>> Paige: Hello, everybody, and thank you for joining us today for the virtual Vertica BDC 2020. Today's breakout session is entitled, Vertica next generation architecture. I'm Paige Roberts, open social relationship Manager at Vertica, I'll be your host for this session. And joining me is Vertica Chief Architect, Chuck Bear, before we begin, I encourage you to submit questions or comments during the virtual session. You don't have to wait, just type your question or comment, in the question box that's below the slides and click submit. So as you think about it, go ahead and type it in, there'll be a Q&A session at the end of the presentation, where we'll answer as many questions, as we're able to during the time. Any questions that we don't get a chance to address, we'll do our best to answer offline. Or alternatively, you can visit the Vertica forums to post your questions there, after the session. Our engineering team is planning to join the forum and keep the conversation going, so you can, it's just sort of like the developers lounge would be in delight conference. It gives you a chance to talk to our engineering team. Also, as a reminder, you can maximize your screen by clicking the double arrow button in the lower right corner of the slide. And before you ask, yes, this virtual session is being recorded, and it will be available to view on demand this week, we'll send you a notification, as soon as it's ready. Okay, now, let's get started, over to you, Chuck. >> Chuck: Thanks for the introduction, Paige, Vertica vision is to help customers, get value from structured data. This vision is simple, it doesn't matter what vertical the customer is in. They're all analytics companies, it doesn't matter what the customers environment is, as data is generated everywhere. We also can't do this alone, we know that you need other tools and people to build a complete solution. You know our database is key to delivering on the vision because we need a database that scales. When you start a new database company, you aren't going to win against 30 year old products on features. But from day one, we had something else, an architecture built for analytics performance. This architecture was inspired by the C-store project, combining the best design ideas from academics and industry veterans like Dr. Mike Stonebreaker. Our storage is optimized for performance, we use many computers in parallel. After over 10 years of refinements against various customer workloads, much of the design held up and serendipitously, the fact that we don't store in place updates set Vertica up for success in the cloud as well. These days, there are other tools that embody some of these design ideas. But we have other strengths that are more important than the storage format, where the only good analytics database that runs both on premise and in the cloud, giving customers the option to migrate their workloads, in most convenient and economical environment, or a full data management solution, not just the query tool. Unlike some other choices, ours comes with integration with a sequel ecosystem and full professional support. We organize our product roadmap into four key pillars, plus the cross cutting concerns of open integration and performance and scale. We have big plans to strengthen Vertica, while staying true to our core. This presentation is primarily about the separation pillar, and performance and scale, I'll cover our plans for Eon, our data management architecture, Mart analytic clusters, or fifth generation query executer, and our data storage layer. Let's start with how Vertica manages data, one of the central design points for Vertica was shared nothing, a design that didn't utilize a dedicated hardware shared disk technology. This quote here is how Mike put it politely, but around the Vertica office, shared disk with an LMTB over Mike's dead body. And we did get some early field experience with shared disk, customers, well, in fact will learn on anything if you let them. There were misconfigurations that required certified experts, obscure bugs extent. Another thing about the shared nothing designed for commodity hardware though, and this was in the papers, is that all the data management features like fault tolerance, backup and elasticity have to be done in software. And no matter how much you do, procuring, configuring and maintaining the machines with disks is harder. The software configuration process to add more service may be simple, but capacity planning, racking and stacking is not. The original allure of shared storage returned, this time though, the complexity and economics are different. It's cheaper, even provision storage with a few clicks and only pay for what you need. It expands, contracts and brings the maintenance of the storage close to a team is good at it. But there's a key difference, it's an object store, an object stores don't support the API's and access patterns used by most database software. So another Vertica visionary Ben, set out to exploit Vertica storage organization, which turns out to be a natural fit for modern cloud shared storage. Because Vertica data files are written once and not updated, they match the object storage model perfectly. And so today we have Eon, Eon uses shared storage to hold Vertica data with local disk depot's that act as caches, ensuring that we can get the performance that our customers have come to expect. Essentially Eon in enterprise behave similarly, but we have the benefit of flexible storage. Today Eon has the features our customers expect, it's been developed in tune for years, we have successful customers such as Redpharma, and if you'd like to know more about Eon has helped them succeed in Amazon cloud, I highly suggest reading their case study, which you can find on vertica.com. Eon provides high availability and flexible scaling, sometimes on premise customers with