Archive for the ‘OUGF’ Category
As always on a conference, I try to write a small summary about the sessions I have attended. This is my first visit to OUGF in Aulanko (Hämeenlinna, Finland), which is a beautiful location. The conference site has a SPA, golf courses and everything you would need to relax. But that’s not why I’m here, I’m here for gaining some more knowledge about Oracle products.
So, back to the sessions. The day started out with a keynote by Chris Date. He had a passionate keynote about why nulls are not to be used. I do understand his issue with nulls (and the more I think about it, I get issues with using nulls too). The number of practical problems I can think of when working around that issue are numerous. To start with: all systems I work with are designed to implement the n+2 valued logic, including Oracle’s optimizer. Anyway, for me it is good to have a lecture, because that is what it is, about algebra and n+2 valued logic again. I try to understand and therefore be able to apply logic instead of replicating “facts”.
Next up Thomas Kyte was up for his keynote (Five things you probably didn’t know about SQL). I said some things in the past about Tom’s presentations and for me that still applies, I am far better at reading his stuff than hearing it. Please go and see his session if you can because that what applies to me doesn’t have to apply to you (and his talks do contain lots of useful information!). Some things I noticed and I shoot look into:
- Tom suggests to use an array size of 100 as a rule of thumb, whereas Cary Millsap suggested a 2048 array size at his Mastering Oracle Trace Data series (that I happened to attend last week). This shows again: there is no silver bullet for performance. The answer is “it depends” all over again;
- Gather statistics after some representative queries, that will raise the quality of statistics. Checkout dbms_stats.seed_col_usage. Actually I heard this before by Maria Colgan too (@Hotsos 2011), but I sort of forgot this.
Third up was Juris Trôsins with a session called “Implementing Oracle Database Vault: from basics to some tricks”. Juris showed that just implementing Database Vault is not enough, you need to implement extra security measures like encryption. Most interesting fact for me was that DBA’s won’t see the values for histograms.
No conference is complete without at least one talk about implementing ExaData nowadays. Since one of my clients is looking at implementing ExaData I decided to visit Sergey Shchukin’s session about deploying ExaData. It seems that ExaData is deployed with ASM high redundancy for its data diskgroups, which is interesting to me. I would appreciate if anyone can tell me why high redundancy is chosen over normal redundancy (is it because of cell server reliability?). Also it seems you cannot install any new tooling on the servers (no changes to the software & hardware, except for switch configuration) according to Sergey. Finally I learned that hybrid columnar compression is “free” when you use ExaData.
Graham Wood had an entertaining talk about scaling with a demo that showed what happened if you incorrectly scale your JDBC pool size. His advice is to queue sessions on middleware instead of the database to get a higher troughput. The rationale behind this is that too many sessions on a host use up memory and CPU cache causing the sessions that need to run on CPU to be slower. This didn’t really show up in the demo, but youtube contains some interesting videos where these examples do work.
Protecting confidential data in and out of your database was presented by Jagan Athreya. This was a sessions about data masking, to be more exact the Oracle Enterprise Manager 12c pack that allows you to mask data for test data. It seems that just today the E-Business suite masking tooling was released.
The next slot I skipped to do some final checking on my presentation. Even though I finished the slides two weeks ago I decided to check upon one thing (nm_data.properties). Seems that changed with 12c.
Final slot of the day was for Andrejs Karpovs. He was presenting about Cloud FS / ACFS. Personally I am a great supporter of ACFS, so I was interested to see what he would tell about it. Andrejs showed me how you can have a general purpose filesystem to be managed by CRS.
and that was it for the technical content of day one.