Cisco UCS: What’s New

I am because I mean I'm in the product management team of UCS at such want to make sure slides or flipping curve so today I am here to share some of the good updates that we have made to use his product line yes UCS is here and still continuing the Thrive so I will start with actually some of the market momentum that we continue to see around it then we'll go into the specifics of product update that we have made and we have on show floor we'll quickly touch upon inter-site some interesting updates there and we'll wrap up time permitting we'll wrap up our session with a look into the future technologies of interest and how they're going to be impacting the platform of tomorrow so first though let me paint a picture on the data center product portfolio that Cisco has and UCS is part of that molecular for you so the very bottom in the gray you have some products listed there Nexus UCS and so forth in green is where a lot of focus goes on to bring the automation as you deploy those product and maintain it in the blue we have you have the software products which bridge the gap between the on-prem and the public cloud for the customer workflow to be deployed and in that category you have at the Cisco cloud mile of cloud center container platform and so forth and in orange you have the security products that we have got security as at the heart and the center of any product design that we do the Cisco UCS is right there and this session is going to be only about the UCS everything else will probably be covered in the separate sessions so just to get started let's talk about the momentum this product was released a decade ago we are pleased to share with you I think this this is publicly available and out we are now number one worldwide share in in blade market number one in Americas in blade revenue and we continue to hold the performance bench benchmark records across various categories customer momentum standpoint over 55,000 customers yes eight or nine years ago actually we celebrated ten years so it's a decade of journey with you seeis fifty-five thousand customers hyperflex that was released as three years ago continues to grow rapidly over 4,000 customers and inter-site which is the latest edition is also growing rapidly and we have over three thousand customers we continue to roll out Cisco validated design for those who are under familiar with Cisco Verity design we not only roll out the product but we also invest time in putting together the solutions which give our customers the recipe recipes that automation scripts to deploy their workload you can stand up line for entire infrastructure OS install all the way to the application level we provide the complete recipe and we provide the front and support if customers are run into any issue so we have over 250 plus of those Cisco ability design out there we have shipped tons of stories through our partnership with variety of storage partners there and hyper flex and today happy to announce Cisco inter-site manages over half-a-million device across the territory it's a significant I have a slide coming up to talk about that interview already starting now so that's quite a lot of devices managed by their side so he's intersection of super setting UCS manager or are they staying side by side yeah so that's an incremental journey the inter site has taken their steps to make the take the manageability to the next level especially because of that fact the scale the cross site and all those things eventually as it is going to provide everything that you see as manager does and more I'll not spend much time on it but typically people ask about the product differentiation this is a slide that speaks about the value prop that customers see we are known for reduction in the provisioning time we are known for reduction in cabling and that comes from the unified fabric that we have reduction in management cost to service reps profile and so forth the reduction of power and cooling the reduction we are known for performance so accelerate your workload and we are known for reducing your deployment time because of the cisco validated design this is a quick look at the Cisco portfolio that has evolved over the ear product for Robo then in the second is where the core UCS platform is which brought the converse computer network to the market building on top of that is converged infrastructure product which we created working with the top tier storage partners we released hyper hyper conversed which again builds up the same underlying architecture of UCS with hyper flex software and then the products on the right three are very application oriented there are applications which are core sensitive so for that we released C 4200 which packs many many cores which smelt of AMD our CPUs then there are use cases where you need tons of local storage so that's where s series goes in and then finally the new thing machine learning deep learning for that we have C 480 ml which packs tons of CUDA cores in that for you box you have 40,000 CUDA cores even though you know you see you may think that there are several different product lines here is one interesting aspect they have been built out of same common architecture we call it it's one system which powers all the workload so our journey started with blade and rack servers and we targeted the mainstream enterprise workload but as we figured out the need for other application requirements we started to add more servers for memory memory sensitive applications we added C 480 be 484 GPU sensitive applications we added 480 ml stories acessories came in and high density core you got the C 4200 to you for node they are all fundamentally powered by the same common architecture that brought UCS to the market so that adherence to that common fundamental architecture has allowed us to deliver the consistent manageability experience the same benefit across all product line and has helped us with the velocity with which we release the product with