local disks get a little jealous of how recovery and sub-clusters work in Eon. Though we operate on premise, particularly on pure storage, but enterprise also had strengths, the most obvious being that you don't need and short shared storage to run it. So naturally, our vision is to converge the two modes, back into a single Vertica. A Vertica that runs any combination of local disks and shared storage, with full flexibility and portability. This is easy to say, but over the next releases, here's what we'll do. First, we realize that the query executer, optimizer and client drivers and so on, are already the same. Just the transaction handling and data management is different. But there's already more going on, we have peer-to-peer depot operations and other internode transfers. And enterprise also has a network, we could just get files from remote nodes over that network, essentially mimicking the behavior and benefits of shared storage with the layer of software. The only difference at the end of it, will be which storage hold the master copy. In enterprise, the nodes can't drop the files because they're the master copy. Whereas in Eon they can be evicted because it's just the cache, the masters, then shared storage. And in keeping with versus current support for multiple storage locations, we can intermix these approaches at the table level. Getting there as a journey, and we've already taken the first steps. One of the interesting design ideas of the C-store paper is the idea that redundant copies, don't have to have the same physical organization. Different copies can be optimized for different queries, sorted in different ways. Of course, Mike also said to keep the recovery system simple, because it's hard to debug, whenever the recovery system is being used, it's always in a high pressure situation. This turns out to be a contradiction, and the latter idea was better. No down performing stuff, if you don't keep the storage the same. Recovery hardware if you have, to reorganize data in the process. Even query optimization is more complicated. So over the past couple releases, we got rid of non identical buddies. But the storage files can still diverge at the fifth level, because tuple mover operations are synchronized. The same record can end up in different files than different nodes. The next step in our journey, is to make sure both copies are identical. This will help with backup and restore as well, because the second copy doesn't need backed up, or if it is backed up, it appears identical to the deduplication that is going to look present in both backup systems. Simultaneously, we're improving the Vertica networking service to support this new access pattern. In conjunction with identical storage files, we will converge to a recovery system that instantaneous nodes can process queries immediately, by retrieving data they need over the network from the redundant copies as they do in Eon day with even higher performance. The final step then is to unify the catalog and transaction model. Related concepts such as segment and shard, local catalog and shard catalog will be coalesced, as they're really represented the same concepts all along, just in different modes. In the catalog, we'll make slight changes to the definition of a projection, which represents the physical storage organization. The new definition simplifies segmentation and introduces valuable granularities of sharding to support evolution over time, and offers a straightforward migration path for both Eon and enterprise. There's a lot more to our Eon story than just the architectural roadmap. If you missed yesterday's Vertica, in Eon mode presentation about supported cloud, on premise storage option, replays are available. Be sure to catch the upcoming presentation on sizing and configuring vertica and in beyond doors. As we've seen with Eon, Vertica can separate data storage from the compute nodes, allowing machines to quickly fill in for each other, to rebuild fault tolerance. But separating compute and storage is used for much, much more. We now offer powerful, flexible ways for Vertica to add servers and increase access to the data. Vertica nine, this feature is called sub-clusters. It allows computing capacity to be added quickly and incrementally, and isolates workloads from each other. If your exploratory analytics team needs direct access to the source data, they need a lot of machines and not the same number all the time, and you don't 100% trust the kind of queries and user defined functions, they might be using sub-clusters as the solution. While there's much more expensive information available in our other presentation. I'd like to point out the highlights of our latest sub-cluster best practices. We suggest having a primary sub-cluster, this is the one that runs all the time, if you're loading data around the clock. It should be sized for the ETL workloads and also determines the natural shard count. Additional read oriented secondary sub-clusters can be added for real time dashboards, reports and analytics. That way, subclusters can be added or deep provisioned, without disruption to other users. The sub-cluster features of Vertica 9.3 are working well for customers. Yesterday, the Trade Desk presented their use case for Vertica over 300,000 in 5 sub clusters running in the cloud. If you missed a presentation, check out the replay. But we have plans beyond sub-clusters, we're extending sub-clusters to real clusters. For the Vertica savvy, this means the clusters bump, share the same spread ring network. This will provide further isolation, allowing clusters to control their own independent data sets. While replicating all are part of the data from other clusters using a publish subscribe mechanism. Synchronizing data between clusters is a feature customers want to understand the real business for themselves. This vision effects