The afternoon had one of the sessions I knew I would enjoy the most before I even got here. It was Carly Millsaps "Thinking clearly about performance". THe reason I anticipated a great presentation is not just Carlys knowledge in the area, but also the fact that he is fast becoming the best presenter in the Oracle area. This presentation was no different, he covered a lot of topics and still made it seem like there ws no rush with anything.
Everyone ought to read the paper the presentation is based on. You can find it here.
The presentation essentially alks through 21 items that makes up things that matters for performance. It also shows why DBAs need to care about the developers area and vice versa.
One key area he brought up was "knowing matters – proving matters more". They reason thi is important is that unless you can prove your theory, it is unlikely to get implemented in an organization that juggles many high priority changes.
Response time does not equal throughput even though it sounds as if it does to most people.
Customers feel the variance, not the mean. This means that it is usually not the average time something takes that is the problem, it is the fact that it sometimes takes 10 times longer that makes the user upset.
Problem ananlysis has to start with knowing both the current state and the goal state. If you do not know both, you cannot get started. Even though it can feel uncomfortable to ask about the goal state, you HAVE to.
Cary used sequence diagrams to show how to find out what it is that needs to be addressed. Without knowing what times each part (not just the DB) takes, it is not possible to know what part to fix first.
Profiling is used to know if the endgoal is possible. If you remove everything that should not be needed (such as waiting for a latch) and the gal is still not met, then it is not just a matter of removing unnecessary things, you probably need to re-architect too.
One of the closing remarks was the classic quote on performance that "The fastest way to do something is to not do it at all". As obvious as it sounds, it is easy to forget when analyzing a problem.
The next presentation was "Going realtime/convergent". It was about Oracle Apps Billing and Revenue management. The presentation was held by a person from Tusmobil in Slovakia. His name is something like Ognjen Antonic. The reason it is "like" is that I have no idea how to make my keyboard produce the special symbols he has above letters in his name.
I was hoping for a little bit of insight into the work with implementing a realtime billing solution on Oracles applications. This being a customercase provided more of their business and market background. We were shown lots of impressive numbers on what it had done and what they had achieved, but for a techie it did unfortunately not provide much insight into the work they performed to achieve it.
My next presentation for the day was attending a presentation on Golden Gate. The name was "Golden Gate: What is all the fuzz about" and was held by Borkur Steingrimsson from RittmanMead Consulting.
It was a review of what you get with the license to start with. One thing that is included is Active DG, which explains part of the license cost. Golden Gate is a nice technology, but it is unfortunate that what was included with the database in previous incarnations is replaced with something that requires a separate license. Still, Golden Gate looks like it has a lot of neat features.
It has recently been certified to be used with exadata for extracting and loading.
GG can be set up to do deferred apply of changes.
The presenter thinks that it has little GUI support, but what it lacks there is made up for in scripting support. This shows that it is primarily geared towards DBAs who traditionally prefers to do work on the command line over just clicking around in a GUI.
GG handles both the initial load to a target system and later incremental loads to keep source and target synchronized.
It has support for DDL and it can change both schema and table prefixes on DDL commands. It will however not change schema prefixes if you hardcode that into procedures or functions.
That presentation was the last one for the day. The evening was rounded off with a five mile roundtrip hike over to mamacita in Marina. If you like mexican food you will be in heaven at this restaurant. THis was by far the best mexican food I have ever had. They have a micheline star and they proved it with every dish we tried and we sampled five things where every thing tried hard to top the previous one.
Take a look at their site if you are or are planning to go to SF. If you go there, you need to try their Crudo de Atun. It is an ahi tuna tartar and it is to die for. This place has what gourmet dreams are made of, I know I'll dream about their food until I get a chance to return.