The evolution of a DIET
Several years ago, I was consulting with a customer who has a very large Oracle database environment that needed a new storage system sized to meet their growing demands. Sitting down with the Oracle DBA and storage team at the customer’s site, I asked them what I thought were some straightforward questions:
“What is your peak IOPS (Input/Output Operations Per Second), and,
“What does your I/O pattern look like, read/write random? Read/write sequential?”
Their responses, or lack of responses, were surprising.
So, I asked, “When is your peak I/O?”
Again, they didn’t have an answer.
The folks responsible for managing and maintain this company’s Oracle database couldn’t answer these questions, AND, what was most interesting to me is, they didn’t even know how to go about collecting the data to find out.
I thought this was strange since every storage vendor in the world seems to sell their arrays around IOPS and performance specs, and the IO patterns for different workloads.
Now, at first I thought this was an isolated instance so I didn’t give it much value. However, a few weeks later at a different customer, also a large company with a large and growing Oracle database environment, I asked the same series of questions and the answers I got back were the same – they just didn’t know.
Despite all the tools and knowledge these organizations had around Oracle, and all the money they had spent to purchase and deploy their databases, they couldn’t answer what should be basic questions any DBA or storage administrator can answer quickly, and with confidence.
I knew as well as they did that there are tools that could provide this data, but these were not metrics they regularly tracked. However, the challenge not having accurate performance, peak I/O or other important metrics cause is – how do you design a new storage solution for something you cannot size or profile? Many times it seemed DBAs or storage admins would guess at what they thought their needs were with not a lot of actual data to back it up. The result of having no data or the wrong data is pretty much the same – the customer ends up with a solution that doesn’t fit and can’t meet their needs.
So, I created a spreadsheet that could show what the I/O pattern of an Oracle database looked like by analyzing the data collected from Oracle Statspack reports. This first spreadsheet-driven tool was simple but it worked for what was needed, and ultimately helped me to help my customers make better buying decisions.
Suddenly my phone was ringing off the hook! And it hasn’t really stopped since – that was nearly 7 years ago.
Over these past 7 years, that simple spreadsheet, which came to be known as the Hitachi Storage Evaluator for Databases, or HiSED, has been improved, expanded beyond just storage systems, and has seen a lot of usage in real-world Oracle environments.
Now, with an all-new, browser-based and greatly expanded utility, and even a new name, Hitachi Database Infrastructure Evaluation Tool, or DIET, can ingest thousands of Oracle AWR and Statspack reports and model the Oracle database workloads of an entire enterprise in literally seconds. DIET can determine when your peak Oracle database performance is and illustrate how it trends over time. It can also show you the types of IO patterns you’re generating and it can very quickly find where performance bottlenecks exist within your Oracle environment – and make recommendations on how to fix them.
Over these past years, as we continued working with customers to right-size the infrastructure that supports their Oracle databases, we accumulated a mountain of raw data about problems customer were experiencing inside Oracle. And, we saw some patterns begin to emerge:
- Hundreds of companies with thousands of Oracle databases were having the exact same performance problems – so these customers were not achieving the goals they set out to in the first place.
- Maybe even more surprising – the vast majority of these customers were not utilizing all the CPU / processing power they had purchased – it wasn’t even close.
That’s a troubling, but all-too common pattern I’ve seen – You’re not getting the performance you need, so you throw more hardware (and expensive software) at the problem but because your utilization is low and your database isn’t optimized, you still don’t get the performance you expect…and on and on it goes.
Over the years, we have taken hundreds of the most common problems with Oracle databases we’ve seen and created a massive library of rules and best practices – this knowledge base and experience is the foundation of DIET. In essence we’ve taken the equivalent of over 500 years of Oracle database performance tuning skills and automated it.
Our experience with DIET, and HiSED before it, includes the analysis of over 1,000,000 Oracle AWR and Statspack reports. The Oracle solutions team at HDS has worked with over 20,000 different databases and nearly 2,000 organizations globally – many of them are the largest in their industries, and near the top of the Fortune 500 list.
It is just amazing what we have learned about how our customers, and even those who are don’t use Hitachi products and solutions use their Oracle databases and what kind of performance, configuration and utilization problems exist. Our experience and the mountain of data from thousands of analyses has enabled us to refine our storage and server solutions, our software and our professional services to address the biggest challenges users face with their Oracle database environments.
So, what’s next, you might ask? What are all these best practices, and how can DIET help you?
Please, stay tuned to this blog and I will write next about the number 1 performance problems we find and the easy fix that can double your Oracle database performance in seconds.
In the meantime, why not give DIET a try for yourself? Collect a few AWR reports from your most troubling Oracle database, and head over to our “Lite” version of DIET – It’s completely free and we’ll send you a high-level analysis of your Oracle database in just a few clicks. And, if you would like more information and a deeper understanding of the health of your database, how to improve performance, increase utilization or lower your costs, and how HDS can help you get there, please contact us.