Saturday, October 24, 2015

On Solving the Right Problem

Dear Readers,

It has been quite a while since I wrote on the blog due to some past things that occupied my personal time that had to be addressed. Recently I began to solve new performance related problems for a customer. Here is the scenario:

Oracle 12c RAC
Large enterprise SAN
Converged infrastructure

The target goal is to achieve a minimal performance level (SLA) for storage and system performance.

The customer is only receiving a low number of storage performance in terms for overall IOPS with their two node Oracle 12c RAC configuration.

So with little information to go by from past experiences, I asked how they measure such performance values.

Enter the Benchmark Tools!

The customer is using a third party performance tool to measure performance for CPU, system and storage figures. So what is this tool you might ask? Well hold on for a minute we will get to that question in a short while. My first DBA spidey sense was to get the data from the horse's mouth so to speak that is from Oracle! I logged into the Oracle 12c RAC cluster and pulled the recent AWR reports from the cluster and noticed that overall there were no serious performance issues!

It turns out that the performance tool is reporting different numbers than the Oracle database!

Hmm well that sure is very odd! I review the infrastructure between customer sites and lo and behold find that configurations are different! So that calls to mind that Oracle per se is not guilty but a more fundamental issue of apples to oranges comparisons. Which leads to my next thought process



I know this seems obvious to most of us Oracle DBA types right? Well you'd be surprised at how many customers spin their wheels attempting to solve a performance issue by mistaking the forest for the weeds. Instead of immediate jumping to conclusions, take a deep breath and step back to look at the big picture. The following comes to mind:

1. Storage configuration- disks, HBA, HBA, multi path configurations
2. Firmware and patch levels for infrastructure- servers, SAN, networks
3. Review OS configurations and releases
4. Run basic tests to collect data points- Oracle AWR, sar, vmstat, et al.

Stay tuned on my next series of blog posts on how to exactly solve these types of problems. Oh yeah and get ready, set and go for Oracle Open World this coming week!!