the heat statement, I apologize for a confusing statement. Electronics of all shapes and sizes generate heat due to impurities in the conductors and the conversion of AC to DC. We need to move that heat out so I do not have in increase in thermal runaway and a magnetic field generation attracting more dust, generation more heat...physics natural conditional loop statement! 60% extra space to remove heat is a nice place to start.
As for the the Nex1K, I would say that all hypervisor based software would fail that test. The difference the Cisco software switching piece adds is really amazing. It performs better side to side then other software virtualizations. Plus I have enhanced security and truthfully, I love the port profiles that allow me to control and manage the inevitable VM sprawl that grows naturally, so much better. Plus, after using ERSPAN to get visibility into the hypervisor for troubleshooting, I just can't go back to the old way!!
Here is the real test. Business is business. Engineers can argue the values of Token Ring over Ethernet all day long BUT we vote with our purchase orders. I believe that in 9-18 months all competitors will follow suit with an architecture similar to the UCS. Mark time....NOW!
Thank you again for your comments.
Jimmy Ray Purser
Just a quick correcting on
Just a quick correcting on one of the RFCs that Jimmy has referred to, it is RFC2889 (Benchmarking Methodology for LAN Switching Devices).
Jimmy Ray - Thanks you very much for your informative posts.
Hope this helps and thanks again,
Sanjeev.
UCS - the truth
Hi all,
the Nexus 5000 is not only a 10GE machine - it supports FC at the same time at the same port alowing, to reduce the cables around 30 - 50 %. You only need 2x Nexus 5k and not 2x 10GE Scwithces and 2x FC Switches.
Nexus 1000V has less impact on the performance like VMWare's distributed Switch but it can deliver much more functionality, helping to step ahead to a full virtualized environment.
The management is done in the two redundant Nexus 5000 (these are other machines for UCS with more DRAM). The servers are stateless allowing virtualization at highest degree.
Always read the comments
For those of you who have not commented yet...I hope you are reading these! My favorite part of a good blog (like this one) is the ability to question the author (like DC Guy did) and actually get a response. You can tell Jimmy Ray uses no ghost writers but he also stands by what he says...even the mistakes. I don't want to shoot down negative comments...in fact, I would bet Jimmy Ray encourages them. It is the dialogue that we can all learn from and kudo's to Jimmy Ray for never getting angry while responding in a very detailed fashion to criticism or even just questions. The real learning is in the dialogue like this and I appreciate detail that is worthy of its own blog post... Keep writing Jimmy Ray and the rest of you...take him on. If you don't agree with him you better call it out or forever hold your peace.
Post new comment