Scale by the Bay 2018: Panel III: Cloud, Edge, and Silver Lining
Recording: Scale by the Bay 2018: Panel III: Cloud, Edge, and Silver Lining
you thanks you know thanks for being with us it's a it's a great you know pleasure to host everybody this was the most diverse conference ever worked people from all around the globe and it was really really cool with that Elmo tracks who had a mini kind of blockchain engineering track we had data track basically fused into this so we had a lot of fairly cool topics and but we never did anything like this panel so this is really this is really new to us right so we basically go very deep and very broad right so the the topic of this panel is cloud agents silver lining and the idea is that as a community we do a lot of traditional cloud computing so we kind of know how to spin up a big hit abolish cluster and kind of launch a bunch of jobs and try to cessation to them and see what's going on and all like try to figure out what's going on in some web browser and so this is kind of thought computing but suddenly we're in the world but nothing is certain right so suddenly we have blockchain and we have age and all these words are showing up and we need to figure out what to do with them right so I think like as a community with a more familiar with traditional kind of you know single processor or multi processor in the box or multi computer data center so DCOs is kind of as high as we go on abstraction level all right but suddenly we get a lot of new things and and what got me interest in this I ran an engineering enterprise blockchain conference color thing trust where we can completely ban the word ICO it was completely about engineering side of this and there is result there is a whole bunch of companies which basically try to build this distributed consensus systems for trade for our kind of energy communication for like any kind of enterprise application with the needs to be an agreement in turns out an agreement can come slower sometimes you know some kind of things can not show up and so so three is there is even talk now that eventual consistency is not important that is like in the world of tomorrow Goomba no eventual consensus of everything forget about it right so basically oob islands of this can the consensus so so all of this is very cool and so we have really good representation of folks on the panel coming from different worlds so I'll just let you guys introduce each other maybe it's like for one two minutes like take I kind of tell us your take on this future work loud a nation i UT are all interconnected where are we going and kind of where you want to be where you're driving this and then we'll pick up on the themes and have a conversation in about maybe like 20 minutes before the end will open the mic to the audience so prepare your questions and really like feel free to come on the shoulder it's not just me asking questions I think this is the first time I have an excuse to look really cyberpunk but in San Francisco thankfully provided that weather but this is the first time the work that I do has been called traditional so thank you I do mostly traditional data parallel systems but I think certainly edge computing is very important and the idea of always bringing everything back to cluster to compute everything is perhaps not the future very inconvenient for someone who sells clusters but that's okay can learn new things and build new tools for people as well and I am really jet-lagged I was in Madrid yesterday I think that was yesterday so I don't remember everything else he told us to do but I can answer direct questions well I'll just pass this on to you all right hello everyone my name is Roman and I actually it's funny because you know I think I I'm trying to now escape this traditional because like yeah you guys are traditional now you know there's called big data thing you know so 2012 age is coming so that's what I'm doing I'm you know I actually have a company that I'm one of the cofounders of you know that we're doing next-generation age computing platform and I also do a lot of the open source so I'm actually working billboard of of the attacher project so if you want to know what the patch fee does in terms of projects you just walk up to me and look at this there's a couple of other things so I also happen to be on the board of directors of a battery foundation Apache Software Foundation and do a lot of things there and I used to the Big Data a lot of it so I used to be a yahoo and close era but like I said I mean it's so 2012 now all right hello everyone my name is Holly people also call me why I'm the fungus see of a Luxio so just a little bit about ourself so we believe the future about the data it's a data driven world and in a data-driven world you have data and treatment applications like motion learning analytics eSATA and also in the data driven world you have a lot of data storage persistent storage like cloud storage multi class storage on-premise storage file system object or eSATA so we believe if you go into the future you need to have a data platform and our technology essentially it's open source technology is between applications and persistent storage and essentially we aggregate and virtualize all the data from storage world and present a global namespace and unify the api's to the data-driven applications so why is it related to the edge I mean first of all you all know is a super relevant to the cloud and in fact many people run us in a multi cloud or hybrid cloud environment and also we have people running us in an agile environment today as well which I'm happy to introduce later on thank you I guess I'm why I'm the CTO constellation labs I'm building a new kind of blockchain called a DAC which is focused on a lot of this stuff and we're taking applications for it a little bit differently than how we're doing things in this space which we think is kind of applicable to the space which I won't talk about today