Tag: Oracle

I didn’t get to update the blog yesterday because I had a little bit of an issue with my hosting provider. The site seems OK now, so sorry for any inconvenience it caused… 🙁

As mentioned previously, I didn’t get to see too many presentations in the first two days because of the “Ask the Experts” panel and “Meet and Greet” presentations, so I was determined to do as much as possible on Wednesday, so here’s what I went to see…

“Recent Enhancements in Oracle’s Cost Basede Optimizer” by Jonathan Lewis – Over the years I’ve read quite a bit of Jonathan’s stuff and it has become obvious to me that my brain is too small and squishy to think on that level for too long. Sometimes I read his stuff and think, “I rubbish and I’m going to give up!” With that in mind, I was a little nervous about this presentation, thinking it may confound me to the point of making me quit Oracle. Fortunately, he dummed it down to mortal-level and an went away thinking, “Cool! I know what I’m doing.” Nice one Jonathan! 🙂

“Oracle Real Application Clusters: Scale Up or Scale Out” by Erik Peterson & Lan Shao – I’ve been using RAC for quite some time and I’ve seen my share of issues, so I was interested to see some comparisons between single large nodes and many small nodes. The presentation was very “big picture”, so it didn’t contain much raw data, but it seems that scale out (lots of small nodes) gives much better performance than scale up (one node). By the presenters own admission, some of the performance comparisons were very surprising. If you are thinking about RAC, it might be worth taking a look at this overview to get a feel for the configurations you might want to consider. You have to remember, these performance benchmarks were specific for the China Lottery, high numbers of small OLTP transactions, so you can’t assume hybrid or warehouse systems will respond the same way. Interesting stuff though!

“Recent Advances in Automatic SQL” by Jonathan Lewis – Once again, I didn’t walk away feeling like a rank amateur, so thanks Jonathan! This was a quick run through a number of the automatic SQL tuning features in 9i and 10g. It’s well worth checking out this paper if you are new to the automatic SQL tuning features because it will help you avoid a number of common mistakes people make when using these features.

“Tuning Oracle9i and Oracle Database 10g with Statspack and Oracle AWR Report” by Rich Niemiec – This presentation contained about 130 slides, most of which had numerous subsections. As a result, it felt like 500 slides in 60 minutes and I’m sure this confused this hell out of the newbies in the crowd. That said, Rich is a funny and interesting presenter. A little time was spent discussing the use of hit ratios for tuning, which you don’t hear a lot of these days. His attitude was, variations in hit ratios can give an indication that something has changed in the system, not that specific values of the ratios mean good or bad performance.

I also spent a lot of time at the trade stands speaking to techies, rather than sales staff. You can get some very useful real-world information from some of these people, and it’s not all biased in favour of their products.

It was a really busy and ejoyable day, but by the time I got back to the hotel I was too tired to go out again, so I just went to bed. 🙂

Cheers

Tim…

I already posted to say this mornings “Meet and Greet” went really well. This afternoons was even busier! All the seats were full and there were people standing at the back watching my presentation on PL/SQL tuning. I was extremely surprised by the turnout considering it was a trade stand. Over the last two days I gave away about 60 books, so there are plenty of people going home with heavier suitcases. It was a really cool experience and I got some great feedback from the people I spoke to. I can see how this public speaking thing could get very addictive! 🙂

This afternoon I went to a presentation called, “What Every DBA Must Know About Grid Control”. As the name suggests, it was a quick run through everything you can do with OEM Grid Control, so there wasn’t really much new information for me. The main reason I went was because Julian Dontcheff from Nokia was one of the speakers. I first met (on the net) Julian a few years ago when we were discussing Oracle 9i OCP beta exam on DBASupport.com, but I’ve not heard much about him since he moved to Nokia. I now know why. He’s part of a six-person team managing approximately 500 databases on versions varying from 8i to 10g. He’s a bit too busy to be frequenting forums these days. It was really cool to meet up with him in person after all these years!

I’ve got the bloggers bash tonight, which coincides with the big OpenWorld party. So I’m going to miss out on seeing Elton John perform live to spend the evening with a load of Oracle geeks. That just goes to show how geeky I truely am. 🙂

Cheers

Tim…

Update: I went and met the guys as the blogger drink. It’s always good to meet up with other like-minded people. A few of us then went on to the OpenWorld party. We were lucky, as we got there just before the doors to the Elton John show were closed. I’m not a fan but I was hoping to hear “Yellow Brick Road” and “Rocketman”, which I did. 🙂

This session by Leng Leng Tan was very interesting, listing all the new self-managing features of the next release (codenamed 11g). These features include…

Change Management:

• Database Replay – Allows the total database workload to be captured, transferred to a test database created from a backup or standby database, then replayed to test the affects of an upgrade or system change. Currently, they are working to a capture performance overhead of 5%, so you could conceivably capture real production workloads.
• SQL Replay – Similar to the previous feature, but this only captures and applies the SQL workload, not toal workload.
• Plan Management – Allow you to fix plans for specific statements, regardless of statistics or database version changes.

Fault Management:

• Automatic Diagnostic Repository (ADR) – When critical errors are detected, they automatically create an “incident”. Information relating to the incident is automatically captured, the DBA is notified and certain health checks are run automatically. This information can be packaged to be sent to Oracle support (see following).
• Incident Packaging Service (IPS) – This wraps up all information about an incident, requests further tests and information if necessary, and allows you to send the whole package to Oracle Support.
• Feature Based Patching – All one-off patches will be classified as to which feature they affect. This allows you to easily identify which patches are necessary for the features you are using. EM will allow you to subscribe to a feature based patching service, so EM automatically scans for available patches for the features you are using.

Performance and Resource Management:

• Automatic SQL Tuning – The 10g automatic tuning advisor makes tuning suggestions in the form of SQL profiles that will improve performance. You can tell 11g to automatically apply SQL profiles for statements where the suggested profile give 3-times better performance that the existing statement. The performance comparisons are done by a new administrative task during a user-specified maintenance window.
• Access Advisor – The 11g Access Advisor gives partitioning advice, including advice on the new interval partitioning. Interval partitioning is an automated version of range partitioning, where new equally-sized partitions are automatically created when needed. Both range and interval partitions can exist for a single table, and range partitioned tables can be converted to interval partitioned tables.
• Automatic Memory Tuning – Automatic PGA tuning was introduced in Oracle 9i. Automatic SGA tuning was introduced in Oracle 10g. In 11g, all memory can be tuned automatically by setting one parameter. You literally tell Oracle how much memory it has and it determines how much to use for PGA, SGA and OS Processes. Maximum and minimum thresholds can be set.
• Resource Manager – The 11g Resource Manager can manage I/O, not just CPU. You can set the priority associated with specific files, file types or ASM disk groups.
• ADDM – The ADDM in 11g can give advice on the whole RAC (database level), not just at the instance level. Directives have been added to ADDM so it can ignore issues you are not concerned about. For example, if you know you need more memory and are sick of being told it, you can ask ADDM not to report those messages anymore.
• AWR Baselines – The AWR baselines of 10g have been extended to allow automatic creation of baselines for use in other features. A rolling week baseline is created by default.
• Adaptive Metric Baselines – Notification thresholds in 10g were based on a fixed point. In 11g, notification thresholds can be associated with a baseline, so the notification thresholds vary throughout the day in line with the baseline.

So in summary, the performance and resource management is very much and evolution of the 10g automatic management features, but the change and fault management is completely new in 11g. Exciting stuff!

Cheers

Tim…