The Future of Software Defined Storage

neighing hi I'm chase McGuire editor of data mission and our topic today is the future of data storage the future of software-defined storage to talk about that i'm joined by a top thought leader in the sector with this is Vincent Siew chief technical officer for storage and software-defined environments at IBM hey Vincent how are you doing today pretty good thank you James and you are there in sunny Tucson correct correct very warm win under you get around your goggle you're in the bay area so you you're putting a lot of miles earlier this year series yes yes I mean it's always exciting so I I know that yeah we talked before him we decided to throw out all the covers all the all the questions and just like start a new which I appreciate that like this so I think that the idea is cognitive capabilities which is interesting because you think about the classic data center with these vast reservoirs but sometimes these are some dumb machines sitting there just holding the holding the data we want to do more with the data than merely store it the question is you know cognitive capabilities how is cognitive capabilities you really transforming the IT process in your view yeah so what is cognitive kV let's be really clear what do you mean when you say cognitive capabilities yeah cognitive capabilities that bill too I think that the fundamental desired fundamental objectives is to be able to precisely capture the insight of the data you know I know is you know that people are collecting data you know people any more and more data right there is growing the exponential rate this was true ten years ago is even truer now the problem is how do you turn that huge amount of digital assets into insight that you can leverage for your advantage and in the past the people has really relatively sort of deterministic algorithms right and how do you or if for example that we talked about the data securities and those are all quite determination but when we start to realize that the world is a lot more complicated then we need the the more intelligence artificial intelligence capability to help us to analyze those data and to extract the right values and co-read in those information that's even harder with all this information what data is written to what information and create a new information create a new values from the combination of those the information and those higher-level capability we call the cognitive is that a series of tools software tools software in combination with hardware what is the actual tool itself yes they are iBM has significance is same significance understatements has a very significant investment on the cognitive software suite like Watson knowledge studios and things like that right now we are integrating that capability with infrastructures so we allow that that that the cognitive soft would be able to harvest the metadata of the system all of your data and allow the user be able to customize those data and be able to Corey allows the metadata to generate a new insight for you for example they can say that how often you touch the file is it is this particular user has a right has the has a normal assets pattern of all those informations etc etc so we can coordinate all this information and create a new insight that people don't have it before and this is very exciting that we are we are looking at how do we apply that in the life science how do we Coriolis research data and when we try to accelerate how power you know the medical communities how to accelerate those medical research which is it's a service like in a pervasive analytics I mean you want to like just you know not just store the data you want you you want to constantly be you know understanding with the data means and you want you want any licks in all your various domains your various devices how does pervasive analytics tying to the idea yeah this is a great term I really like this term called pervasive anali anali is no longer just a very very narrow defined in the past when people talk about Big Data some people would define that wallet just had to sort of MapReduce an overload or spark its bark but no longer that's the case I think that we talked about remember we need the the initiative of the vision of ecology to be able to get the best insight from the data aggregated data correlation on the data then we need to be able to run analytic in in lots of places you cannot in a past that no as I say that people are thinking about big data then load the big data they load the data to the Hadoop clusters and run the HDA hadoo computations and get the data those things are if the data becomes evidence that become you know large it's really difficult asset difficult thing to do so what we do here is enable that people be able to run that had to peel run spark be able to run no sequel database that across all data sources your data sources object stores no problem we have we have an accelerator allow you to have a 10x first a performance improvement allow you to build on spark on object storage that in the past people has a perception that object storage is slow and but if you would if by far more your data is the object storage we need to find a way to allow you to be able to run your and added capability function on it and that's what we are working out our research team is working on right now is there a name for that accelerator at this point or is it still in research it's his research name it will be a a spotless iterative but i'm not a marketing uh uh god- he I don't want to come with a new name for them no we just called spark super fast oh yeah I'm spot superhighway something good yeah just a sure I shirt changed about ahead dupage sparking that she like you know Hadoop really changed analytics in many ways but it seems like I spark is more on the upswing than Hadoop was because it is perceived as sparking so much faster than I do that it's not so much spa get faster it's I do really require you to load the data to the hunted clusters you cannot I mean sure right now we mean by the way we start to enable a duel runner object storage but up until now the prac that common practice if you want to analyze the set of data you load the data from whatever the data source you have put on the Hadoop clusters uh-huh right and they data that process you know let's say take two hours once you have a few terabytes but one thing's go beyond petabytes or the job gonna run for like you know wheeze a month and then become a burden I do a bass part doesn't have that sparkles and it sort of it so break the paradigms for people to run the analytic and where the data is and that's where the enablement that we our research team is working out to allow people to be able to you know be able to run you know sparks in any of their data sources hmm what about the idea of you know hybrid cloud and it comes into you know the software-defined storage and I think you know suffered to find a story just sometimes siloed partially because the idea of an entirely software-defined datacenter is still somewhat of a futurist dream but if you think about software-defined storage in combination with the cloud there's all sorts of possibilities there I mean for the hybrid cloud where does that go for you these those things are not disconnected the thing we talked about cognitive too pervasive and edit too hyper car those are not disconnected the reason hyper cloud is important is given that I from the original the initial vision that I want to be able to harvest the insight on all my data then my data is spread across unplanned opera in the cloud the private cloud and I need to have a federation allow me to be able to seamless operate my data center across all this not just one particular cloud not the particular on print technology I want those technology to be um to be working in harmony if you will all I need is that tool exists I guess that's what I wanted I'm curious is that tool you just talked about that pervasive does that actually exist that our I my teams and our research team