Oracle Exadata X8 Overview

[Music] hi I'm Juan Luisa and today I'm gonna introduce Oracle Exadata x8 the foundation for mission-critical database both on-premise in the cloud and for autonomous database now the vision of exadata is very simple it's to provide a dramatically better platform for all kinds of database workloads and there's three parts to Exadata one is the hardware the database team designed the ideal hardware for a database and that includes a scale out database optimized compute networking and storage for fastest performance and lowest costs the most important part of exadata is the software the smart system software well we've implemented dozens of specialized algorithms that vastly improve every aspect of database for OLTP analytics and consolidation and then the third important part is automated management so we've automated all parts of EXA data and automates things like updates configuration and ultimately we're also delivering a fully autonomous database using Exadata an exudate it is identical whether you run it on premises in the cloud or as an autonomous database ok so exit data has now been in the market for ten years it was shipped in 2008 and it's very popular in the market we now have four out of the five top banks telecoms and retailers running Exadata it's extremely successful and its successful across all the different workloads so we run petabyte data warehouses super critical applications things like online financial trading ecommerce also hugely complex applications like s ap Oracle e-business suite Siebel PeopleSoft and also lately there's massive database consolidations being implemented on X data so incredibly successful across all kinds of workloads today over three quarters of the Fortune Global 100 run Exadata and already over a quarter of the Fortune Global 100 companies have adopted exa data cloud and Exadata cloud with exit data is growing at a rate of over a hundred and fifty percent a year so it's been thusly successful so first I'm going to talk about the Exadata x8 hardware and then we'll get into software and some other things the x8 hardware continues our tradition of bringing all the latest state-of-the-art components into Exadata every year so I'll briefly talk about what we changed in Exadata x8 so remember in Exadata is everything a customer needs to run a database so this includes servers network and storage so on the servers will use either two socket or eight socket database servers and of course this year we've adopted the very latest processors they're about 15% faster based on cascade Lake Intel processors and these processors include the latest security fixes so that we don't have any more software overhead we run in exit data with indexed data in ultra-fast InfiniBand internal fabric and then out to the data center we use standard Ethernet and from a hardware perspective the most interesting part of exit data is the scale out intelligent storage so an exit data will use servers for storage we use two socket servers this year we've adopted the latest cascade Lake CPUs but we've also increased the processing power by over 60% which makes exit data offload even more effective and then we've updated to the latest higher capacity district so 40% higher capacity in Exadata x8 versus x7 now there is one larger change in exit data x8 we're introducing a new type of Storage Server so traditionally we've had a high capacity storage server in an all flash us storage server we call it extreme flash now we're introducing a new type called the exit data extended or XT storage server and the idea this is to provide a much lower cost for exit data storage for in frequently accessed older or regulatory data so it's like an exudate a storage server much lower cost for lower performance needs so for the cost of a Hadoop or object storage you get something that integrates seamlessly into the database and into the exit data so it's very efficient very high capacity uses our hybrid column compressions it's very simple because you just drop it into an exodus completely transparent to applications transparent to the operational model the security model the encryption all you got to do is drop these servers into the exit data rack and we'll use them for extended storage also the extended storage server doesn't require exit data system software so that's optional you can use it if you want to do smart skin but you don't have to if you want to lower cost even more so that's the new the new server that's introduced with exit data x8 okay now exit 8x8 of course because it's taken the latest hardware components and more advanced software we're able to achieve even better performance than we've ever achieved so an X is a 2 X 8 a single rack can achieve 560 gigabytes of i/o throughput that's 60% more in the all-flash config then we did a year ago with exadata x7 and we run over 6 and a half million OLTP iOS per second that's 25% more per storage server than before and it's not just the total number we can also run them at very low latency so we can run over 3 and a half million iOS per second while keeping latency below 250 microseconds so extraordinarily high I apps and low latency and of course this is much faster than other competing products and it scales linearly as we add more recs now as a quick comparison and Exadata x8 outperforms and all flash EMC power Macs across all the different metrics so the extra data provides seven point four times more throughput over two times faster io latency and achieves over one and a half million more I ops and of course the exit data scales beyond the single rack so now that I've talked about Hardware what's new in Hardware let me talk about what we're doing on the software side which is actually the real secret sauce in an exit data that that drives performance availability and cost so I'm going to talk about first what we do in software for analytics then OLTP then consolidation so starting with analytics there's a number of features that we've developed over a lot of years to greatly speed up analytics the foundational feature of X data for analytics is the ability to offload data intensive sequel so it runs directly in storage by passing any bottleneck and this is called smart scan technology and with it were able to achieve hundreds of gigabytes per second of throughput also we have a unique algorithms that we offload for data mining decryption aggregation backups dozens of algorithms that we've added over the years in addition the offloading IO we're gonna actually greatly reduce the total amount of i/o so we have something called database aware flash caching the caches iOS in flash the large majority of iOS go straight to pci flash and also we have something called storage indexes which can eliminate io for data that's not relevant to a particular query and often this can eliminate ninety or ninety-five percent of all the i/o lately we've also focused a lot on columnar formats so initially the next day and many years ago we introduced something called hybrid columnar compression and the goal of that was to greatly reduce the space needed to store data and as you reduce the space it can