 So now we're into the next afternoon session. This sort of represents the time that Mike moved from Berkeley, again, after being conjoaled by his wife, coming here to MIT. So we're going to have talks from Stan Zodanic, talking about streaming. We're going to have talks from Sam that discusses impact, Mike's impact in the New England Davis community. Then we'll have Andy Palmer talking about Seastore and Column Stores and Vertica. And then we'll finish up with Magda and myself talking about what it's like to be a student. So one last final answer. If you have not turned in your Stonebrick or quiz answered questions, this is your last opportunity. You can pass it to the side when we come get it. After this, we'll be telling results and we'll announce the winners afterwards. So at this point, I'm happy to introduce Stan Zodanic. Do we have any announcements? Anyway, Thomas, do we have any announcements? Oh, um. He's a good man. I'm sorry, what was the question? Thanks. OK. Yeah, he was looking up online. So we have an announcement from the MIT parking services. There's a car outside that needs to be moved. They were going to tow it. They told me to do this after lunch. The license plate, it's a New Hampshire. It's Quebec, Unicorn, Echo, Lima, 4, Lima, India, Foxtrot, Echo. If this is your car, please move it because they're otherwise going to tow it. Thomas, is it on the website? Yeah, OK. Is this anybody's car here? Please, please move it. OK. So with that, we're taking things over to my advisor, Stan Zodanic, who is a good man. He showers daily. He's never physically abused me. OK. Physically, yes. Yeah, make sure I'm all in order. I have to apologize. I'm getting over a cold, so I have a little bit of laryngitis. So hopefully my voice will last. This is the... OK. Well, it's great to be here. We're all having a good time. And when Sam mentioned that we were going to have a party of sorts for Mike, my imagination went wild. I thought, party? Wow. I know we'll have DeWitt jump out of a cake. But, you know, that's just my way of thinking. It turns out that if you hand over the duty of arranging a party to a bunch of academics and researchers, they basically redesign SIGMOD. And that's OK, because it's fun. We all like SIGMOD. But it wasn't what I had in mind. OK. Let me just see if I can get this to work. There we go. So it turns out I've known Mike for a very long time since the mid-80s. And we'll get to this move to the East Coast in a minute. But I'd first like to digress to the early days when I knew Mike. And basically back in the 80s or so, the database community likes to do these discussions, these two camps, these wars, right? So there was cortisol in relations. And then there was a time when there was relations versus objects. And this is when I first met Mike. I was an object guy. You can guess what he was. So we would find ourselves on various panels or head-to-head across the aisle from each other. And so these discussions would get a little heated sometimes. And basically, I would say things like that. And Mike would respond with that. And so I was a young faculty member at the time. I was a little intimidated. And so I figured that Mike hates me, which he may well have, but he didn't show it. There's some guy in the front here that's jumping up and down. I don't know what that's like. By the way, I'm not sure what's going on with Mike's head. But anyway, there was this guy here. The audience didn't often agree with us either. So there's this guy who basically didn't like either of us, thought we were all wrong. And he said that Datalog will set you free. It wasn't until much later that I figured out this was Jeff Almond. And so this was the context of... And then, miracle of miracles, we became friends. He didn't really hate me. And the major life lesson here from Mike Stonebreaker, or about Mike Stonebreaker, is that he likes a good fight. He's not going to hate you forever. He's actually going to appreciate that. So I hope that's true. So basically around 2000, as you all know, Mike and the family got on an airplane, and they flew to Boston, I guess. I got a little carried away with clip art here. Sorry about that. And I knew he had moved here. He moved to New Hampshire, and that was a good thing. And then one day, my phone rang. And that's how phones look back then, by the way. My phone rang and it was Mike. He was in New Hampshire, and he was feeling kind of lonely. He was looking for a playmate, and MIT, which was much closer, didn't really have much in the way of databases going on, I'm sorry to say. So he suggested maybe we should get together, and maybe we would do something together, have a project together. So I went into MIT. We met on the fifth floor of 545 Tech Square, and we talked for the afternoon. And that kind of looked like that. And the miracle that happened is we concocted a project. And the project was about streaming. That's what I'm supposed to talk to you about today. Now, many of the slides that I'm going to show you in the next bunch come from slides that are more than 10 years old. It was really kind of fun to go through these old slide decks that you hadn't seen in 10 years and realize what you were trying to sell. Please, they're old. These are historical documents. These are not new things to be criticized. This is an actual slide from the old days. Hard to say old days. It seems like just yesterday. But basically the thing on the left was how the world used to work, and we're going to change the world in the streaming context to look like the thing on the right. And I think it was Mike who came up with the term human active system passive, system active human passive. That's how I remember it. And this allowed basically the data to trigger events in the CQ engine and answers would go to the user. The data was the active party. And so that led to Aurora. And this is the slide, the sort of cartoon slide that we always showed to tell you what Aurora was about. Basically