I’ve attended SIGIST’s (Israeli testing forum) convention (26/5) on performance and load testing. As with HP’s last convention, I’ve twitted live during the presentations, and here are these tweets. I’d like to thank all of you who replayed and bounced their ideas by me during the convention (@GlennHalstead, @steve_in_denver, @xerocube and @Michaeldx).

You can always follow my tweets at: http://twitter.com/yassa .

The tweets are presented from first to last (reversed from the usual twitter display):

At SIGIST - The silence before the storm.

Setting things up http://twitpic.com/5znk5

Alon Linetzki is now up, giving a short introduction.

Today’s conference will focus on load and performance testing.

Signs of our time - about 1/3 of the place is empty. Last conference there was a full-house.

Alon is promoting the near convention - 27-29 July. Should be a blast - a lot of high-quality content.

Alon says Agile is falling from grace at Europe, and perhaps will go the way of TQM and other failing methodologies. Do you feel the same?

Now on - Ariel Salperter, founder of the Performance Center for excellence in Comverse.

#SIGIST Ariel says he came to the performance engineer position after seeking a wider look on the application.

#SIGIST Ariel says performance engineers are stuck as "Performance Tests", disregarding the more technical aspects of DB, network & storage

#SIGIST - Ariel says the performance testing excellence center is dedicated to cultivating engineers that tackle a system from end-to-end

#SIGIST - According to Ariel, companies are starting to understand that end-to-end performance testers deserve their own pay-grade

#SIGIST - Ariel presents a definition of performance which doesn’t neglects the user’s perceived performance & experience.

#SIGIST - The out-of-the-box code segments and objects of modern languages are producing programmers with no performance considerations.

#SIGIST - Ariel gives an example of how complex can a performance analysis be (example from the cellular business).

#SIGIST - Ariel predicts the performance engineer is going to be a pivotal position in any n-teir application.

#SIGIST - Comverse has paid over 350K $ in new hardware for a client because of programmers who were not performance oriented.

#SIGIST - The financial client makes companies "jump" on deals without properly asking themselves "can these requirements be fulfilled?"

#SIGIST - The creed should be "Proactive, Systematic & Quantitative approach".

#SIGIST - the oral presentation is lingering, but the handouts are rather insightful. Will try to get a soft copy and post it on AdvancedQTP

#SIGIST - The reason Comverse gave performance engineers power is that product failures could be traced directly to performance.

#SIGIST - The journey this field has undergone is similar to what QA undergone 10 years ago.

#SIGIST - From within the dev team, to a no-go independent unit directly under the project management staff.

#SIGIST - openning bottlenecks in SMS processing has saved Comverse over 20M $ in the last year. God damn…

#SIGIST - In order to encourage the dev teams to "think performance", Converse has deployed "devload" mini-load labs inside the dev teams.

Now off to networking break

I think next time I’ll open a dedicated twitter user for live reports - this must be hell for some of my followers.

#SIGIST - Now: Elad Sender from V-Ness on performance and load test case in C2/C4 systems (command, control, computers & communications).

#SIGIST - The app in questions was meant as one of the central C&C; systems in the army.

#SIGIST - The previous iteration had such poor performance, it actually induced "had habits" and poor judgment in the client.

#SIGIST - the client had preferred updating nylon maps rather than using the app. (~15-30 minutes to load).

#SIGIST - This was supposed to be near real-time app!

#SIGIST - we’re all eager to hear how they’ve managed such a colossal failure.

#SIGIST - This caused an overall restructuring of the entire development approach, and rebuilding more than half the components.

#SIGIST - The main point - do not test performance before deployment - test performance always and continuously.

#SIGIST - If at first you only have a few disconnected components - test them. There’s no excuse for neglecting to run performance & load

#SIGIST - The failed iteration was tested, but with almost no data, which proved to be the critical bottleneck in production.

#SIGIST - In performance & load testing, there’s no room for assumptions or cutting a little slack. Every detail is multiplied tenfold

#SIGIST - this lecture turns out to be less spicy than I expected.

#SIGIST - You’ll always have at least one log you forgot to clean, and it will clog and hang after a week of continuous run. So true.

#SIGIST - You cannot extrapolate load - there’s always the infamous "knee" lurking about.

#SIGIST - Load engineers should always present their findings with DRP analysis.Never put the max available users in case of cluster jump

#SIGIST - Monitoring can radically change the server’s behavior. Remember this.

#SIGIST how can you reconcile the hardware costs of replicating the production env. ? Build special drivers that test only specific aspects

#SIGIST Even on mission critical C&C; systems, you must insist on testing the production env., even if you have to go up to the joint chiefs

#SIGIST - Profiling and integrating performance testing into the entire dev. cycle had quadrupled the max possible uers.

#SIGIST - Well, the convention continues, but I must depart. Thanks for following, A summery of these tweets will be posted to AdvancedQTP.

#SIGIST - Bye all