Cisco Live Milan ACI Discussion

you so I have no slides I can probably draw some stuff on that little flip chart we got but I sell sdn supposedly for Cisco supposedly ACI is the product line that I mainly focus on so question for you is what do you think of Sdn where are you at with Sdn do you have any real knowledge of it do you think it's something real where do we start this conversation I have 40 minutes to waste of your time I think with most people there's sort of the the big unknown it's the it's still the answer of am I going to be a network engineer or programmer and how do I marry the two together and so I think that that that I think that's still at least when I talk to my customers that's still the the big unknown in their heads which is okay what does this really do for me and then how how how do my engineers of my employees skillsets how should we be developing those to reflect that need so they're still trying to get over the need part before they even tackle the well my guys just they just assume that their employees are simply not going to be suited for it and they're gonna have to be enabling training replacing something so I just want to make sure that we rephrase everything so we're on the same page so the idea is that you're absolutely in the understanding that something is going to change that the network is going to change how do the people that do it adapt okay especially knowing that change is still as of yet undefined that's fair is that is that a common thing or is there something else we want to cover there are some confusion about Sdn and network crystallization because they can be little different as the end can be applied also to our level bare metal switches a different approach of the necessary in overlay to the network but and under linking to the physical part so there are a lot of aspects and probably when Network people are looking at the overall but other people are just looking at at what is a network virtualization and maybe does not understand the order of the picture so then the second piece is when we start to look at Sdn how do we look at the physical box that routes the package versus the software that sits on top of that and the control system that sits on top of that and the rest and then again I think it tails back right into your question is how does my normal network engineer management that fair ok so it's a where I would see Sdn right now is a very interesting place and I'm not here to actually sell you anything that comes from Cisco learning of the product and if I get fired at the end of this presentation it would probably not be the first time the idea is to take a look at the word cloud how many of you got very very irritated with the word cloud five years ago and it made your skin crawl still today most of us sat around and we like argued about what is and what isn't a cloud and what's not and now cloud is just ubiquitous we understand it it's a term it's out there right look at cloud if I said cloud was a methodology of how we deliver compute resources how you can deliver a service or an application would you agree now you can do that privately you can do it publicly you can do it in a hybrid cloud it's at all fair underneath that I can deliver you infrastructure I can deliver you software or I can deliver you platform is that all fair as the ins kind of getting to that point it's it's the methodology of the fact that the network itself is broken we have not properly fixed it and moved it forward over the last 20 years how many of you are networking professionals have spent most of your careers in the networking space so that's what we had about half the room or over half the audience here so I call you to different things and I've spent a lot of years in networking so I call you do four things one I will call you a goat but that takes a little longer explanation the reason I call you a goat comes from a gentleman who I respect a lot who's at chef goats are a very adaptable animal the goats can figure out and move and do things to differentiate into whatever the environment gives them the thing I prefer is MacGyver and I know I'm in Europe so I'm gonna just guess that you may have seen or heard of the American TV show MacGyver from the 80s now MacGyver could sit in any room over a 30-minute episode and given anything laying around grab that microphone and that core didn't get this set any other and save the day that fair enough assessment as networking professionals we've done the same thing for 20 straight years there's a difference we have not been able to reset every 30 minutes we have not had the episode the TV series issue where every 30 minutes we come up with a new problem and the duct tape and bubble gum that MacGyver uses for us is IP and spanning-tree any problem we've ever been thrown we fix it with IP and spanning-tree in some new and different fashion you look at any overlay that exists today how many of you are looking at or using it overlays with your customers or in your own environments anybody nobody no nothing no overlay no MPLS heck VLAN 802 dot1q you're not using that that's these are all overlays they're an encapsulation of a protocol that we started with 20 years ago and we keep trying to reiterate the same thing and move forward so Sdn I don't care which solution you go with is an idea that the network became broken that the network itself is slow to change your virtualization admins how long does it take them to bring up new servers oh you