Advanced Exadata Performance

Troubleshooting and Optimization Master Class by Tanel Põder

This advanced seminar is meant for DBAs and developers who already have at least the basic experience of working with Exadata and have seen some of the common performance problems that may occur in production or performance tests. In this class we will dive deep into details of how Exadata works internally and then how to systematically troubleshoot performance problems caused either by application, database, storage cell software or Exadata hardware issues. The demos and troubleshooting examples are shown on Enkitec’s in-house Exadata X2-2 and V2 machines and we will use both standard Oracle tools and advanced Exadata troubleshooting scripts written by Tanel.

What students will learn:

  • Understand how Exadata features, the storage cell servers and the DB-cell communication work internally
  • How to use database, OS and cell metrics for systematically troubleshooting Exadata performance issues, from basic to complex problems involving multiple components, bugs and hardware issues
  • How to identify which workloads and SQL statements are not Exadata-friendly and what to do about it
  • How to design for optimal Exadata performance and avoid design and coding pitfalls

What this class is not:

  • This class is not an introductory “what is Exadata” class, but an advanced seminar about getting the most out of your Exadatas and resolving even complex performance problems systematically and fast
  • This class is not an introductory SQL or database tuning class, but focuses on troubleshooting only Exadata-related performance issues of your workloads
  • This class is not a general Exadata infrastructure troubleshooting class (like hardware, networking and Grid Infrastructure configuration problems), instead we focus entirely on performance

 


Course Topics

Day 1 – Exadata Internals, Systematic Troubleshooting and Smart Scans
    • Understanding the low-level internals of Exadata data flow all the way from cell disks to the database nodes – and the physics behind its performance
    • Using basic performance tools like V$SQLSTATS, ASH and SQL Monitoring views on Exadata
    • Troubleshooting why Smart Scans (and direct path reads) do not get used
    • Measuring the efficiency of Smart Scans and how much work gets offloaded to the storage cells
    • Troubleshooting advanced Smart Scan performance problems, like passthrough mode and selective fallback to block IO – with Exadata Snapper
    • Using CELLCLI and cell-level metrics where database & session metrics aren’t enough
Day 2 – Parallel Execution, Hybrid Columnar Compression and OLTP
    • PX issues, including the In-Memory Parallel Execution feature
    • Optimizing hash joins for efficient bloom filtering in the cells
    • Troubleshooting Exadata Hybrid Columnar Compression issues
    • Troubleshooting long commit and disk write latency issues on Exadata
    • Troubleshooting flash caching issues and optimizing for tiered caching
    • Learn to use the Exadata scripts, tools and OS commands Tanel regularly uses for Exadata performance troubleshooting
    • Q&A and free-form demo session and/or reviewing Exadata performance case studies
Upcoming Sessions
There are no sessions scheduled at this time. Contact us to schedule a private or custom version of this course.
Duration
Instructor Led: 2 days
Online: 4 days
Classroom training typically runs from 9:00 AM until 4:00 PM daily. Online training typically runs from 12:00 PM ET until 4:00 PM ET daily.

Please note that actual times may vary.
Prerequisites
The students should be familiar with Exadata fundamentals and either have basic performance optimization experience on Exadata or have already attended Enkitec’s Exadata Performance Optimization class.