also speed up the analytics by an order of magnitude now we've gone much further than that in developing columnar technology so because of the high-performance i/o of Exadata the bottlenecks have shifted from what used to be io bound to now it's mostly CPU bound so we need to bring in new algorithms to break through this new bottleneck so what we've done is we brought our in-memory database technology into exit data storage so we use the exact same formats that we use in the database servers to speed up analytics with an in-memory format that can be used with fast vector processing operations in the chips so that greatly reduces the amount of CPU needed to scan data and greatly accelerates analytics and recently we've actually expanded that so that those in memory columnar formats work even for OLTP databases and for hybrid columnar compressed databases and shortly we're gonna release a new version of EXA data that will adopt the Intel obtained DC persistent memory to provide even more columnar storage and more speed-up for analytics and exit data okay so now I've talked about analytics let me talk to them about OLTP so again there's a lot of algorithms that we've introduced to optimize OLTP and exit data first of all exit data completely eliminates the traditional bottleneck to LTP which is ios per second limitations on io per second and it does this because exit data has completely scale out storage so it scales linearly we use ultra-fast nvme flash and we have our ultra-fast infiniband network so it allows us to deliver millions and millions of iOS per second in an exit data another unique thing in exit data is we have special algorithms to speed up database logging and commits so we can greatly reduce the latency of those making transactions run much faster in addition to speeding up iOS we also have done a lot of work to eliminate any kind of stalls from failed or SiC components so for example we can detect when a server is failing much faster than in a normal configuration to avoid any kind of long time out when we're victim the server from the configuration another important thing is that we're able to redirect iOS around any kind of component that is sick or failing now most recently we focused on eliminating the internode communication that's required to coordinate across the nodes of a cluster so we're working on speeding up and eliminating any kind of bottlenecks in the cluster communication so first of all we've added a special protocol where the database can talk directly to the networking hardware by passing all the protocol stacks which gives about three times faster internode OLTP messaging a second big improvement is something we call smart fusion block transfer where we eliminate the log write that's normally necessary to transfer data between nodes and then a third thing our most recent enhancement is we've added a special RDMA protocol remote direct memory access for coordinating transactions between nodes so when one node wants to know the status of a transaction in another node that node can directly read the memory of that other node to find out the status without doing any software at all and that greatly speeds up this coordination okay so what's coming and what's new and OLTP we're gonna be introducing something called persistent memory in the extra data storage servers so the idea here is that we can use that memory both as a tier in front of disk and flash so that we can access data faster but acting more importantly because it's memory we're able to access it remotely using the same our DMA protocol remote direct memory access and it's that protocol that really provides the bulk of the speed-up because it again it bypasses the whole software stack so from the database servers we can reach into the storage server through Hardware directly access that persistent memory and get the result and that gives a much faster latency up to 20 times faster latency for remote i/o operations and of course an exit data we mirror the persistent memory across multiple storage servers so that any kind of failure doesn't affect it and we use it as a shared cache meaning that we move data that originally starts on disk in the flashes it becomes hot and as it becomes really hot we move it in the persistent memory so we dynamically adjust what data is in the persistent memory so only the hottest data is there and that essentially gives us terabytes worth of persistent memory the equivalent capacity so we can run multi terabyte databases with the performance of persistent memory okay so now I've talked about analytics and OLTP so let me talk a little bit about consolidation because I mentioned a big trend now is for customers to create massive database farms on Exadata where they consolidate all their databases in an excavated configuration so we've done a lot of work on handling mixed workloads and consolidation in Exadata a few examples here we automatically prioritize latency-sensitive operations so because we own the whole database and storage we know what's latency-sensitive and what is it so when we have a latency-sensitive operation we can automatically prioritize that all the way from the database through the network and all the way to storage so that's a very big difference there's nothing you have to do that just happens we can also prioritize for example OTP operations over things like analytics or batch we can automatically prioritize important workloads based on user policies so you can say this is more important than that and we'll prioritize that all the way through CPU i/o we can do that by job by user by service by pluggable database by session by individual sequel statements also we've done a lot of work on providing isolation across multiple databases or tenants in a single system so this is what we're using in our public cloud to greatly isolate different users so that the performance of one user doesn't affect the performance of another user okay so now that's a quick view through the consolidation we've also introduced new versions of our exit data system software so we call it 19 1 and 19 two versions of our exit data system software and in that software we've provided at least two 20 unique features and enhancements since last year so there's features in there that improve analytics transaction processing consolidations security we've improved upgrades both at the database level and at the firmware level and also we've added support for latest generation of our database software called database 19c and of course all these software features are available not just in our latest exit data x8 but also in previous generations of exit data included in our system software 19 is update to oracle linux 7.6 and latest generation of oracle vm which are both faster and more secure and we've added a couple of special features one is something we call advanced intrusion detection so what we do here is we're constantly monitoring our system software and if there's a virus or something that tries to change that software we detect those changes and automatically alert when we see any kind of change to our cell phone so that's one thing another big thing that