gave me the only IT answer it depends it okay if you give me it depends at least give me a time frame take away change management people need 15 days to to spin up a new virtual machine just because the process is very long yeah and others that take 50 minutes they get a template and they deploy a template and service it really depends it it's all about processes in their company's IT you can do everything good IT today at the same level I think for Sdn is the same kind of stuff I can see today oh I'm a storage guy so I think that the plug-in guys have have at the best life and where they can lose networks and they don't lose data ok and if you lose storage you lose there ok so I I'm sorry guess what I'm not seeing today in Sdn is the simplicity there is of you there I call it democratization of networking so there is a lot of inner ignorance today in the IT world and we have a lot of generalists it's very hard for most of the companies getting the right people to do the right stuff for storage networking and visualization at eye level so they prefer to deploy simple infrastructures they are trying to avoid complexity even if the infrastructure is becoming bigger and bigger and bigger they try to simplify everything and this is the problems of as the end because you can deploy very complex infrastructure probably but what I'm not seeing today is the simplification of the process who's that that's from they're probably from the wrong point of view which is the storage point of view it's your question towards the physical appliance of a server or the concept of a server to move a new workload to which could be a VM my my question I guess maybe I timeframe right that was a question what's the time to deploy a new server yeah I'm so I'm very bad articulating whatever it is I'm trying to say what I'm trying to what what is the problem we're actually trying to solve with Sdn what what Sdn is great to type and I can sell it and I can go make money for my company what is the actual problem we're trying to solve everything everywhere apparently unicorns and rainbows that's that's how I sell it by the way but what are we what are we fixing usually there's two parts to these conversations that I don't think anybody really acknowledges there's the forwarding of traffic how do we you know the cool stuff that we can do at that level distributed firewalls either in software or you know on a fabric level whatever it is you know that's one side of it but then there's also the way that IT interacts with such a system right and I think that Sdn is trying to solve that problem too it's just that most of the time both of these things are kind of bunched together as one that's what helps you know confuse people I think that's where that's where we end up with a problem so you come from the storage side sorry are you on the the disk side you're on RI ops type guy like this is how I'm delivering you then what needs to happen from the performance of the disk one moved to him not do you even speak anywhere no but it's so to all of us our jobs ever is they everyone here what everyone here consider themselves to name infrastructure person someone who is mostly involved with delivery of infrastructure does anyone here consider themselves irrelevant irrelevant irrelevant as in having no relevance yeah I am irrelevant because I have a title that is in marketing which means I am in relevant the rest of you actually do something on a day-to-day basis that matters what we deliver though is an application or service that our users use there's the data and and that matters right when you hit Evernote when you hit date whatever it might be you don't care about their servers that are infrastructure or who they run on or the rest of that you care that your notes and whatever you need are there on every device you use that's the same thing we deliver is IT infrastructure professionals the problem of Sdn is trying to solve its that networking has fallen far far behind the rest of it and you can give a million reasons for the why I don't want to argue that but the fact is networking needs to get to a point where it can deliver on demand as fast as the rest of the infrastructure can I spin up a VM in a minute take away people and change process can I spin up a VM in a minute can that VM come up with a package so that the there's even an application and a patch management place in place can I then put on the back of that puppet chef or cisco ucs or whatever else and make it come up with the physical server and the host and everything else comes up and now we're at four minutes that's fair now if I want to deliver whatever the service is from that VM to a user I have to hit network switches network ports firewalls and load balancers how long does everything other than the VM take to get up weeks months forever never and when it does we have such a problem in language translation and such a problem in the way in which we deliver the way that actually happens how many firewalls have you looked at that have a firewall rule that you could not even begin to tell me why it's there I only know the all all that's mine this is something that happens all the time networking stayed very archaic when you look at the world today the world today we have cars that can drive themselves does anyone own a car that can parallel park itself that's pretty amazing I can't parallel park a car myself even when I'm sober which has never happens but there's cars I can parallel parked themselves