well thank you guys so I will have maybe kind of one you know first simple question right so the we see kind of with I think the rise of blockchain brought with it very interesting set of open-source things like us a Roman explained to me a lot of AOS I looked at it and and there is a lot of open-source activity people basically want to to build built you know the distributed applications can you summarize for us you know kind of would you think of the main distinction right between dabs and kind of traditional apps and what does it mean like can these things coexist for instance I can we where do you put over there the basket like we cannot put much debt on the blockchain it's not very capable what's your thinking about where this is going you know if the apps are gonna take on how are they can occur exist of the cloud or they cannot supersede the cloud or place it can I like see them back-to-back what what do you think is gonna happen oh sure sure well traditionally people tend to use this concept called like a smart contract but I think that especially like you know as we're moving towards you know gathering a lot of data from a bunch of different you know sensors like the edge as I think we're calling it um I think there's a new opportunity to sort of use consensus as a way to validate data in general and sort of you know actually thinking about like am i passing around transactions or I meant passing around like bits of data that I want to get validated and we think that that you know that's a you know our approach there are different technologies to view the data stores for security for consistency for scalability such a and based on what we're seeing we talk to you many enterprise customers and most enterprise people today no matter what type of technology they're using the ended up today in a multi storage environment for example they typically have one primary storage could it be HDFS yes ECS ice along like GPFS many many right and they ended up with this very like messy data environment data silos people call that and they have many data duplication so then the came with a more emerging of the club from the smallest small company perspective alike a startups midsize company they may use more of the cloud for storage but what we're actually what we're seeing from the big company perspective on in the end of the day I think people realized and they understand that data is actually the most valuable asset in the company today and particularly moving forward with more applications are being innovated so because of that actually what we're seeing is that these companies they don't actually put data in a cloud so they still put the data in the unprimed on prime environment and with this they actually use the cloud as the computation bursty computation compute environment say they have several thousand machines in the on-premise environment they want to do a big model training so they want to spawn several more thousand machines so they do that in the club that's the way we see that and we believe at the for foreseeable future like say five years from now probably this would not change like it's big enterprise they would still host their data on Primus with various technologies black blockchain or the previous protocols and and they leverage the craft for the bursty workload but this is only applies to the enterprise all right well just a couple of thoughts because I think you know my colleagues are exactly right on some of those points but the way I look at it is you know our call industry is going through you know phase shifts all the time right so you know it's between distributed and decentralized all the time so I think we've just asked you know the decided distributed you know sort of phase and you know sort of went into this call centralization one right you know the big cloud in the sky is kind of like a mainframe right so it's about overdue that we actually go yet another step and you know do the full-on you know this internalization we then bring the blockchain right you know it's like if you do fall on this and realization like how does your database look well it looks kind of like a blockchain and how does your stored procedure look well it kind of looks like a smart contract and then you know some of the learnings from the distributed systems it's like what's interesting in the distributed system actually two things only storage and scheduling right you know and then you basically do scheduling you know through something like oh so there's other blockchain examples like that and you do your storage you know storage is really tricky yeah if you want to replace HDFS with the blockchain you have a really sad time or a lot of venture capital money possibly both venture capital money can fix some sad times but not all yeah I mean I think our industry is one of fashions and trends and certainly we have a new fashion trend and it'll be a party we'll have fun I don't know I don't really have anything else to add except this is gonna really suck to write the code to do Oh God debugging is the worst it's okay they pay us I hope so maybe you know al kind of maybe make it more concrete so I had everything trust look at the company Dianetics which basically wants to be the first genomics of the blockchain and so the they spent a lot of time compressing DNA right and and they put some of this on a book chain but obviously the problem with that is you know if you lose it once right you lose you know once you lick it that's it it's a it's out there and and it's a big problem so so the basically there's been a lot of thing you know like let's say we have something like like you know genomic sequence so we kind of in our traditional now kind of cloud world we can put it into spark right so this is one of the top things you know we want genomicist to do with spar because they still at the pearl stage right like if you look at bioinformatics it's threes like five years behind the what we are doing here