|Keynote||Have you ever wondered why the wings on a fighter jet are designed substantially different from the wings on a 747? Why a bridge over a creek is engineered differently from a bridge over the San Francisco bay? Surprisingly, the reasons are similar to why your database applications on engineered systems should be different from what you've currently implemented. We'll explore those differences and the reasons behind them.|| Tom Kyte|
|DBFS performance||DBFS is often recommended by Oracle as a clustered filesystem of choice when running on the Exadata platform. The session will focus on in-depth performance overview of the entire DBFS stack, starting with dbfs_client on Linux all the way up to SecureFiles inside the Oracle database.|| Alex Fatkulin|
|Everything You Ever Wanted To Know About Exadata Patching||If you have read about Exadata on the Internet, you have probably heard the term “patchmadness” before. What does this mean? In this session, Andy Colvin will apply the latest patches to a live Exadata system. While the patching is going on, Andy will provide a deep-dive into how the patching process works, covering the various patches to be installed on Exadata, how each type of patch is applied, and how to minimize risk during the patching process. Learn about when to apply patches rolling versus taking an outage, how to patch a standby system first, and a few more tricks that aren’t in Oracle’s documentation. Having patched more than 60 Exadata racks for various customers, ranging from V2 racks to X2-8 and X3-2 configurations, Andy will share some stories about problems seen in previous patches, along with workarounds used to resolve said issues.|| Andy Colvin|
|The Future of Big Data in the Enterprise (A Model for Maximizing Value and ROI)||Hadoop and NoSQL clusters have now become standard “wrenches” in the data analytics toolbox, helping companies to derive value from the rapidly expanding and very diverse sources of data encountered in contemporary business. Many early and current uses of these tools grew from POCs and are now configured as standalone systems. But except for a few use cases, large scale enterprises are immediately encountering limitations which limit their use and therefore their value. The big data analytics systems of the future will be end-to-end systems with much greater integration. This presentation will discuss such an integrated end-to-end model, its implications on data center storage and networking, and how we expect this model to evolve in the near future to give customers must greater flexibility, ease of use, and therefore value from their data. || Dan McClary|
|TBD||Coming Soon|| Doug Cutting|
|Exadata ecosystem in TURKCELL||Turkcell, Europe’s 3rd largest telecommunication operator has purchased his first Exadata in April 2010. Now having 5 exadatas supported by Sun ZFS 7420 Storage Appliance, GoldenGate for real time data transfer and Oracle Enterprise Manager 12c for monitoring , Turkcell has created a fine tuned ecosystem for DWH environments. The biggest database is 75 TBs in HCC compressed and is running on 12 nodes. Total size of all 8 dbs are 200 TB HCC compressed. Attend this session to get a feeling about day-to-day operations and customer experience change after Exadata.|| Ferhat Sengonul|
|Multi-block I/O Talk||This presentation is about how the Oracle database implements multiblock reads on Linux systems. Whilst this may look like a simple and easy to understand topic (the system fetches multiple blocks conforming db_file_multiblock_read_count blocks instead of one, right?), in reality it isn't. The description of the former line is mostly true for version 10 non-PQ multiblock reads, but with Oracle version 11 it has changed.
Oracle silently introduced true asynchronous reads with version 11, called 'adaptive direct path reads', which happen under specific circumstances. This session outlines these circumstances. One of the most eye-catching features is reading blocks to the PGA, which makes the reads non-shared, which is different from the traditional reading to buffer cache/SGA.
You probably wonder where Exadata fits in, the description doesn't mention it. Well, the direct path read function is the point where Exadata smartscan function is considered. In other words: having a solid understanding of direct path reads is key for using smartscans|| Frits Hoogland|
|Is Hadoop for You?||The Big Data revolution is here, and everyone wants to drive better insights from their data. But do you really need a whole new architecture to do that? Oracle ACE Director Gwen Shapira will discuss the benefits of using Apache Hadoop as part of a Big Data architecture, and the associated costs. The session will cover variety of use cases and will discuss when Hadoop is a suitable solution and Oracle RDBMS offers a better alternative. Operational and development challenges such as scalability, real time performance, storage requirements and rapid development will be discussed in depth, to help IT managers choose Hadoop for the right projects and ensure successful implementation.
|| Gwen Shapira|
|The Myths of Exadata: The Most Common Mistakes made by Beginners||Too good to be true? Sometimes, it seems that way when discussing the implementation promises of an Exadata environment. These promises include no system outages, incredible throughput and no management required. While some of these are true, they do require experience and system knowledge for avoiding the pitfalls.