are designed for ancillary aspects, how we should assign resource pools, security policies and balance client connection. We will be simplifying our data segmentation strategy, so that when data that originate in the different clusters meet, they'll still get fully optimized joins, even if those clusters weren't positioned with the same number of nodes per shard. Having a broad vision for data management is a key component to political success. But we also take pride in our execution strategy, when you start a new database from scratch as we did 15 years ago, you won't compete on features. Our key competitive points where speed and scale of analytics, we set a target of 100 x better query performance in traditional databases with path loads. Our storage architecture provides a solid foundation on which to build toward these goals. Every query starts with data retrieval, keeping data sorted, organized by column and compressed by using adaptive caching, to keep the data retrieval time in IO to the bare minimum theoretically required. We also keep the data close to where it will be processed, and you clusters the machines to increase throughput. We have partition pruning a robust optimizer evaluate active use segmentation as part of the physical database designed to keep records close to the other relevant records. So the solid foundation, but we also need optimal execution strategies and tactics. One execution strategy which we built for a long time, but it's still a source of pride, it's how we process expressions. Databases and other systems with general purpose expression evaluators, write a compound expression into a tree. Here I'm using A plus one times B as an example, during execution, if your CPU traverses the tree and compute sub-parts from the whole. Tree traversal often takes more compute cycles than the actual work to be done. Especially in evaluation is a very common operation, so something worth optimizing. One instinct that engineers have is to use what we call, just-in-time or JIT compilation, which means generating code form the CPU into the specific activity expression, and add them. This replaces the tree of boxes that are custom made box for the query. This approach has complexity bugs, but it can be made to work. It has other drawbacks though, it adds a lot to query setup time, especially for short queries. And it pretty much eliminate the ability of mere models, mere mortals to develop user defined functions. If you go back to the problem we're trying to solve, the source of the overhead is the tree traversal. If you increase the batch of records processed in each traversal step, this overhead is amortized until it becomes negligible. It's a perfect match for a columnar storage engine. This also sets the CPU up for efficiency. The CPUs look particularly good, at following the same small sequence of instructions in a tight loop. In some cases, the CPU may even be able to vectorize, and apply the same processing to multiple records to the same instruction. This approach is easy to implement and debug, user defined functions are possible, then generally aligned with the other complexities of implementing and improving a large system. More importantly, the performance, both in terms of query setup and record throughput is dramatically improved. You'll hear me say that we look at research and industry for inspiration. In this case, our findings in line with academic binding. If you'd like to read papers, I recommend everything you always wanted to know about compiled and vectorized queries, don't afraid to ask, so we did have this idea before we read that paper. However, not every decision we made in the Vertica executer that the test of time as well as the expression evaluator. For example, sorting and grouping aren't susceptible to vectorization because sort decisions interrupt the flow. We have used JIT compiling on that for years, and Vertica 401, and it provides modest setups, but we know we can do even better. But who we've embarked on a new design for execution engine, which I call EE five, because it's our best. It's really designed especially for the cloud, now I know what you're thinking, you're thinking, I just put up a slide with an old engine, a new engine, and a sleek play headed up into the clouds. But this isn't just marketing hype, here's what I mean, when I say we've learned lessons over the years, and then we're redesigning the executer for the cloud. And of course, you'll see that the new design works well on premises as well. These changes are just more important for the cloud. Starting with the network layer in the cloud, we can't count on all nodes being connected to the same switch. Multicast doesn't work like it does in a custom data center, so as I mentioned earlier, we're redesigning the network transfer layer for the cloud. Storage in the cloud is different, and I'm not referring here to the storage of persistent data, but to the storage of temporary data used only once during the course of query execution. Our new pattern is designed to take into account the strengths and weaknesses of cloud object storage, where we can't easily do a path. Moving on to memory, many of our access patterns are reasonably effective on bare metal machines, that aren't the best choice on cloud hyperbug that have overheads, page faults or big gap. Here again, we found we can improve performance, a bit on dedicated hardware, and even more in the cloud. Finally, and this is true in all environments, core counts have gone up. And not all of our algorithms take full advantage, there's a lot of ground to cover here. But I think sorting in the perfect example to illustrate these points, I mentioned that we use JIT in sorting. We're getting rid of JIT in favor of a data format that can be treated efficiently, independent of what the data types are. We've drawn on the best, most modern technology