with the with one of the fine the finest engineering teams that we have got and building your own CI solution alright obviously a solution is continuing to be partnered with the parties right sorry you're going to do that with your partners not as cisco itself so historically the CI solutions we have always done with the partners we roll out the reference architecture so we will continue to do that okay yeah and not that you you showed a stack like this one could ask itself if you're going to build your own CI stack as Cisco Oh got it that's a good catch I should have probably included CI right in the middle one because that's the product that we came out with so we absolutely intend to continue with the CI a product line the way it was depicted in the prior slide good catch yeah okay so now let's talk about some of the platform updates okay so first our fabric interconnect is entering the fourth generation so we're increasing both the port density as well as the port bandwidths there so this one will have seven point four data bits of switching performance 96 ports providing 10 to 25 bit capability and 12 uplink ports with 4100 great capabilities those 16 unified ports could be either unless Ethernet or fiber channel fiber channel with 8 16 or 32 gig depending on the customers scenarios fabric interconnect is complemented with corresponding Vic our Vic's are known to reduce the number of HBAs that the customers need because you get on a one Vic you can instantiate and number of Ethernet cards or n number of logical FC cards so they're at the heart of you know convergence of the fabric and the reduction of the HP is the customers would otherwise need they are entering the fourth generation and we are introducing the wigs which are 10 25 40 or 100 Gig capable they will have nvme over fabric belt on rocky v2 Fibre Channel over nbme support and the VX land the envy GRE vmm virtual machine multiple QT cam filters and so forth just a quick question on fabric interconnect update which isaac is inside this first generation what's the magic the adzick the nesic oh what's the ASIC inside this yes is it a cisco one or submergent silicon so the ASIC we built but I don't know which vendor we are using so you'll have to excuse me for that one but I'll get you the answer with the man is in there my apologies then we are introduced this is a net new addition to our our our family of products we are introducing and this is on the show floor we are introducing a short depth rack server this is purpose-built for edge and remote locations this is going to provide up to four drives with two PCIe GPU or six drives sort of nvme SAS any one of those choice and this is I'm going to be introduced sometime summer this year and this is on the show floor the c-4 80ml our server for machine learning is getting a refresh it will have cascade Lake and opt-in Drive support and finally our high coal density server is getting the AMD next generation processor refresh to drive up the tenth of the performance there so in total this is the update that we are sharing with our partners customers at the Cisco live this great effort that has gone in by our engineering team and this is a reflection of our commitment to the customers in the UCS product line that the comment was made earlier and just a quick question regarding this because you have now MD editors epoxy for you do you see a high demand on that because at the moment and these more or less catching up when it comes to horsepowers on the cpu great question so the products that we have if I can I would really argue on your on your argument because I think that's in in the data center world we are really excited about AMG because you are getting much more power than you're getting with an Intel CPU and you are also getting all of that the same amount of course in a single CPU you know so you're you are from now you can cut your costs on virtualization yeah so by granting one CPU instead of two so all of the major enterprise organizations are really evaluating at the moment in the realm so I think that it's a it's a great thing it's a great thing to have because you can end up so the thing is what's going to happen if for example VMware decides that they change totally their licensing number you know method they can very well install not a VM MV memory tax but V core tax and just render the wall argument irrelevant but I think that there's really huge difference for MD yes maybe maybe don't get me wrong I see you aim D has now it's now the position that they have after 10 years were interred Amon ated the market a very competitive product yeah and I think that was kind of my cursor I'm not the server person but I can imagine that a lot of people are looking into AMD and is keen to get that box into their data centers here yeah I think you're right we're beginning to see the demand I wouldn't say same level yet as in Dallas and has max pointed out their exploratory phase and we learn more as we get in the market as we drive it probably in the next tech field session I would be happy that I'll be able to share more data but it's an early phase we do see the demand that's why we are so what we're on these slides and talking about hardware and what's coming soon I have to ask this question and you don't have of course to answer it but as you say it's UCS has been around for 10 years I think we're still using the far the same 5,000 or 108 chassis which if I remember when I was still making UCS quotes was given to be end-of-life around 2020 2021 so I think we're getting at the edge of a refresh which may not be now maybe this samurai yes but how all of these components play with what's coming that's because something has to come next right I mean if I have a chassis which is expiring and if you're going to do that so not asking you to answer and earth person but is there some kind of backwards or forward compatibility and these elements we're seeing here absolutely so I will answer it in two parts one right now and