and in your network you have a person who sits on a switch who puts in archaic commands any of you ever configured QoS I don't care who switches you're buying QoS configured it this is amazingly complex and if you don't hit it right on every box it doesn't work as a system at all networking is not kept up with the rest of the world because we have an importance that's different or whatever you want to say but it just hasn't Sdn doesn't matter which flavour you choose where you're going what you're buying what you want to do is about making the network operate at the speed of where everything else in the system goes that you just took a job with a company that is advanced in an IT and I'm not asking you to talk anything too deep into your job but your job is about what didn't give me a brief description enabling enabling infrastructure to be more agile and responsive enabling infrastructure to be more agile and responsive network is the least agile and responsive piece of your data center today everything is growing into a more agile and responsive piece so there if I look at Sdn in this way if you look at Sdn as an umbrella the way you do cloud esteana is a methodology of how we would look at how I want to deliver networks in the future I want to be able to look at a network as a system I want that system to be controlled in a holistic fashion and I want from that a way to programmatically instantiate that in case I want to use OpenStack above it or my own homegrown tool above it or whatever it may be above it is that a fair definition now underneath that there's different implementations you've got implementations I sell you've got implementations other people so you got in patience you don't have to buy there's a billion different pieces but the whole purpose is to get us to a point where the network is not VLAN config and switch port config on a single given pork butt anyone here configured an access control list on a switch connected to a server anyone in the room we get to head nods that's a very simple easy task right it's really easy to maintain once you put it in place you know no problem if you have some change you move it around and I love you for your sarcastic head not then continues right these network change management processes are terrible Sdn is there to get us past there the matter which solution you choose or where you go it's to get you where you already are with server virtualization it's to get you where you already are with even Linux containers and next generation of where servers are going it's to get you to where you are with bare metal servers and with storage when you look at storage day but storage perfect are you a lung guy do you carve out Lunz and deliver it via scuzzy present protocols for the most part a few years ago not anymore though so how do you deliver now I won't believe fortunately ammok of something so I can do that people are doing things wrong but how'd it go I mean the best place now but now I'm in sales I'm in actually the best place cuz I walk away as soon as you buy it I did not say that please for the cameras but but the idea is even storage there's a lot of complexity behind it but but it's pool that's virtualized right we haven't done that in the network and that's where we're going and the whole purpose is that as infrastructure we are the most important piece of keeping things running but we're the slowest piece of getting them there and when things go wrong it's hard to troubleshoot and on the networking side then for the networking people in the room you'll like this for everyone else you'll hate this we're the first to blame when something goes wrong it's the network's problem right always a network so how do we go back how do we troubleshoot that how do we look at what goes on how to mean time to innocence is what I call it and whatnot I didn't name that it's not mine but I use it and clothing just don't go together it's same sentence fair enough but the whole idea is is we need to get to a point where infrastructure delivers the services and applications that the business requires on time and on demand and I don't care if you use network virtualization or Cisco ACI or whatever is you choose to use the idea is that we need to move infrastructure out of the way of getting things up and running but without sacrificing that second day operation without sacrificing when things go wrong because they do how do I troubleshoot and not just troubleshoot or mediate how do I find out where it is that fair enough all right so who has a really like who wants to be the jerk who has a really really difficult question because I think I have like three minutes left based on my having no ideas I mean so customers that you can talk about with the trialing and waltz and all scenarios for the products I sell specifically the products I sell nobody's bought any I ever know there's a the use cases are typically people that are looking for a change in the way in which they operate what they need to do now if you have a data center that consists of four or five switches that has a very low change rate you don't need my solution or my competitor solution or any of these other things because you can have one person with the CCNA or whatever their certification is go in and make a config change once a month it's not really that big a deal when you have a change rate and you actually have some scale to what you're delivering and your applications move now we're starting to see some adoption so