we've added is monitoring across the whole stack including CPU network in memory using new machine learning to try to detect when things are going wrong so now we can look through the entire stack and detect things like stuck processes memory leaks flaky networks etc and will either fix them or Alert on them so this is a brand new much more effective alerting mechanism so now that we've talked about that software and Exadata let me talk about the next big thing which is cloud how we're using EXA data in cloud to deliver game-changing mission-critical databases so first of all we have in Exadata in our public cloud something called the exit data cloud service and quite simply this is an exit data running in our public cloud that you as a customer can just sign up and use so the idea here is we've run the full Oracle database with all the Advanced Options or you can bring your own licenses and it's a hundred percent compatible with the databases that you have running today on-premises excavator of course our fastest most available database cloud platform with the best quality of service in the cloud and the advantages in the cloud is that you get a fast elastic and web driven provision so you can go to the cloud you can click and you can get an exit data very quickly we also handle a lot of automation things like updates backup data guard etc and we Oracle cloud experts manage and deploy all the infrastructure for Exadata and because it's a cloud-based product it also uses a cloud-based subscription model so you don't pay a firm for it you pay on a monthly basis in Exeter in the cloud we have excellent elasticity so we can burst the computer online and you only pay for what you actually use so this has now been production since 2016 so in addition to a public cloud service you can also run the exit data cloud model inside your own data center and that's called exadata cloud a customer so for customers that can't move their workloads to the cloud we essentially bring the cloud into their data centers so the ideas it's very simple it's the exact same cloud hardware software interfaces api's and control plane that we run in our public cloud but they run inside the customer data center and also it's the same pay-per-use subscription model that we use in our public cloud that brings pricing based on a naval core is also into the customer data center so that's exadata cloud customer so we can bring extra data to any kind of model you want you can buy a traditional exa data and run it yourself in your data center you can get extra data in the cloud at customer model which is subscription-based but it's in your data center and we run all the infrastructure or you can get a full public cloud Exadata where it's all operated by Oracle and again you pay by subscription you don't need a data center at all so it's customers choice now of course when we talk about cloud on the next really big thing is autonomous database and cloud so that's we're extending automatic management all the way through the database level so in the traditional model of deploying a database the user deployed everything built and deployed everything so the user built the databases the infrastructure the network and had that manage tune repair everything now with autonomous database the full database lifecycle is completely automated even for highly mission-critical databases and even for highly sensitive data so now you can just basically get a sequel prompt and do your sequel and you don't have to worry about anything in the backend you have to worry about managing any kind of infrastructure or the database itself so the big advantages of autonomous database are you get to focus on innovating more it's actually much lower cost also and Oracle worries about all the security and availability of your data and of course autonomous database in our cloud it's all built on top of our exudative technologies so all the benefits and all the software features I've talked about you get them all in the autonomous database so I've talked about the technology behind exit data the hardware the software the cloud model but overall exit data is really powering the future of database at Oracle so first of all exit data is used for all our SAS applications so that's ERP HCM etc the future of all our applications we're running tens of of customers today are all running on Exadata I mentioned that autonomous database technology which is really the future of database all that is running exclusively on Exadata and we'll be bringing that autonomous database technology to the exadata cloud of customers so customers can have that in their data centers as well but we haven't forgotten about Exadata on-premise that remains a major focus at oracle and we are adding many exclusive database features that were developed for autonomous database so that customers running Exadata on-premises can use them directly there are a lot of specialized algorithms that we've developed that only run on exa data and here are a few of the new ones for example for analytics we can maintain a standby database even if there are non like direct loads we can automate the population of data into the in-memory columnstore we can run in memory technology against external tables we talked about a lot of specialized cache fusion algorithms for OLTP we've also introduced especially optimized key value style access for OTP and optimized iot data loading in the consolidation area we have a lot of enhancements for pluggable database multi-tenant database including refreshable pdb copies per pluggable database key store and we can scale up to 4,000 pdbs on EXA data on a single database and we can also dynamically grow c DB memory with our database 19 c most recently we've also introduced the technology we use in autonomous database as a feature you can also run on on-premises x data so for example the two big features are the ability to gather statistics in real time as the ml executes so that statistics never become stale and the use of machine learning to implement new indexes automatically so both of these capabilities are available now for use by Exadata on-premises customers now x2 the data has been in the market for 10 years and it's really very different from any other product it has so many unique capabilities that there really is almost nothing that competes with it there's no second place and we continue to enhance exadata every single year to provide dramatically better performance cost availability and security than any other platform on the market thanks for joining us today for more information on exudative technology and Exadata x8 I invite you to visit us online you

Channel: Oracle

Views: 12,465

Keywords: transaction processing, exa data, exidata, x8, x7, oracle, engineered systems, Autonomous Database, data warehousing, English, Engineered Systems, Juan Loaiza, 30867875, oracle database, EVP, online transaction processing, search-gallery, oracle exadata, converged infrastructure, Marketing Video, EXT-19-59, Exadata, OLTP, 3play, retire-2024, exadata, integrated platform

Id: szlfbBXoXYs

Length: 22min 52sec (1372 seconds)

Published: Thu Jun 27 2019