partner to work on those Federation to allow us to be able to seamlessly connecting the data from the unprinted at you off up to the cloud and be able to spin up the services on the cloud those things is coming is comfortable all right so it sits on the horizon correct in the near what else about hybrid cloud and pervasive computing how does I worked so the thing is that I think that I just want to bring up two more aspect of that one is the workload management okay when we right now we just talk about how we're going to manage my data on the arm Prem private cloud public clouds and that you have to ask your questions is how does i distributed workloads which workload unique and which you look workload should be wrong at which kyle is the most optimal solutions isn't right you don't want your workloads running on particular storage platforms and the data is in particular occasion but data is all far away right so you need to figure out how do you manage your workload to optimize the efficiencies right the second thing is providing insight of now with all this cloud connected the next questions I'll get this question from my client a lot where exactly my data is yes right it's hard to know now I need to meet and by the way some the ITSM that need to be able to provide a meeting of those be able to ride the the health check of those data before the provided data protection so those things are two more important topic that and know when we open up these hybrid car capabilities yeah you know I want to get a further sense of where you think storage is going this year but look can just do it a little bit about flash because flesh may not be as exciting as someone as futuristic it's like you know pervasive computing but seems like flash has had a number of years here on the upswing either of your three to five years I wonder did I mean flash is still extremely important well I thought I said I wonder if is flash in your view is a kind of plateaued in terms of its adoption is still on a rapid upswing okay so I want to expand the definition of flashy little bit beyond just the nand flash okay i want to call it a solid state storage technology right including the name flash including the phase change memory mmm okay the phase change memory that allow us to have a light you know ten times better 100 times better endurance of the flash and sure it's more expensive but has much lower latency which you will provided next tier will create a new tier between a dram and fastest flash storage so we will see that happening we will see the software-defined storage to take advantage of that I phase-change memory capability the storage class memory capabilities on a flash I I my protein for this year's we are going to see the pervasive adoptions on morning in a way in the past you will type of flash very focused on the audio TPS or close that transitional no database kind of were close but we are seeing this across the adoptions on right all kind of different workloads from you know across the whole storage tier I mean you're a just absolutely especially I we believe in 2018 when the qoc which is for people cells I know when I say four people sell may not be very very sexy it is it is very sexy to see how what you open up the new opportunity when people start talking about a cold data is a very economic way to store the code data mm-hmm right so we'll open up even the object storage and you know if you if you look at the problem not just purely from dark picking my perspective you look at the overall how many people you need to you need to how many people you need to invest to just managing the drive failure and system there there and right changing their to all in a solid-state technology will have a it's a game changer in terms of the TCO calculation right i mean people have been concerned because the price has been hired but the price is falling so it seems like it's you know talk about pervasive its deadliness arabia but we're talking about the sort of big leak of the the price reduction which the toc this year his three pit / cells next year we'll see the 4p per cell and those are not a huge leap in terms of the densities and the huge leap in terms of the cost of those systems what else about where data storage is going to this year what are you back to happen they say the next 12 to 24 months to be big and data storage and software-defined storage I think that I think one of the big trend i will see is we don't look at the software-defined storage in silos software-defined storage and software-defined Network these tight integrations to allow the optimize the infrastructures and the workload placements those things work together you will see the true benefit I always and caution you know my client that don't just try to solve one piece you know you can have the best software to my storage but but if you don't have a right workload management right network and our integrations you might not feel the best result you like right right what else I mean something that just we were gonna be totally surprised about oh my gosh you know that's gonna happen a data storage where they give us the secret Vincent not just like you know what's up yeah well I think that let me go back to the first hobby with about collagen yeah cognitive is big we will see the very creative way to use the cognitive capability somebody called artificial intelligence to how do you complete change you were you were your old process for example the security right as i say that in the past well we determine securities kind of mechanical and kind of deterministic right your daughter has were right ids alright encryption keys but now you will see that people start to analyze your asses patterns you analyze your IP address analyze the correlational data and be able to you know truly protect your data even from the people that has the right authority if you will right right so babies are the most dangerous people those be so right now we are we see a lot of innovation airs you know I be inside there there's a lot of research in this area and we spend all the time to work with our client to look at what is there what is their ultimate goal to protect their to protect their most important assets and a do they what insight they are they they need to to create new value for them and I think this is I know this is not so much traditional story speaking but there are a lot of storage requirements there are our requirement lay on a storage layer to support that for example if you want to do all these correlations your storage you need to be able to be able to be able to be able to help the software calculates software to collect those metadata to manage the data efficiently and there's a performance requirement you don't want to analyze stuff take take long time to analyze it you need to have a right some of the storage technologies be able to facilitate the demands on the underperformance if you were right and you need to have a right infrastructure to be able to store them secure security at the data availability because Laura remember I early am I talking about they are they are not aware you're going to come from the correlation of the data right the data generate data okay right Ryan you you know really those data need to be all available you cannot just put the data block of data in in some mountains and in somewhere that you know those that are already to be available and we need to be able to real-time curry all this information to generate the best insight and that is a grand challenging and that's that has a lot of demand on the store systems that is that is I think that is the future of storage right there yes yep yep I visit I think you have said it's a lots a lot of good stuff i will send you the the interview when it's done we can we can tweet about it but i think i really appreciate your expertise thanks for talking with us today well thank you i really excited this is never a dull moment here I'm sure thank you thank you so

This transcript was generated automatically from the video's captions and may contain errors.