we're I'm seeing the biggest adoption of Sdn and it's multiple flavors and all I'm happy with all of them is when the applications and services that I have to deliver change very quickly my business demands something different from me I need to make it happen the network itself today is probably that slowest and when I look at the network I don't look at that layer 1 through 3 Network I look at there's also load balancers and firewalls and other things that go through that packet needs to go through to deliver to users that's the kind of the slowest piece so as soon as you start to see any type of change in that that's where the use case is so it's access control list management it's security management it's firewall management it's micro segmentation these are the places that I see Sdn being adopted and is it being adopted widely throughout the entire customer base you all tell me how many of you work with a customer right now that that is adopting Sdn no and that's the problem right of the global IT organizations what percentage of them are actually relevant to a conversation around Sdn proportional to the number of IT organizations that are be having it the term shoved down their throats yeah I think that's right it's it's the the massive problem is that everybody is is is saying is this right for me and in reality most of it most businesses it's hard to quantify most businesses right so you have to say you know of the businesses that exist today how many of them really actually need that agile response to on-demand network services generally speaking very few of them yeah I would agree with that I would say that every every customer I work with on every datacenter I've worked with needs a better way to do VLAN management and access control s management every single one there is a problem how much they are willing to pay for that because if you have to change a VLAN every week in a traditional midsize company so they how can this account how can they spend so much money in a news reach capable of being configured magically and not not that meaning burden isn't isn't typing the archaic command versus clicking the button the burden is the process surrounding getting to that point once you get the year the approvals process that's the burden right it's once you get down to hey this service needs to live this place on the network it's you that's the business relevance of the conversation and and whether or not we've wrapped it around a button that the person can click or an archaic command they can type into a switch approach not gonna fix it that's right I was looking for trends of a sense of safety we have now one place where we can go and push this off we you know do we now trust this more than loads of people going off and configuring loads of boxes it's more layer 8 thing thing is I I do see myself asking for a VLAN or a trunk or whatever after the approval it still takes 10 to 15 days before you oh I now I got a hole in my calendar to do it so if I put it if I can put in an automated request and it is executed on the approval that would save me 15 days but how many networks are how many people are actually going to say yes we will approve that that process because if there's not a human sitting there doing that there is errors there's this little process I don't yeah how do I know that that script actually ran and executed and completely made the change it needed to happen they're always gonna have you know we can kick off a script to turn up how many servers and how many storage connections and everything and have it happen in five minutes but you're still gonna have that Hulme that goes back through and verifies everything was done and I think that's I mean this is where the Sdn it sounds so great because I mean I love to script things if I can replace the human with a script I'm gonna do it but you still have to go back in and reput and check it Beverley should be approver be the one that's actually executing the widget right regardless of which problem about a decade ago when when all QA was done manually right because because the QA department was a completely separate team aside from the dev that actually created the product right yeah and they sat down and they're like all right I'm going to be I'm going to emulate a user and I'm gonna try to break this break what you've done right and everything's done manually and if they miss a use case then bugs get made into the product and you know customers see it and test-driven development changed that you know dramatically because now all of your scenarios are documented is actually part of the codebase and now you have a feedback loop right and networking the tools actually still I my opinion the tools can get there quickly but the people need to understand that the same value can be realized because now you solve the problem of automating changes everybody write scripts that's not a big deal how do we verify that and then say I know I I have to interrupt all of you because I believe somebody in my company paid for me to talk so I don't really care about your own opinions but no to get back to that and and in everything you all addressed it where do I see Sdn getting adopted right now Sdn gets adopted the most in universities who are getting government grants to adopt and test it that is the biggest place I see it getting adopted beyond that where do I see it getting adopted in the very large scale companies in the eBay's the Amazons the other places the world who