so you know it's like they still very excited about pearl so eventually the news will reach you know that now it's it's a cluster pearl is cool man I agree no I love to write like soso but the guys just to kind of show the the timeline right so the I've been at the PI forensic is a pin and ask who is using spark nobody used spark everybody heard about it right so that's kind of where we are so so we want people to put you know DNA into spark and in the cutting edge folks do that so we have started doing that and they can cure cancer so this is kind of clear hopefully but they kind of the question goes away we don't discuss the question what happens like whether they store this DNA because we kind of assume there's traditional security model because like it's on it's on a secure disk it's under lock and key it's encrypted and we kind of assume it's okay but although we know companies lose data all the time and all the cloud companies lose data right so so so there are two questions here right so first question can we achieve the compression what do we do this kind of data so one idea is that put it in a sidecar and the clouds put a pointer on the blockchain and have some some provenance discussion you know what who can see like who can see things in the cloud now we can look at them from both places right and and and and and so the the other approach is to put everything into the blockchain then there is a question of what is you know it persists there what if there is an error right and so so we can have this new kind of distributed system right but there but it's it's permission better right it's kind of we can reason about who can read this so so what do you think of the implication of this I wonder like if you kind of valleys about it I thought this is being recorded isn't it why four five six do they want to put DNA information in the blockchain so these I think the reasoning is that you know now you have granular control over who can basically you know you can say now the the pharma company can see it if you allow it and basically you can granularly open it to the parties to the know like you know to be no to wish it will go if you know you can have a smart contract of some sort or some tactless right so so basically you will programmatically define very fine-grain who can see what and do betrayal of exes right so if you know you grant access you know to your physician or to your far like your analytical trial right so I think this is what they want they want like fine grained control and then let's say I think another use cases let's say somebody's developing experimental cancer treatment and they need to run on a batch of cancer data so it's like an ideal world you'll give this research organization permission to mine your data in a specific way which may be anonymized and so basically like you opt in into some kind of data mining through there through that instead of just uploading your data and hoping people will you know follow some rules sure I don't see how the blockchain fixes that problem so maybe I'm just not cool enough I mean personally if that was my worries I would look more at public key cryptography just start my data and then control who I give my keys to rather than putting it in the blockchain I mean if I needed to raise money in San Francisco yeah sure I definitely put that in the blockchain but I wouldn't do anything with the blood chain I just store a copy of my data there for venture capital money I don't know maybe I'm cynical but I think Spotify has a really interesting thing where they talk about their solution to GDP our compliance and how they essentially we're looking at the possibility of and this is me talking about another company's technology so of course there will be some mistakes in my recollection of it but essentially rather than trying to propagate their deletes because they don't trust themselves necessarily to propagate their deletes in a timely manner they just revoked the keys and the keys are no longer able to access the data and so the data itself is still there but the the keys are gone and eventually they'll get around to cleaning up the data as they're required to but in the immediate term the the key server will no longer granted a valet key to look at that data and I think that's maybe a much more reasonable approach but maybe I'm cynical what what do you all think I really agree I just wanted to say like I I think that that happens a lot and it kind of sucks and I mean it's like let's just put everything on the blockchain let's like put your sandwich on the blockchain thank you thank you so much blockchain Samuel yeah anyway yeah so just borrowing that it's alright but all these cases outside I mean yeah obviously also the GDP are you know kind of hip you push I think is really the one that people do touch on but I think traditional methods are really just what we want and for most cases really I think it's you just gotta really just realize that this is almost kind of I mean it's not necessarily just something about validating transactions it could be like other kinds of data you know what I mean so like this approach of really thinking about stuff in terms of like let me just use it to you know I you know put this thing on the blockchain cuz I can think of it as some new way to like be a database it's like there are different databases for certain use cases and certain tools you know I mean it's all really just about like getting that like truth that like you know validity of some kind of data center yeah yeah I guess my only comment would be speaking of you know cryptography I think we actually saw a very interesting company and the everything trust conference I believe they're basically a research you know the Russia you know some place and they came up with this really cool way of essentially re encrypting to a different key without actually decrypting so you know they have a full encryption that is possible without ever revealing you know