Aurora took a bunch of input streams on the left, passed them through a bunch of boxes or transformation engines or SQL operators if you wish. And out the other end would come data that was useful to these applications. And you'll notice, let's see if I can do this, way up here, is that visible? A little green dot. Way up here there's this thing called a profile. And that essentially expressed the quality of service that you wanted out of your system. And by the way, down here there was some tip of the hat to historical data. But we'll say more about that in a minute. Okay, so the operators were your typical friends, things that you might have put in a relational query plan. And we added a few extras like WSort and resample, no need to go into that now. And also user defined procedures or functions. Okay, let's this quality of service thing all about. There were things that the user would specify that would tell you what kind of results you wanted to see. So for example, the one that we spent most of our time on is this idea that delay is, the more delay the less happy you are. So basically think of quality of service as happiness. Okay, I'm going to go through this pretty quickly. I don't want to get too technical. Think of this as, do you remember Saturday Night Live? There was a character called Father Guido Sarducci. So this is five minute. He had this thing called five minute university. Five minute economics was a biolo sell a high. This is a five minute university on streaming just to set the context. So there were other kinds of QoS that you might want to specify. There may be certain values that are more important. So that was how you would deal with that. Turns out in practice we really only dealt with this one here. This was a way to basically get the user to tell you what they wanted. Okay. So another, something that we studied was load shedding. Nesime Tatbo, who's here somewhere. This was for a PhD thesis. And basically this was a way to optimize queries by eliminating the bottleneck by throwing tuples away. So this was an early attempt to do approximate query processing. It's just not throwing tuples away. The result isn't particularly accurate. So you get an approximate result. Okay. But it gets worse because the network is not always your friend. Things can be delivered in, be delayed arbitrarily or can be delivered out of order. And basically that kind of stuff plays havoc with these constructs that we call windows or computing aggregates. Okay. So our solution was to basically hand this problem back to the user. And we would basically let the user tell you what they wanted. For late tuples that arrived, we had this notion called timeout. And timeout basically said that after a certain amount of time goes by, you just close the window. And you might miss something that's coming later, but it's better to not wait forever blocking for something that might never come. We also had this idea of we called Slack. And this was essentially saying that if a tuple came early, then you would wait a certain amount of time before you'd finally decided that it was time to switch gears. Okay. So that was sort of the high points of Aurora. Borealis was Aurora operating on multiple nodes. And so this is a picture that we used a lot. I think Orr came up with this. Are you here Orr? Yeah, there he is in the back. This was Orr's picture of what Borealis was supposed to look like. Okay. So here are the main, I told you that to tell you this. Here are the main technical problems that we addressed, basically push-based processing, these notions of windows, which was a way to take an unbounded data stream and chop it up into pieces that you could compute with. Load shedding, I mentioned that, dealing with disorder, I mentioned that. And this idea that nothing, no customer that we ever talked to had a pure streaming application. They all required storage of some kind. And so we had a lot of discussion about how to build storage into the streaming system. And I think we might not have actually gotten it right. So we're doing some work now to try to get it right, maybe incorporate streaming into a transactional context. Okay. Borealis addressed things like load balancing and high availability. And then there was this benchmark. This was an amazing story. So I remember walking through the Stanford campus, talking to Mike after an all-day meeting on streaming. And Mike said, you know, the problem is that we don't have a benchmark that people would salute. And so he went home. And over the weekend, he came up with this benchmark that had to do with cars and highways and the congestion of cars on the highway would determine the tolls. And back then, people talked about this, but it really didn't exist. And Mike, because he's Mike, turned this into a community-accepted benchmark. I don't know how he did that, but he did. And so you'll hear people talk about linear road. That's the linear road benchmark. Mike made it up in two days. And even though it didn't correspond much to reality, people believed it. Which is pretty cool. Pretty cool. So here's a photograph of the early Aurora Borealis team. You probably recognize some of the faces there. Others, you may not. I don't know what happened to Or and Mitch. You weren't there that day, but Or, Chattentumel and Mitch Triniac were definitely part of this group. One of the most amazing things about this project is we got a lot of students working on it. We had this group, plus a bunch of other ones, when it was time to commercialize it. We had to make a list of all the students who had worked on it for the purpose of distributing the equity. And I think there were about 20 students on that list. And at one of the conferences, Sigma, I think it was, when we demonstrated the system, Paul Larson came up to me and said, I have one question. How did you keep such a large group, 20 people, or 20 people, together on this one project in a university setting? And I guess the answer that I would give is that Mike is such an inspirational guy, and I mean this with all my heart, that he was able to keep the juice flowing in the group together. And that's pretty impressive. So that's the early team. Then we went to demonstrate it at Sigma, what was this, 06? 