other as I go to get into the Millea my last two slides on the future technology trends first in the UCS the reason that jessie has not changed over the ten years is not because we have not innovated and evolved the way it was designed that chassis allowed us to keep on evolving and innovating on the plates that's why every time Intel required sir required as the blades changed so it has kept on changing we had enough real estate to keep on fitting in more and more deer amps and also it has completely been changing but the founders and I can't take the credit I wasn't there did a really an excellent job of of evangelizing a chassis that will do good to us over ten years and still be the market leader and we do send the second part I'll answer in the second round we do see even as we move forward there is enough substance in that chassis took let it continue while allowing us to innovate at the places that it is desired to be evolved and out over that some of them so you see no various who have maybe and in the Rome plate at one point in time that would be completely market driven right now we saw the high-density demand that's why to you for and came in and we want to learn with the market as well but no technical limitations from your and to add this in the exist that's correct that's correct the way the system is designed yes the CPUs could be yeah all right so a quick update so this is not an hyper flex session but since this pertains to the platform I have one slide on hyper flex all nvme hyper flex in particular so as if you recall all nvme hyper flex was launched last year when it was released it was the first fully integrated all in vme hyper converse from any vendor with full res features so and we immediately start and SAS they were challenged on the hot plug and and the LED management and firmware upgrade because and nvme drives connect directly to the CPUs so someone had to solve the issue of hot plug surprise removal ever led light and all those things us owning the platform allowed us the opportunity to do that we own the platform you own the BIOS we have a strong relationship with Intel and we own the software so we released it there's a very big focus on that we focused on performance and we continued improving it through the last year and we if you compare our all nvme hyper flex with all our flash we are the market leaders in all Flass in terms of performance even considering that all nvme takes the performance to the next level you get 90 percent more i ups 50% lower latency and latency has been one thing that we have been laser focused on and sustained latency they are off we released solutions over the last year on Oracle on Oracle RAC sorry you're saying you're the leader in something in performance in the app yeah so and they're publicly available a white paper and power benchmark results from independent you know vendors we are leader in terms of delivering the performance both in war I often latency yeah okay but for storage for hei for its happy ok hyperreflexia so this hyper-converged same hyper-converged yeah structure yes ok then recently we released 8 terabyte drives for all nbme a terabyte nvme drives so that takes your per node raw capacity 264 terabyte so couple that capacity with the performance now users have unique opportunity to consolidate their via VMs even further in a smaller footprint and all this thing and this is the key announcement that we are making we have worked with our dry partners to take all nvme and make it mainstream because one of the prohibiting factors in mass adoption of all nvme has been the cost so far we had been selling it at 3 X and plus primarily driven from the nvme drives happy to announce today you get all this benefits at only 5% premium which is a throwaway premium over all flash and for 8 capacity including the software license so we are not taking all and be any mainstream okay what's that question sorry okay sorry about that I'll just sit back okay all right so with that let me cover some few statistics on inter-site so inter-site as you may know it's the cloud hosted management platform it is changing the management experience completely in terms of how people view and manners and install and deploy the UCS and fabric interconnect all over all over their data centers today we have over five over four hundred fifty thousand plus connected UCS servers being managed by the centralized management platform over 46,000 fabric connects fabric interconnects are managed with inter site over twenty six thousand five hundred hyper flex nodes are being managed through into site and 6000 plus hyper flex clusters are being managed the power of inter-site actually becomes very evident when customers are happy that we are reporting to them the problem that they may see or they may already have with the resolution and the way it works is we have centralized telemetry being done we are constantly looking into the challenge that customers may you know run into and if there is an error we automatically open the tactical if there is a known instance of that issue we also you know send the guidance on how to fix that so it's been just truly a differentiating experience from the customer standpoint which is driving the popularity and adoption of inter-site quick questions on these slides when you see connected you're referring to the cloud-based version of inter-site are you talking overall as a mix of cloud-based inter sites plus on-premises versions only cloud-based it may be more than that in here yes actually the on-premise percent of inter-site is not publicly available yet so this is I think this is the number yeah so you've got twenty six thousand dish connected type of flex service yes or notes let's put it that way yep and you've got I think it was forty five hundred customers four thousand plus customers here oh that's around six oh no no no no no so not all hyperflex are being managed with inter-site okay the numbers are much more but these many nodes are being managed