sort of the source material so then you can store sort of the original copy and you know big cloud in the sky blockchain like you don't care and then you basically essentially issue these you know one time operational keys that would be encrypt it to a particular user and you know that's that's just fine so I would I would look into things like that you know maybe because just like blockchain by itself is probably not enough to solve some of those these cases so blockchain is a trend and it's a sort of new technology from the research community lots of paper has been published a long time ago and for the particular use case we mentioned here actually the existing technology can solve it as well and we have seen some users like a sudden a similar problem so but from the future angle it's like years ago people talk about gold cloth right so that was years ago and the first wave of going cloud like we talked we just came back from New York this week I talked to quite a few CEOs from the some the largest bank in the country so the first way of going to the cloud it was a I mean from their world is a disaster so in a sense of it's a I mean many conversation but in the end the value would not really realized there were some realized but in the end of the day it's hard to say every anything we'll all go certain things and that's a reason like at the beginning people talk about the clock clock laughs and today people talk about a hybrid cloud it's the same room similar similar thing again it's say I relieve in the end of the day things will end up somewhere in the middle like it's not necessary to be all like like one thing and the one technology it just never saw the world like that and and probably at foreseeable future won't be like that is my view actually just a quickly comment on that if I may I think it's actually really easy to be cynical so I cannot believe that I'm playing the role of the you know blockchain guy I guess that you must be saying this but I think it's actually really easy to be cynical and here's why right because you know like every new platform that comes it's actually typically really really really bad and really much worse than anything else that's already available on the market so it actually takes a lot of insight and block you know frankly to realize because 99% of that is just crap like I would absolutely agree with that but there is that 1% that is still worse than everything else and yet will be the next big thing right it's like it's blotchy and I don't know but I also wouldn't like to succumb to this temptation to say like well yeah I mean looking at all the prior experiences you know we can use stuff that's currently available sure but you know maybe we're missing something here maybe I mean well maybe like yeah I will ask you what do you think is the killer app for blockchain maybe connected with the age like what is what is the app that will clearly and demonstrably run on blockchain and it's hard to think what you know how else did can anybody think that it could have been your hat man you never know like I'll just try and I don't know you know I apps are interesting it's a you know we're we're very tied to this rudimentary rudimentary li-like tied to you know the smart contract idea but like what is it really about under the hood you know what I mean like have you guys ever really thought about like someone's gonna like have the ability to like you know maybe drive a car at some paywall and it gets paid and you won't I've seen these you know demos before but like like what is it it like goes and like sends us a little transaction to a thing and it waits it's like okay like there's totally a way that we could actually integrate this notion of cryptographic security or like just getting consensus to like you know solve this problem before you like figure out if your data is actually factual which is a real problem you know it's garbage in garbage out for a lot of people's use this so yeah we think it's gonna be like that honestly I think it's trust you know that's you know what I think you know that conference was actually really telling you know if there's anything really exciting to me about the blockchain and age at the same time it's actually thinking about the trust implications right because for both you know age right you know let's let's think that one is the first example like we actually so used by now you know for a data center to basically have you know all these security elements that just don't exist on the edge right well for one you know there's typically a guy with the machine gun in front of a datacenter right so you have a physical perimeter security then networking wise well you typically have the big Cisco firewall in front of your data center you have that network security now again even those get penetrated from time to time but at least you know you have those on the edge when even in the industrial setting right you know it's like Iranian you know nuclear fusion example you know where they get hacked by Stuxnet like on the edge anybody can walk up to your server and just plug a USB stick in right so I think you know this implication of trust is very central to the success of the edge and I think again that's why it's sort of like very tempting for me to sort of keep looking at the blockchain because if there is a single feature that blockchain has that not really has been sort of seaman that sounds before it's managing trust in a really very weird sort of multi-party competition environment right when it's not like we're all adversarial but it's not like we're all friends either so if we can hire I cannot match the two together then maybe we actually can sort of really build this cyber physical trust you know where you can actually get the data point and then trust the data point whether it came from a cyber Fitness system or a physical system so I spent a lot of time on airplanes because that's my life and there's a really depressing statistic