Utilizing experiences of over 130 Exadata implementations, this presentation provides the strategies for avoiding the common pitfalls of implementing Exadata.|| Hank Tullis|
|The Optimizer, Exadata, and SQL Tuning||By design the Optimizer is not specifically aware of Exadata, but you can help the Optimizer create optimal plans for Exadata. The Exadata features that make significant performance gains possible requires shifting both your thinking and the approach used to achieve optimal performance. In this session, traditional non-Exadata tuning methods will be compared to those most effective on Exadata. Examples will be shown of how old, reliable strategies failed and had to be adjusted and new strategies learned in order to successfully tune SQL running on Exadata. We'll also discuss what options you have to help the Optimizer be more aware of Exadata's capabilities.|| Karen Morton|
|Performance data visualization with Oracle R and Tableau||In this session Tyler and Karl will do an example based and problem/solution approach in showing how to visualize performance data from Exadata benchmarking and some of their most interesting production workload characterization and sizing scenarios. Each problem/scenario will be visualized using Oracle R and Tableau and will drill down deep on how to effectively explore the data and match it with the different ideal visualization types leading to quantitative information and solid solutions.|| Karl Arao, Tyler Muth|
|TBD||TBD|| Kerry Osborne|
|Infiniband for Oracle DBA's||Normally you don't have to look at your Infiniband infrastructure, it just works but what if it suddenly stops working? Is your backup appliance on infiniband slow, but you don't know how to troubleshoot it? Do you think you have a misconfigured infiniband subnet but you don't know what to check? This presentation will help you to go from there, explaining how infiniband works and what makes it tick. Learn how LID's are distributed within the infiniband subnet, how the GID is constructed and how traffic is routed from one HCA to the other. See how you can use infiniband tools to test, troubleshoot and diagnose your fabric. Furthermore we will take a look at on a range of protocols that can be used on Infiniband like RDMA, RDS, IPoIB and how they interact with each other.
We will dive into where in the OSI model infiniband operates, go into the physical layer, data link layer, network layer, transport layer and find out where it is significantly
different from the ethernet architecture. || Klaas-Jan Jongsma|
|Explaining the Exadata Explain Plan||Note: This session by Maria Colgan is Part 1 of the joint session, "Understanding Exadata from the Top Down to the Bottom Up." For Part 2, please see Roger Macnicol's presentation "Exadata SmartScan Deep Dive."
SQL statements in Exadata environments often have complex execution plans involving optimizer transformations, parallel execution and operations that can be offloaded to the Exadata storage. Determining if the execution plan you are looking at is the best plan you could get or attempting to improve a poorly performing execution plan can be a daunting task even for the most experience DBA or developer. In this session, I will take a top down approach to tackling SQL tuning in an Exadata environment and using real-world example demonstrate how you should navigate these complex execution plans to determine exactly what Oracle is up to and what you can do to improve the performance of your SQL.
|| Maria Colgan|
|Managing Exadata and the Human Factor||Managing Exadata can be an intimidating experience for teams who previously used a silo-DBA approach for database management. Over the course of the last decade though RAC and especially Exadata quickly rendered the strict separation of duties (storage, network, database and operating system administration) obsolete. This talk focuses on some of the potential shortcomings in the current IT landscape and how an Exadata system could be implemented within the organisation.|| Martin Bach|
|Exadata SmartScan Deep Dive||Note: This is Part 2 of the joint session, "Understanding Exadata from the Top Down to the Bottom Up." For Part 1, please see Maria Colgan's session "Explaining the Exadata Explain Plan."
