June 2016

Someone approached me at a recent conference to ask if I was available for hire and, in the course of the conversation, pointed out that he hadn’t really been certain whether or not it was possible to hire me for consultancy work. This made me realise that I don’t advertise very much, I usually forget to remind people that I can be hired, and I hardly ever remember to hand out (or even have available) business cards at the end of the presentations I give.

I am, however, available to do trouble-shooting, education, reviews, or design almost any time, anywhere, and at short notice. My speciality, though is solving performance problems quickly, and there are times when I find it agonising to watch the slow progression of question and (failure to) answer on OTN taking days, or even weeks, to get to a solution that was “obvious” from the original question and needed only a couple of hours to check and implement.

Folks that have read my blog for very long know that I routinely point out that Intel Xeon processors with fewer cores (albeit same TDP) get more throughput per core. Recently I had the opportunity to do some testing of a 2-socket host with 6-core Haswell EP Xeons (E5-2643v3) connected to networked all-flash storage. This post is about host capability so I won’t be elaborating on the storage. I’ll say that it was block storage, all-flash and networked.

Even though I test myriads of systems with modern Xeons it isn’t often I get to test the top-bin parts that aren’t core-packed.  The Haswell EP line offers up to 18-core parts in a 145w CPU.  This 6-core part is 135w and all cores clock up to 3.7GHz–not that clock speed is absolutely critical for Oracle Database performance mind you.

A question came up on the Oracle-L list server a few days ago about a query whose plan showed several bitmap operations. The problem was that the A-Rows column reported by a call to dbms_xplan.display_cursor() was showing numbers that semed to be far too small. In fact the query was producing a parallel execution plan, so the “actuals” for the parallel server operations were reporting zeros because the OP had used the “allstats last” formatting option rather than just “allstats” – but the numbers were still far too small even after this error had been corrected.