about the percentage of fake airplane components in circulation they even found fake airplane components in circulation on whatever the president's airplanes called which is like interesting you would assume that there's like higher quality standards associated there and I think certainly Trust is one of the things where you have a bunch of different sort of independent players who don't necessarily like to talk to each other but if we can convince them with the magic of blockchain that they should talk to each other because it's the future regardless of whether or not it is the future that would be kind of cool cuz would be really nice to fly on an airplane that wasn't made of counterfeit parts I don't know I mean the counterfeit parts are really well done so like that's probably okay but I still would feel more comfortable flying on non counterfeit parts just just personally so if we can fix that problem with the blockchain I'd you know write some shitty Scala code to help yeah actually I so if if you're interested in airplane crashes there's this wonderful show oh no actually so sometimes they don't put it in the like normal in in-flight programming but if you're on one of those airplanes with DIRECTV they don't blacklisted and so um and it's like one of those like National Geographic productions so sometimes you'll just be like oh yeah I can watch a show of that airplane crashes on my airplane and also protip I found out talking to people about airplane crashes and the airplane makes them uncomfortable even if they work for one of the companies which made a faulty part which causes an airplane to crash that doesn't make it better but yeah anyways the program is called made a really good program I encourage you to check it out yeah so and there is an episode about the faulty parts that caused one of the planes to crash it's it's really exciting but like somehow most of the faulty sorry not faulty parts counterfeit parts most of the counterfeit parts are well and I've done because the people making the counterfeit parts don't want to go to jail they just want to make money so yay capitalism I guess I don't know so actually this you know this is religion now I understand better because you know I seen whispers of Airbus in charge of security and I you know he media talk to me about blockchain in that context because a lot of airplanes are leased so what happens the companies wish these airplanes when they return a lot of stuff is banked and bashed and broken and and it was not like before but there is because there's so many different parts the companies would return their planes they hide it and send them they release it to some other company internal crashes so so they want to use blockchain basically to for each component to authorize another component so all the pieces of the airplane are on the ball chain and so you cannot replace one without you know making it obvious so it's just like very simple IOT certification right so so it's actually very legitimate so good good good problem yeah and actually there's some like really hilarious ones with the aircraft leasing or the Canadian government loaned a bunch of money to some people who then like took the airplane and just didn't give it back uh-huh and like there's there's some interesting things there can government make some really high-quality investment choices yeah so maybe you know I'll ask one more question and then we'll open it there so one topic which is really interesting to me kind of which connects all these things eventual consistency so I've seen you know this ideas from I should disability systems community right because they have to take all of this stuff there's more and more of it so the onus is on them to basically process faster but still maintain transactional consistency financial consensus so suddenly you know you have to have more and more data and then I UT will start produced business or more data points and and the kind of expectation right now is like okay so I put in a giant cluster count it all and it's all should add to zero right basically like there is a notion which we can hear it from Oracle and kind of financial accounting that you know everything should add up and so it's in this Harlan harder to do and so if like we humans we do not insist on everybody green like right like we basically if you ask us local consistency right who kind of expect to be paid and fed and kind of have a house but we don't kind of you know it it's not it's like not everybody believes this or so we have homeless people because all the people you know like it doesn't propagate far right so so humans basically and we have different cultures and so forth so have agreements we have concerns which is inherently local so we kind of trying to get global planetary agreement on like global warming and still we could not get it right so so there are very few things which like we don't have the curricular warmth probably but but so so humans are not actually made for eventual or global consistency so but somehow as computer scientists we think it's cool and like we're trying hard to do it and we're kind of a spanner right like let's synchronize everything atomically and then suddenly we have magical consistency everywhere so easy and so I think it looks just an interesting issue you guys put it all in memory so you can't do that right but memory is transient and so with the age and we have a difference in there are things go offline in and they don't come back for a long time and inside of this they pop up and say okay I'm here I'm not dead take me back and know you're dead say that right so the question is so I think there is this interesting trance that more and more IOT devices will connect to this global Internet of Things and and they will have to agree on different things right like they need to be paid or they need to be tracked if their cars cars are all gonna be a IOT devices of cars like what why they described there is a mobile