SmartScan lies at the very heart of Exadata’s performance advantage.
After reviewing what SmartScan is and how it works we’ll take a deeper dive looking at: 1. What is and isn’t offloaded and why (and how that has changed over the various releases.) 2. The parallelism model and how SmartScan interacts with PQ. 3. How SmartScan works with ASM, Storage Indexes, and Zone Maps. 4. The benefits of columnar processing on Exadata and the centrality of SmartScan to HCC. 5. How to read the AWR stats for SmartScan and HCC. 6. How to isolate problems with SmartScan and the various parameters available to help diagnose issues.
|| Roger MacNicol|
|Oracle Exalytics - Tips and Experiences from the Field||Rittman Mead were the first UK Oracle Partner to acquire an Oracle Exalytics server, Oracle's in-memory engineered system for business intelligence. Over the past year we have implemented and prototyped Exalytics solutions for customers around the world, and have built-up a body of knowledge around the best way to load and refresh the TimesTen cache, create highly-interactive dashboards, and support large communities of users all running complex queries concurrently.
We've also encountered the usual "Version 1.0" issues, and worked out how best to deploy Exalytics alongside other engineered systems such as Oracle Exadata, and how to deploy Exalytics so that it's highly-available and secure. In this session, we'll go through our experiences, share tips on what works well and not so well with Exalytics, and talk about some of the utilities and accelerators our team have developed to speed up the deployment of an Exalytics system. In particular, we'll focus on the TimesTen in-memory cache layer, and look at how well it performs compared to a regular Oracle database in terms of data loading, aggregation and data access.
Attendees will learn from this session the basic proposition with Exalytics, and how the key features (in-memory analysis, support for large amounts of concurrent users etc) work. We'll discuss what approaches suit the Exalytics platform best, how new features since V1.0 stack up, how to deploy it alongside Exadata, and how to integrate Exalytics into your wider enterprise architecture.|| Stewart Bryson, Mark Rittman|
|Resource Management||Database consolidation and mixed workload databases are a common trend on Exadata. In these environments, one database or workload can use a disproportionate amount of resources, due to a surge in workload or a misbehaved query. This session describes the features in Oracle Resource Manager for controlling CPU usage, controlling disk I/O usage, detecting and controlling runaway queries, and scheduling parallel statements. These features allow the DBA to guarantee resources for critical workloads while maintaining efficient resource utilization. The session provides best practices for consolidated environments, mixed workload environments, and data warehouses.|| Sue Lee|
|Hadoop Internals for Oracle DBAs||In this session we will take a deep dive into Hadoop, HDFS and MapReduce internals from an Oracle Database administrator's perspective. We will explore the similarities and differences between a Parallel Query running on an Oracle RAC Database and a MapReduce job running on a Hadoop Cluster.
This is not a high-level session talking about the future of Big Data, but a detailed look into Hadoop parallel processing and IO flow. || Tanel Poder|
|Real World Parallelism on Exadata||While Parallelism is not an Exadata-specific feature, it is something that should not be overlooked on Exadata. It's unique architecture allows Exadata to perform parallel queries like no other Oracle platform. This session will focus on determining the best strategy for Parallelism and how to avoid the pitfalls. Live demonstrations of Oracle's Parallel Query features will be used to show effective testing methodologies. Attendees will take away knowledge on building and maintaining a solid Parallel Strategy for their Exadata environments.|| Timothy Fox, Tanel Poder|
|Bloom Filters on Exadata||Bloom Filters are a critical component of performance on Exadata in support of the concept of "Filter Early". However, it's a topic that has received a lot less attention than other features of SmartScan.
This session will start by briefly defining the concept of bloom filters at a high level with visual examples. It will quickly move to examples of bloom filters in action on Exadata, using metrics to emphasize the key concepts. These examples will help answer the questions:
- Why are bloom filters important?
- How can I identify bloom filters?
- What is bloom pruning and how do I identify it?
- What impact can erroneous statistics have on bloom filters?
- What is the comparative performance impact of disabling bloom filters?
- What are some of the nuances of identifying bloom filters in query plans?|| Tyler Muth|