by inter-site okay inter-site in the cloud yeah right yes if somebody is doing is on-premise then you don't have the number so yes so that is the number we don't have absolutely right okay and our hope is as more and more customers see and they've gonna start using it even surely all of them will be managed through intersect so this is phenomenal given just two years of inter-site being there and with incremental product updates that is coming in we are seeing very good traction and the other thing that inter-site based management model has provided us as the velocity with which we could roll out the features just from the time it was released to now look at the number of features that we have rolled out and and the cadence and the velocity the velocity has been just amazing this is both a reflection of the architecture and the genetic team as well as our commitment to drive it and lead the transformation of the management experience for our customers okay so with that let's switch over to however before I switch over to the futures any other questions so far because I'm going to turn into a little bit of a look into the future and what we see here is happening and so forth any questions I was just curious you showed before the P certs advisories yeah the other one peace earth yeah can you just give me some background information so house is working if if the P cert releases an advisory or volure nobility you immediately have this in inter site and then you say okay customer you have to patch this in this device because they are vulnerable to this new released absolutely now we don't have to make a call and all this is fully automated we automatically know what all systems are impacted we already have contact information a way of sending the email and everything you are just leveraging that infrastructure to push down the security vulnerabilities and in cases where the agreements are in place patches as well so that you just don't even have to worry about it okay so it starts with that then we'll get over there that's great yeah well I think in the past the pain yeah there's a vulnerability you have to check all your devices which versions make comparisons and then the patch level and Sonia and that's a very important point that you picked up I think that was the the first order of that problem was having an ability to the inventory and maintain every component that goes into the computer system so that's what UCS manager salt years ago that anything that goes into the UCS manager it's all we maintain and manage it down to the former level so we had good visibility and to what's in there now we are coupling that since we are bringing it into inter-site now we have collection of all the UCS system that are there with that level of control down to the drive from their level so regardless of where the security implications are we can instrument and with them you know we do not even the drive from their vendors that we work with they have to adhere to certain interfaces and all that we require to manage their firmware so we leverage all those to do that so very excellent point that you picked up okay so with that let's transition into future a little bit so max was right that future is interesting if I we just started putting together on the whiteboard what are the key words that we hear not all of them are compute but this is not even half the key words that I have put in there several different things are going to so then we said all right let's at least bucket eyes them into some categories so that we can start working towards that to start assessing our UCS system our our product line as to where the implication could be and the four buckets that we came out are this sorry for the color coordination here these are four buckets that we came out with in compute sight the impact that is going to happen is because of the high watts CPUs hiwatt CPUs from Intel they're touching 350 watts and north GPUs 400 watt and north of it persistent memory just a game changer smart neck fpg offload increment Lee finding the use case all over again in the fabric side 400 gate coming up PCIe for ten years to come Jen five is right around the corner now see Excel which runs on PCI memory fabric Genji and willing and so forth so looking at this bucket allowed us to start looking our UCS system and as we pointed out earlier the good thing that we have yes we have to evolve yes we have to do the things but we don't have to dramatically just change everything because if you map it out into what the ideal infrastructure should look like it calls out for it should have the space for accelerators so it should be designed with the real state and power in mind which our UCS already currently provides that all those things delivered in a very composable fashion could make the composability itself very complex if it's not done right that's where our founding service profile architecture with expansion will come handy so we have experience we have foundation to work on and as VC hn5 and other four hundred gate comes in if you have it can you imagine having another set of fabric coming in and complicating things so that's where our existing FY and Vic architecture comes handy we are just going to evolve it further and then infinity to scale it's a factor of the management that's what inter-site is making the steps so that it doesn't matter you can have n number of units all those things half a million can go to easily to half a billion no problems so the point M drive was trying to make is yes we have you are looking into the futures but what we are blessed with is you know the founding team and the engineering team has given as an architecture which will evolve but we do not really require to do a dramatic change into this one

Channel: Cisco

Views: 7,333

Keywords: TFD, TFDx, Tech Field Day, Cisco Live, Cisco Live US, Cisco Live Europe, CLEUR, Tech Field Day extra, UCS, Compute

Id: kergKplZ48Y

Length: 26min 32sec (1592 seconds)

Published: Wed Feb 12 2020