alliance which hopes that you know let's say you need to make a left turn and you're stressed out mom and you have kids screaming in your backseat and so you certainly make a left turn your dad like this is no it shouldn't really happen and they go Society so the left turn car will negotiate it will pay tiny tiny cancellation tokens to the cars in front of it right and they will slow down right like basically you know what was what is it worth to them to rush and kill the mom instead of just waiting right so basically like a lot of things were currently stressing out the theory is that it will go away because first cars will pay each other to accelerate and decelerate and there will be local concerns maybe it will be flocking behavior to propagate across the whole highway you know i-95 and so that all will flow right so but the question is so I'm kind of you know I'm pushing to everybody in their own sense right because I think it affects everybody do we in even spark so cuz we can assume it's all it's all there we kind of don't imagine like if so there is an RDD so it's kind of some failure mode is there so it's something failed but we're gonna think it's a bad thing let's restore it let's replay everything from the beginning let's get back well we're so should we really be doing this should we just agree that stuff is gonna you know disappear and it's okay so not a number is a good thing right like should you think of this probabilistic world where stuff is not gonna be eventually consistent it maybe consists of some degree right other different degrees of good is like you're not dead you have food you know you enjoy yourselves like there is this pyramid of needs so should we just establish everything or should we you know relinquish some of this consistency I guess I'll try to maybe just highlight it from a different angle so look I mean I think we've reached you know the level of complexity interest like everything that we do you know the industrial processes the social processes like everything that we do the right level of complexity is now such that it is very difficult for us to actually track things through supply chain right and again I use this term you know very sort of in a very flexible manner all right it's you know it's like supply chain management applies to just about anything right so I think we will see a lot of kind of like you know going back to basically trying to manage this the local behavior right and then just hope that the emergent behavior will be not something that would end up as destroying you know earth or whatever else right but I think we just not capable of managing the global sort of state of things like we've just reached that level of complexity and maybe we will at some point you know start being able to do that but I think the best option for now we you sort of try to orchestrate the local behavior and hope you know for the best so what you described with the car makes me think of like really depressing libertarian dystopian future and like oh god I don't want to build that if we build that we're bad people please no blockchain hey you know or I think we're all just trying to have fun out there and I think we're gonna try a lot and hopefully some work out you know like well we're all of it oh I don't know have some fun so so the world is always complex it's always simple as well and it's so so so from from our from our perspective probably you want to look at from the simpler way so so there are lots of data that a consistency is really hard and and people need those data to be consistent for example like your banking data and it's transactional and you be consistent if it's not consistent you always hope for a way not consider better for you I say clienti of the bank of course it's rarely happened if it happens it cracked up right so that's one side of the world and the other side of the world is that if you look at the data itself like majority of the data today more than 95% conservatively speaking or even 99% there are still like immutable data and that type of data is growing tremendously along the way and those data are video data all the data like immediate data right and the Machine logs all those type of data for those type of data like because it's immutable and in the end of the day from the consistency requirement go just require much less work so if that is a case I mean it is the case and then from the system designer angle like we always try to leverage that and use it as much as possible and make your system as simple as possible and that's the probably that's a I go to Rome Romans like point like in a sense of it's different you wanna what did you say like someone's localize it right so you always want to say try try to solve a localize the issue now localizing you categorize the problems and for this type of problem yes I mean so use a Luxio use ECS use this and that it solves already and for the other part like global global consistency if you want it pay the cost and breeding people will make it happen so let's say that's how we see it market let the market decide basically yeah yeah exactly oh I also I mean I forgot to mention I just wanted to that we also rely heavily on an eventual consistency and it's a you know kind of core to what we do as well this like in terms of time what look like how long is eventual so we don't do things in in in terms of time we do them in terms of well we just we have this finality window that we call a snapshot and just up until sort of the point where we gathered enough data to form one we try to form the you know most optimal the one that we can yeah it's all it's all really based or under a mutation system and sort of how we like self-organized let's go to folks have questions so when you think of questions I'll ask one more question that what is the currently most exciting technical problem you're personally focused on my boss told me is where less so I think one of the things that I'm most interested in a