In a previous life I was the Exadata UK Team Lead within Oracle Advanced Customer Services. I spent a lot of time installing, patching and generally talking about Exadata for UK customers, as well as training other Oracle engineers to support it. I also used to deliver high-level Exadata training to Oracle sales and delivery staff and the management teams – I wrote a presentation which involved hauling members of the audience up on stage and getting them to throw coloured plastic balls around to demonstrate the way that Smart Scan works. You had to be there.

I (once) knew Exadata very well, both from the technical perspective and also in terms of what it’s actually like to try and implement at a customer site. I have the utmost respect for the people in the Oracle Development organisation that created the product – and an equal amount of respect for those hardy souls in the Oracle Support and ACS organisations that go out there into the battlefield and perform miracles on a daily basis.

I loved the Exadata product when I worked at Oracle. I thought it was very clever and a real example of innovative thinking. However… I do have a couple of issues with it. Firstly I disagree with the strategy of describing it as an OLTP solution, or as a Database Consolidation solution. It was designed as a Data Warehousing solution and that is what it is, in my opinion. I cannot see how it can be accurate to describe it as “the strategic platform for all workloads“.

Secondly, I believe that one of the fundamental drivers behind Exadata’s most heralded feature “Cell Offload Processing” was the slow speed of disk. All Flash storage arrays make those slow disk speeds a thing of the past. I guess time will tell on that one.

And finally, the marketing of Exadata leaves me feeling slightly embarrassed. With thanks to Kevin Closson, here is just one of many great examples:

Manufacturing resource planning (MRP) processes runs significantly faster. To quote Garmin, “Each day at 19:00 M-F we run the Garmin TSO/PMA Certification Set program. Before Exadata, the program took an average of 5 hours and 2 minutes to run. After Exadata, the program regularly takes 1 hour 25 minutes. By my calculations, that is a 475% improvement.”

– Garmin International Inc. Oracle Exadata Database Machine Technical Case Study (page 2)

Reducing the runtime from 5 hrs 2 mins to 1 hr 25 mins is a 72% improvement. That is a fantastic achievement, truly an impact to be proud of. Why describe it as a 475% improvement? It just doesn’t make sense to me.

(And you know, it’s not just Oracle that does that – I see it increasingly in our industry: “By adopting technology X we were able to achieve a 50000X performance increase”. Here’s what I think: publish the underlying facts and let them speak for themselves.)

After leaving Oracle I wrote a blog series called “The History Of Exadata” which outlines how the database machine initially came to the market and describes the underlying hardware and software technologies.

You can also find any blog posts tagged with Exadata by clicking here.


4 Responses to Exadata

  1. Alexey says:

    What was your calculation way for 5hrs 2mins and 1hr 25mins which gave 72% improvement?
    If we take 100% as a 2-fold improvement here we have 302mins vs 85mins and it’s really more than twice.

    • flashdba says:

      First of all let’s convert to minutes to make it easier:

      Old time of 5hrs 2mins = 302 minutes
      New time of 1hr 25mins = 85 minutes

      So, working out the new time as a percentage of the old: 85 / 302 = 0.28 or 28%, i.e. it ran in 28% of the time.

      Expressing that as an improvement: 100% – 28% = 72%

  2. Alexey says:

    ok, thanks – how then did they get 475%? 😉

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: