Uncategorized

Update 7th Oct:

The run is over, and Simon completed his first half-marathon in 1:45:33, raising a little over £725.00. He was pleased with his performance as he had decided that 1:50 would be a viable target and 1:45 a desirable, but slightly optimistic, target.  He was also very pleased that the generosity of Scratchpad readers and other members of the Oracle community that helped to lift the sum he raised to well over double his original target. Our thanks to anyone who donated.  (There is still a little time to donate if you were thinking of doing so, the web address is still open.)

======================

Here’s the content of an email I sent to Packt back in February this year:

Please ensure that I don’t hear from Packt again.

I have been approached twice in the past and explained that I don’t have the time, and I’m not interested in reviewing books where I have had no involvement with the authors.

This elicited an apology, of course, then on 4th August (after two more pieces of spam from them) I sent them another email quoting the above with the following introduction:

Further to your recent emails, please see below the contents of a note I sent to one of your fellow “executives” at packt a few months ago.

Please note that if I hear from packt again I will publish a note on my blog pointing out that you are a disorganised, incompetent and unprofessional bunch of spammers.

For an interesting read that includes a discussion of what Exadata is and isn’t, the arithmetic of data processing, and the balance of power between hardware and software, you should visit Kevin Closson’s website where he’s republished the interview he gave to the Northern California Oracle User Group a couple of months ago.

The posting includes a list of the questions he answered, and a link to a pdf file of the magazine. Topics that come out of the interview include: what was all that flash cache really for, how much benefit should you see from a hardware upgrade whether it’s to Exadata or to non-proprietory hardware, how much more can you read from a proof of concept than the authors thought they were putting into it, how does a symmetrical system differ from an asymmetrical system.

This is known in the DBA trade as being “in the brown stuff”

Here’s an amusing little question that appeared on OTN a short while ago:

From the performance point of view, which of the following is better and why :

(1) select FILE_NAME, TABLESPACE_NAME, SUM(BYTES)/1024/1024/1024 from dba_data_files

(2)  select FILE_NAME, TABLESPACE_NAME, SUM(BYTES)/(1024*1024*1024) from dba_data_files


Assuming we add the necessary “group by” clause to the end of the queries, how could we find out if there is any difference (other than testing the queries on a system with a very large number of data files to see if we can spot a difference in the CPU usage caused the by change in the arithmetic expression ?

So this is slightly off topic, as it doesn’t deal with anything directly related to Oracle, it’s more a reference to those who are using Supermicro boards like I do. The nice thing about professional hardware is that you do not need a keyboard or mouse, or even a monitor. The board I have comes with an IPMI interface, and has a KVM over IP ability. The interface can be accessed from a dedicated software or a web browser.

IPMIView

The suitable application to access your KVM is called IPMIView, and it’s available for MacOS, Linux and Windows.

Since I didn’t want to install a lot of application on my laptop I opted to install IPMIView for Linux on my Ubuntu 12.04 LTS desktop running within a virtual machine. However I couldn’t get it to install at all until I found a great source how to do so. Here are the steps:

A customer called and wanted to know why the development database was so much slower than production when both databases were on the same type of machine and same type of storage.
To analyze the situation, the same query was run on both databases with

       alter session set events '10046 trace name context forever, level 12';

and sure enough, development (SID=dev) showed average I/O almost twice as slow as production:

… Read an email I received today from Oracle Corporation Hello Niall, I hope you are well. I would like to introduce myself as your Oracle Linux and Oracle Virtual Machine Account representative for I would like to set up a 5 minutes introduction call with you to [...]

In my previous post about perf I showed how it can be used to determine the function which is used the most systemwide.

One thing it helped me with, is determining the impact of swapping on a system. The testcase I build for that is kind of a no-brainer (setting up a 3GB SGA with 2GB of physical memory in a test VM), but think about a few things: I regularly encounter systems who have an amount of swap allocated. This can mean memory is simply plain over allocated, but can also be older pages are just swapped out by Linux “just in case”. This probably will happen more with the new oracle recommendation for “swappiness” (the eagerness of Linux to swap) to set it to 100 (used to be 60). The amount of swap transfers can be measured, but the regular tools cannot measure the impact. With perf, this is possible.