from academia and industry. We've got our own analysis and testing, you know what we chose, we chose parallel merge sort, anyone wants to take a guess when merge sort was invented. It was invented in 1948, or at least documented that way, like computing context. If you've heard me talk before, you know that I'm fascinated by how all the things I worked with as an engineer, were invented before I was born. And in Vertica , we don't use the newest technologies, we use the best ones. And what is noble about Vertica is the way we've combined the best ideas together into a cohesive package. So all kidding about the 1940s aside, or he redesigned is actually state of the art. How do we know the sort routine is state of the art? It turns out, there's a pretty credible benchmark or at the appropriately named historic sortbenchmark.org. Anyone with resources looking for fame for their product or academic paper can try to set the record. Record is last set in 2016 with Tencent Sort, 100 terabytes in 99 seconds. Setting the records it's hard, you have to come up with hundreds of machines on a dedicated high speed switching fabric. There's a lot to a distributed sort, there all have core sorting algorithms. The authors of the paper conveniently broke out of the time spent in their sort, 67 out of 99 seconds want to know local sorting. If we break this out, divided by two CPUs and each of 512 nodes, we find that each CPU so there's almost a gig and a half per second. This is for what's called an indy sort, like an Indy race car, is in general purpose. It only handles fixed hundred five records with 10 byte key. There is a record length can vary, then it's called daytona sort, a 10 set daytona sort, is a little slower. One point is 10 gigabytes per second per CPU, now for Verrtica, We have a wide variety ability in record sizes, and more interesting data types, but still no harm in setting us like phone numbers, comfortable to the world record. On my 2017 era AMD desktop CPU, the Vertica EE5 sort to store about two and a half gigabytes per second. Obviously, this test isn't apply to apples because they use their own open power chip. But the number of DRM channels is the same, so it's pretty close the number that says we've hit on the right approach. And it performs this way on premise, in the cloud, and we can adapt it to cloud temp space. So what's our roadmap for integrating EE5 into the product and compare replacing the query executed the database to replacing the crankshaft and other parts of the engine of a car while it's been driven. We've actually done it before, between Vertica three and a half and five, and then we never really stopped changing it, now we'll do it again. The first part in replacing with algorithm called storage merge, which combines sorted data from disk. The first time has was two that are in vertical in incoming 10.0 patch that will be EE5 or resegmented storage merge, and then convert sorting and grouping into do out. There the performance results so far, in cases where the Vertica execute is doing well today, simple environments with simple data patterns, such as this simple capitalistic query, there's a lot of speed up, when we ship the segmentation code, which didn't quite make the freeze as much like to bump longer term, what we do is grouping into the storage of large operations, we'll get to where we think we ought to be, given a theoretical minimum work the CPUs need to do. Now if we look at a case where the current execution isn't doing as well, we see there's a much stronger benefit to the code shipping in Vertica 10. In fact, it turns a chart bar sideways to try to help you see the difference better. This case also benefit from the improvements in 10 product point releases and beyond. They will not happening to the vertical query executer, That was just the taste. But now I'd like to switch to the roadmap first for our adapters layer. I'll start with a story about, how our storage access layer evolved. If you go back to the academic ideas, if you start paper that persuaded investors to fund Vertica, read optimized store was the part that had substantiation in the form of performance data. Much of the paper was speculative, but we tried to follow it anyway. That paper talked about the WS with RS, The rights are in the read store, and how they work together for transaction processing and how there was a supernova. In all honesty, Vertica engineers couldn't figure out from the paper what to do next, incase you want to try, and we asked them they would like, We never got enough clarification to build it that way. But here's what we built, instead. We built the ROS, read optimized store, introduction on steep major revision. It's sorted, ordered columnar and compressed that follows a table partitioning that worked even better than the we are as described in the paper. We also built the last byte optimized store, we built four versions of this over the years actually. But this was the best one, it's not a set of interrelated V tree. It's just an append only, insertion order remember your way here, am sorry, no compression, no base, no partitioning. There is, however, a tuple over which does what we call move out. Move the data from WOS to ROS, sorting and compressing. Let's take a moment to compare how they behave, when you load data directly to the ROS, there's a data parsing operation. Then we finished the sorting, and then compressing right out the columnar data files to stay storage. The next query through executes against the ROS and it runs as it should because the ROS is read optimized. Let's repeat the exercise for WOS, the load operation response before the sorting and compressing, and before the data is written to persistent storage. Now it's possible for a query to come along, and the query could be responsible for sorting the lost data in addition to its other processes. Effect on