technical problem is right now sparked as a really suboptimal job say politely of handling scale down situations or loss of nodes and trying to make that less of a bad experience we do a pretty good job if you like occasionally lose nose but there's this thing where you know you might frequently be losing notes sort of very consistently and then just always recomputing on losses is no longer gonna be optimal and there's a lot of interesting things that we can do because often we'll actually have some idea that you're going to lose the note soon and we could do smarter things there and I think it's a fun problem to solve because it lets me like screw on with a bunch of Scala code and hang out with the kubernetes people who have nicer stickers so I guess what I'm mostly sort of preoccupied these days is it's not even a problem it's sort of these you know kinda like fundamental realization that if you look into the more traditional data center cloud it's actually got really boring right you know it's like everything is kind of like not really cool anymore like you know the server's don't really get to be cool like you know we've sort of solved most of the things you know it's like the data center architecture is sort of settled like the computing in a data center really stopped being cool like it used to be cool I mean like some time ago but it's not anymore and I think you know that's by the way why you know companies like HP are struggling and you know almost all you know everybody is buying in a hallway service or something but I think you know with an edge we actually have this really nice chance of essentially making you know computing cool again in the same way that's sort of like it got to be cool again back in the 80s right you know Sun Microsystems was actually founded on these principles like let's make computing cool again so I'm actually thinking about how we can't just re basically edge gives us this possibility of getting rid of you know at least some amount of legacy right you know we no longer have to think about you know some of the issues that you know make data center computing not cool and how can we use that chance you know because that's once in a lifetime chance right so don't screw it up and actually come up with an architecture that would make computing cool again and I really mean the m2 end architecture so that's why for example I'm so excited about risk five because I think risk five has a chance of being to the edge what arm has become to mobile computing right and with risk five and you know companies like sci-fi we actually do have that final chance it's like it feels like being back at Sun again right we can actually design everything from the crowd up like even the hardware is investable again like try investing in CPU design you know five ten years ago like you would be left out of the room but like with all of these emerging trends from the edge and you know some of the other stuff like even the conquer is investable against only let's use that chance to build a really cool system you know from the hardware all the way up because we are going to get stuck with it for the next 10 15 years so very huge disk as you know so uh there are there are many cool stuff and there are many interesting stuff i like like blockchain like AI at computing all the type of cool technologies but my personally i spend most of my time of course I'm former Luxio and I'm super excited about the thing we are doing Alex yourself in a sense of the way as we see the world is about the data it's it is about the data in the end of the day ai is an application of the data like analytics it's an application that data like IOT you get you you you general master data you need to leverage the data you cannot make it wrong like a Tom's driving is a lot of data in the end of the day so then in that world would be better well I would be very interesting or make the developer's life much easier that if you have a system that you can just put the data in don't worry about anymore and later on you want to take a bit you want to retrieve the data back and the system will guarantee you like you can you can make it happen and you do not in to worry about the details it's like when you are using iPad or when you're using a Mac why do you need to worry what CPU you are using or memory you are using you do not to care you just need to make sure you're sending email email sent your experience is nice and when you are watching on video Tata's smooth hopefully its 4k as well that's that's that's the thing in the end of the day in a data world like say if we can't shoot that state that'd be beautiful and that's the thing we're working on and and we hope we'll realize it one day so that's that's the what I personally like spending time oh cool yeah I'm gonna break the mold actually and oddly enough keep it group drive I've been tasked or I'm and doing the oddest thing I never really thought I'd ever do which I'm like do making something called token Alex so I'm tying together some kind of how do I put it like a model that will actually like go over in sort of the way in which you know distribution will flow through my network as it kind of kicks off and it was like something that had ended up being like oddly connected to how I like built this thing so it was kind of interesting and I like of having to solve this like oh all right like a math problem but like at least like make some plots yeah thank you do we have any questions if not then I will say this is the time this is the time right so it's been a long productive day let's thank all the panelists [Music] and please do not forget the CFP for scalable debate trends and it opens May 1st and we basically get in the month of May already like three times more talks and I can fit so please be like quickly with your best talk so my first mark your calendar submit your best talk get into the program we'll meet you back here or Twitter November 2019 thanks a lot [Applause]