can hire people to actually build it for them and take what's already out there and then like Matt does Matt you are what now a network programmer you're not a network engineer is that correct and I apologize for committing to call you out but though that's where I really see the biggest adoption now does that mean that we're not gonna go there no but the big question is with any big hype what is the problem with my business and the delivery of my infrastructure that you can help me solve if Sdn is the term fantastic what is it you're actually doing for me and in some cases it really works now it's moving the needle forward right now a network is an archaic state if you've configured QLS you've literally like shipped pictures of dinosaurs into a rock with another rock right it's and we're still sitting there that's still how it's done so we are moving forward it is moving the needle it will go slower and for most of us it doesn't need to happen overnight and it shouldn't happen overnight we'll figure out the little places in the places it goes and where it makes sense but the overall idea is that you make the network something that is a piece of the tool that delivers the application and service that you're looking to deliver is that fair okay all the other parts of the infrastructure you've got like pools of resources right we don't talk about servers anymore really we don't we talk we talk about compute pools storage pools we still talk about the network in terms of you know discrete devices the box not even discrete devices we get further it's like VLANs and protocols and ports and this set and the other when you want to troubleshoot exchange you have to get back down to some spreadsheet that tells you which subnet or VLAN that exchange mailbox server is in them that that is incorrect that's not how we work in any other part of the IT infrastructure IP management via spreadsheets needs to die I sorry I have to understand because for for what I see historically speaking now we had the my frame then the UNIX box is done trying servers now open stack yada yada and I think that infrastructurally I'm speaking if I can use this term and we are creating another side over here so as the end is something that works very very fine with all that OpenStack docker and that kind of stuff just if you if you step back in the virtualization environment is a little bit difficult if you go into legacy environment it's impossible to adopt so you are building something you okay great but you are not addressing still 50 60 70 percent of the problems that traditional enterprises have because you it's quite difficult to manage that kind of machines and and you know IBM as reason knew my friend a few days ago I'm not a fan but most of the banks all around the world are still there and they will upgrade their my frames and they will stay with their UNIX machines for a few years more and answer the traditional visualization because now it's no more the new stuff visualization is the old stuff so we are building anything that works very well with all these DevOps all these can tell where the developer asked so when I deploy a new container probably I need to do some operations somewhere else in the stack to make it working properly so it worked there but the older the rest of the SDN doesn't work with the old stuff I know something in there in the networking space but I think that the UNIX is Unix and you can do a lot of right and there are there are things that we're not gonna pull off mainframes if you look at mainframe sales over the last what five years they're flat they're not declining they're not going up there's people are still buying them they're UNIX systems AIX system there's a lot of different things that we're not gonna go rewrite and rebuild and I've been in a hospital within the last four years that there was a Novell server sitting under a rack that nobody knew what it did or what was going on the light was still green and they wouldn't touch it because they were pretty sure that it touched something else that mattered that was patient care right so St Anne's not gonna touch everything very very very quickly and all the rest but if you remember when virtualization came around when virtualization came around every application developer said no damn way my application is being virtualized every application owner said no damn way I can run on anything but bare metal and now virtualization is what 70% if I throw out that number is everyone in a happy prison 70% of workloads Sdn is getting to that point I'm talking about a world in 2007 or in what 2015 now if I my math is correct it's not usually we're moving forward is St in for everybody no do we have SD and fully defined know what we do have that is unanimous right now and any of you please disagree with me I'm not gonna take offense is that the way we do networking is broken in comparison into the way we do the rest of infrastructure you do not go do command-line configs on a day-to-day basis any other piece of your infrastructure other than networking and so I don't care what flavor of networking STM we go down I don't care if it's a open flow or whatever it might be we should not be touching a box and hitting that command line to make a config change so that a new application or service and come online and it's not just because it slows us down to get there it's when something goes wrong because guess what something always goes wrong so when you have that data breach and you're now