query isn't predictable until the TM comes along and writes the data to the ROS. Over the years, we've done a lot of comparisons between ROS and WOS. ROS has always been better for sustained load throughput, it achieves much higher records per second without pushing back against the client and hasn't Vertica for when we developed the first usable merge out algorithm. ROS has always been better for predictable query performance, the ROS has never had the same management complexity and limitations as WOS. You don't have to pick a memory size and figure out which transactions get to use the pool. A non persistent nature of ROS always cause headaches when there are unexpected cluster shutdowns. We also looked at field usage data, we found that few customers were using a lot, especially among those that studied the issue carefully. So how we set out on a mission to improve the ROS to the point where it was always better than both the WOS and the profit of the past. And now it's true, ROS is better than the WOS and the loss of a couple of years ago. We implemented storage bundling, better catalog object storage and better tuple mover merge outs. And now, after extensive Q&A and customer testing, we've now succeeded, and in Vertica 10, we've removed the whys. Let's talk for a moment about simplicity, one of the best things Mike Stonebreaker said is no knobs. Anyone want to guess how many knobs we got rid of, and we took the WOS out of the product. 22 were five knobs to control whether it didn't went to ROS as well. Six controlling the ROS itself, Six more to set policies for the typical remove out and so on. In my honest opinion is still wasn't enough control over to achieve excess in a multi tenant environment, the big reason to get rid of the WOS for simplicity. Make the lives of DBAs and users better, we have a long way to go, but we're doing it. On my desk, I keep a jar with the knob in it for each knob in Vertica. When developers add a knob to the product, they have to add a knob to the jar. When they remove a knob, they get to choose one to take out, We have a lot of work to do, but I'm thrilled to report that in 15 years 10 is the first release with a number of knobs ticked downward. Get back to the WOS, I've said the most important thing get rid of it for last. We're getting rid of it so we can deliver our vision of the future to our customer. Remember how he said an Eon and sub-clusters we got all these benefits from shared storage? Guess what can't live in shared storage, the WOS. Remember how it's been a big part of the future was keeping the copies that identical to the primary copy? Independent actions of the WOS took a little at the root of the divergence between copies of the data. You have to admit it when you're wrong. That was in the original design and held up to the a selling point of time, without onto the idea of a separate ROS and WOS for too long. In Vertica, 10, we can finally bid, good reagents. I've covered a lot of ground, so let's put all the pieces together. I've talked a lot about our vision and how we're achieving it. But we also still pay attention to tactical detail. We've been fine tuning our memory management model to enhance performance. That involves revisiting tens of thousands of satellite of code, much like painting the inside of a large building with small paintbrushes. We're getting results as shown in the chart in Vertica nine, concurrent monitoring queries use memory from the global catalog tool, and Vertica 10, they don't. This is only one example of an important detail we're improving. We've also reworked the monitoring tables without network messages into two parts. The increased data we're collecting and analyzing and our quality assurance processes, we're improving on everything. As the story goes, I still have my grandfather's axe, of course, my father had to replace the handle, and I had to replace the head. Along the same lines, we still have Mike Stonebreaker Vertica. We didn't replace the query optimizer twice the debate database designer and storage layer four times each. The query executed is and it's a free design, like charted out how our code has changed over the years. I found that we don't have much from a long time ago, I did some digging, and you know what we have left in 2007. We have the original curly braces, and a little bit of percent code for handling dates and times. To deliver on our mission to help customers get value from their structured data, with high performance of scale, and in diverse deployment environments. We have the sound architecture roadmap, reviews the best execution strategy and solid tactics. On the architectural front, we're converging in an enterprise, we're extending smart analytic clusters. In query processing, we're redesigning the execution engine for the cloud, as I've told you. There's a lot more than just the fast engine. that you want to learn about our new data support for complex data types, improvements to the query optimizer statistics, or extension to live aggregate projections and flatten tables. You should check out some of the other engineering talk that the big data conference. We continue to stay on top of the details from low level CPU and memory too, to the monitoring management, developing tighter feedback cycles between development, Q&A and customers. And don't forget to check out the rest of the pillars of our roadmap. We have new easier ways to get started with Vertica in the cloud. Engineers have been hard at work on machine learning and security. It's easier than ever to use Vertica with third Party product, as a variety of tools integrations continues to increase. Finally, the most important thing we can do, is to help people get value from structured data to help people learn more about Vertica. So hopefully I left plenty of time for Q&A at the end of this presentation. I hope to hear your questions soon.