This blog is a summary of the items I thought really jumped out in the technical talks at Momentum Vienna.
As in one of my previous blogs, it is real clear that the story EMC is telling is: The server is the best and fasted in the ECM world but that is not the reason to buy our software.
And all the presentation had that same atmosphere. It is the best and we are proud of it, but we are not shouting about it.
I like this. If you see all graphs of Ed Bueche you are stunned how fast the new D7 server is. The amount of performance gain on the most fundamental places in the communication of the whole stack is incredible. But everybody presents it as if it are normal expected improvements.
But what are those major improvements:
- Context change. The major problem in a lot of performance problems is the start-up time. The first time a user does something after a certain amount of inactivity, the system waits for multiple seconds before the new session is available and the user gets a response back. In D7 this context change improves not only in percentage but also, the more concurrent sessions there are, the more improvement percentage increases.
- Type caching. The super solution to make the overall performance significant faster also had a big downside. Once the system starts caching stuff for you, it currently stops your system until all is in. I hoped there was a way to do this on the background, but that was (still) impossible. What they did is change the way they get and store the caching. Looking at the graphs again, the performance gain will mean that the terrible start-up for any logon will be gone, even with large sets of users.
Note: what is still a comment on this is, that in the graph the performance went down significant with 400 concurrent users. This number 400 is very related to the system set-up, but it will mean that there is always a limit where your system will (probably suddenly) become very slow because the number concurrent sessions or actions on the server.
- xPlore 1.3. There is a lot of benchmarking done with xPlore. The takeaway is still: it is a lot better and new patches come out very frequent, but there is still the possibility to run into performance issues. Your system support team should be able to understand the way xPlore handles queries and how to interpret the reporting of xPlore. Ed has a lot of blogs and white-papers about possibilities and test options, but you need to know Documentum, SQL, full-text vs database searches, combining the two and the pain for combining the two and so on.
So overall it is great to see that the major focus for the server is performance not new functionality.
I’m looking forward to my playtime with this new version and test the graph results myself.