05. Sigma 05. And Mike looked around and noticed that the Berkeley guys had t-shirts that said Berkeley database group. And the Stanford guys had t-shirts that said Stanford database group. And we needed something to basically say who we were. So it was too late to get t-shirts made. So we went to the local mall and we bought these hats. You see the hats? They say Aurora Borealis. And because I'm a pack rat, I still have mine. And thank you, thank you. I won't say who was modeling it there, but you can take a guess. I think you can see some whiskers down here. But that's okay. Anyway, we wore these hats. Here was the demo room. That's Nessie Mae, that's Eddie, and I don't know who the other people are, but it was quite well received. And there's Nessie Mae giving the presentation. Oh, this was, I'm sorry, this was taken at Sigma 05. I think it was 05. Oh, it says it right there. Sigma 05, we won the best demo award. Yay. Thank you very much. And here's Mike getting a demo from Yanif Ahmad. And he looks pretty happy. But I think something's going on in his head. Something is percolating this beyond just making a good show at Sigma, right? And here's another picture, and you can see he's pretty happy there. I actually just included this because it's a good picture of Mike. But he's thinking about something. And what could it be? Well, he wanted to build a company. And the company was based around Aurora, essentially. And you'll notice here that it says, formerly Grassy Brook. Well, what's that all about? Well, we needed a name for the company. And Mike has a house up on Winnipeg Saki. And it was on Grassy Pond Road. They said, well, Grassy Pond, there's a good name. But, you know, ponds just kind of sit there. And streams move. We needed Grassy something. What moves? Well, Grassy Brook. That sounds good. So Grassy Brook was the first name of Streambase. And it was named after the street where Mike and family live. OK, so this was, by the way, this is also a slide from a deck. I didn't just make this up. Now, when you get into a company, and this is what I've learned, you have to worry about things beyond just the technology. You have to worry about marketing, for example. So this was my first real immersion in the wonderful world of marketing. And this was one of the slides from our pitch deck. You'll notice that the stick figures that I had on the previous slide have changed. Now we have powerful images like bullet trains, right? Yeah, that's what we're talking about. And the words are all superlatives, right? They're all very strong language. So that's what the marketing people teach. Now, this next one, I swear this is the truth. You'll remember the picture where I had the streams coming in on one side and the answers coming out the other side. Well, our genius marketing people translated that into this. This is for real. And, you know, what came out the other end wasn't streams, it was just, you know, goodness and nirvana and all that good stuff. Okay, so I was poking through some other slides from these early pitch decks and I found one that had Mike's footprints all over it. And that was this. And, you know, I don't understand what it means exactly. I'm not sure what an EIA system is, although I looked it up once. And here's the famous quad chart coming back and there was no predetermined, this was just, it's just speaks of Mike, right? And so, yeah. And so the culmination of all of this marketing stuff, in my opinion, was something that was called the Da Vinci Coder. And I don't know how many of you people know about this, but I'm going to try here to hopefully this will work. Ah, why am I getting no sound? No. Where's that project I needed yesterday? Does it sound like I care what you think? Managers. The system is down. You have any idea how much this is costing us every second? And it's code. I mean, this code is spaghetti, my friend. You can actually read code. And code it. Stop working on that urgent project and start working on this urgent project. I wish we had time to see the whole thing. The computers react instantly to real-time events without custom coding. Come on, let's do this. You can't do this to me. No. But now the tide has turned. So if we fast forward a little bit here... We're in C++ or Java, she's... See that again. Hurry with no latency. Hurry with no latency. Well, you might recognize that gentleman. And so not only is Mike a brilliant computer scientist, but he's an actor. We filmed this at a studio somewhere south of here, and it was pretty amazing. And just to show you that he likes to share the wealth, if we fast forward a little more... The code still. Show me the money. Show me the money. He tells me we're getting close. Now where? I will use my superior French cryptography. I'm not sure you saw that, but... Couldn't keep up with the data rates. Code still. That's me. So I've learned a lot of things from Mike. My life has been changed radically because of knowing Mike. But really, I mean, he shared his moment in the sun on the big screen. And I'll be forever grateful that Mike... Harry Balakrishnan was also a monk in this movie. I'm not sure where he shows up, but trust me. Okay, so... Let's see. I want to go back to here. Okay. Yeah, I don't know. Something's going wrong here. Maybe... Let's see if this works. Okay, good. So my first story here is Mike... When we finally decided we wanted to commercialize this, we went out to some VCs along 128, and I was with Mike, and we got about halfway through the pitch, and they had their checkbooks out. It was really pretty amazing, and it wasn't because of the pitch, it was because of Mike. They trusted him. They called him a serial entrepreneur, and I had just