on CNN Headline News how do you go verify that you had the correct tools in place to say hey I was in line with laws today you go to a command line and you pull all that information back out s teen is about getting us back out of there and it doesn't matter which vendor you talk to build it at home build it yourself go anywhere you want to go it's about moving us back off that command line you don't do it with your server anymore you don't do it with your storage anymore even storage you now pull it together you deliver it to a group of VM so as Williams spin up why are we doing this with what I'm not getting at this point is why for example all the hyper converters players are talking about this pool of resources in a cluster or machine and they are not hyper converging networking - so there are a lot of utilization a lot of storage virtualization and this machines but no networking at all it's so I would like to see the simplicity guys or the Nutanix guys with a and I per converged at appliers also for networking because hyper commands are really simplify the IT infrastructure I would like to see a CIO and a sex or whatever the product embedded in in this kind of appliances it's not address it today that's your drawing in two different conversations so hyper convergences is a great thing at home I use the system called so I'm not allowed to talk about but I love the system they're fantastic so Knossos is great I would consider it a hyper-converged home audio system I put a speaker in any room I want I call it which room it is and I play all the music throughout my house and they all play the same at the exact same time I can walk through the house that never skips a beat or I can have 17 people in 17 different rooms my house isn't that big I'm lying I'm just trying to make myself sound better but uh and they can all play different things right it's a very hyper-converged speaker system now if I was running this hotel Sonos wouldn't work for me so hyperconvergence has a place at a point where you're going to scale and it has a cap when you look at hyperconvergence at some point your CPU or your memory or your land or your whatever it may be or your disk space is going to exceed the requirements of one other piece and so hyperconvergence works within its realm better than anything else but outside of that realm the same as my music system at home and a million other things I don't necessarily personally assist me I don't think that has a fit so I wouldn't I wouldn't draw hyperconvergence into an SDN because if you get a very large hyper-converged system so all the resources you have so many resources that you can sustain very different workloads and big workload of any kind yeah I agree with you that if the system is very small you probably run out of storage of RAM or whatever very quickly because probably you have a single work workload running on that system but if I have say 100 servers I can run VDI alongside I don't know a traditional visualization probably because some of the resources are used for VDI but other other servers are less impacted in fact less on on the sorry on on the storage but more online for example so it's not true and on the other side I'm reading a lot about Facebook Google and the other kind of guys they are thinking a lot about very flat networks now you know Facebook came up with this Dirac interconnected Rex we just reach on top of Dirac and they are building this very flat networks so that there are one Facebook one Google yeah probably one of they clear the others okay I get it and then I want I wanna and that's the end more so for the rest of us for the traditional enterprise and it's not today I wouldn't like to see that not in five ten years on this right and you said it's my exact point and I have to close I've already run over time my apologies to you in the audience but Facebook is a Facebook Google is a Google there nobody else is gonna run a network or a data store the way they do the people that build their data centers build their data centers down to in some cases the server components Sdn is looking to help you solve a problem if you have that problem look at s the end if you don't have that problem and switch configuration still works for you don't move forward yet it's not time but I have a case and how many a CI customer do you up today what I would love to have the name John Chambers and if I had the name John Chambers my car would be much prettier and I can answer that question that's an S that's not no common questions yeah that's nothing like that I'm sure but well but you were trying to get fired by the end of the presentation show why not give the number anyhow yeah zero we have not sold a CI to anybody Cisco's been trying we've been trying very very hard to sell a CI and everyone said no it's this is craziness what are you doing no I can't I literally cannot answer the how many a CI quite can't do it what we would say is that we are very happy with our growth and our progression which is meaningless market texture and that's what I'm allowed to say and I very much appreciate your time thank you very much for letting me talk to you for a little over than what I was supposed to

Channel: Tech Field Day

Views: 5,435

Rating: 5 out of 5

Keywords: Tech Field Day, Tech Field Day Extra, Cisco Live, Cisco Live Milan, Cisco Live Europe, Software Defined Networking, ACI, SDN, Joe Onisick, Insieme

Id: PHKYKDqRbSw

Length: 37min 48sec (2268 seconds)

Published: Sat Jan 31 2015