never seen anything quite like that. Once the company got going, we managed to get an audience with CTOs and all the big financial firms in New York City, so I spent a lot of time on trains going to meetings with CTOs to explain the technology, and I learned something very important there, which is that the end users are a lot smarter than university people sometimes like to pretend. They can tell you, that's not a problem for us. Oh, but that is, or but you're missing this. So I always went home from these countless train trips a little bit smarter than when I went, so my hat's off to end users, and oh, by the way, stream-based had hats too. The whole thing was an amazing experience. I learned a lot about business. I learned a lot about commercial world, and it was really much better than going to Harvard Business School for a couple of years, and none of this would have been possible without Mike, so I seriously want to thank him publicly right now. How did this happen? I think I missed a whole bunch of slides in the middle. All right, so in any case, Mike has a hidden talent. I don't know how many of you know this. He plays the banjo, and you might say, what, the banjo? The banjo, everybody knows the banjo is kind of a strange instrument. It's not very popular. It sounds horrible. It's kind of in the same equivalent class with the bagpipe, right? So why would Mike, a brilliant computer scientist, and now we know an actor as well, why would he want to play the banjo? And if you were in the bluegrass world for any amount of time, as I have spent some time in the bluegrass world, you know that all the jokes are about the banjo players, and so just to give you a sense of how this stereotype works, and by the way, I'm not accusing Mike of any of this, right? So here we go. What's the difference between a banjo player and a large pizza? Does anybody know? It's really good. A large pizza can feed a family of four. And we know Mike has no problem feeding a family of four. So that's not it. How about this? How can you tell if the concert stage is level? Mike is drooling out of both sides of his mouth. Oh. Come on. That's exactly right. Do you play the banjo? No, I play drums. I get it. I get it. Okay. What's that? Pull yourself together, Mark. It would suggest to me that you're in fight if Mike goes and drools out of both sides. Well, I've never seen Mike drool at all. But anyway. And it gets worse because the mainstream media has picked up on this. So New Yorker Magazine is one of my favorite cartoons of all time. But as you can see, banjo players get no respect. No respect at all. So why would you want to play the banjo? I'm just a good question, right? And I asked myself that about Mike. And, you know, if you look around these days, there were some pretty high-profile banjo players in the world. You all recognize these folks, and they're both playing the banjo, looking pretty good there. And I think that Mike just wants a piece of the action. So he took up the banjo and organized a bluegrass jam among a few of his friends, including myself. And every two to three weeks, we get together in the stream-based conference room. There's a little link to the previous stuff, right? Stream-based conference room. Started it back in the Vertica days, and we even had one public appearance. Yep, that's true. And anybody want to guess what we call ourselves? That's good. We're going to remember that. Do you want to guess? It's resumed aboard. Resumed aboard? Yeah, we call ourselves. And that's it. So here's a picture of Mike with his banjo. I think this may be the only picture on the planet of Mike holding his banjo. So this is rare stuff. He has more than one banjo. It's true. But then again, people who are into this music tend to have more than one of what they play. I'm not going to tell you how many mandolins I have. So this, Mike, I'm going to introduce you to the rest of the band here. It's not really a band. It's a jam session. This is JR Robinson. He was a former Vertica engineer. And by the way, the original idea was we were going to smuggle a banjo in here, right? And we were going to surprise Mike and say, okay, here you go. Let's play something. But it turns out that JR Robinson, the lead singer and rhythm guy, couldn't make it, so that idea went out the window. But hang on for a minute because there's more to be told. Robert Hoffman, former stream-based engineer. He comes most of the time, not all the time, because he's a very much-in-demand symphony bass player. And so he's been sitting in with us playing this, I don't know, bluegrass music. And he's just getting used to the idea that bluegrass musicians don't use sheet music. And so he's just kind of adapting to that. But he's been a good addition to the group. That's me. That's that funny-looking instrument that I have. That's a mandolin, as conceived by the Gibson Company in the 1920s. And occasionally we have this guy sitting with us, Dave Reiner. Does anybody here know Dave Reiner? Yeah, some of you folks do. And just for those of you who don't know him, if you're worried that maybe he doesn't have the database cred to join this group, I think he's got some papers. He was doing parallel computing actually before a lot of us were. So he's okay. He's a very good fiddle player. He forces us to up our game. And he comes every now and then. And just to pull back the curtain a little bit. I hope this works. Let's see. Yeah, right. But it's not showing up. The cursor's not showing up on my screen. A little standard called Crypto Creek. In case you don't recognize it. A little singing lead. That was recorded, by the way, of my iPhone. I turned it on and surreptitiously grabbed a little bit of what we were doing. And as you can see, Mike's a pretty good banjo player. Another talent, really uncomputer scientist, actor, musician. Wow. So I think that's the end. I'm not sure why. There's a few slides I missed in the middle. But I think that'll, that'll suffice for now.