Uncategorized

by Matt Hayward

Given a description of gas dynamics and the atmosphere, you would be hard to pressed forecast tornadoes. The term emergence denotes the phenomena of surprising behaviors arising in complex systems.  Modern storage systems are complex, and full of emergent behavior that make forecasting application I/O performance fiendishly difficult.

People have been asking me a number of questions about Delphix and technologies in the market place

Delphix allows companies to make copies of Oracle databases in minutes with  almost no storage overhead by sharing all the duplicate data blocks between database copies. Sharing duplicate data blocks massively reduces storage consumption, frees DBAs up from the repetitive work of copying files , allowing them to concentrate on innovative work and most importantly accelerates project development by eliminating the time and resource consuming bottleneck of provisioning databases for development, QA and reporting. With Delphix the databases can be provisioned in  minutes to any point in time from the source database allowing one to easily go back before an incident on production to retrieve data before a problem or stand up a full copy of production at a time that production had a problem allowing easy analysis of intermittent issues.

Mogens presenting in a bathrobe before 800 people

Lots of good presentations, discussions and interactions with people at RMOUG. Kudos goes out to all the RMOUG team for putting on an awesome conference. Thanks to Tim Gorman, Kellyn Pot’vin  and all the dedicated volunteers.

My favorite presentation  was by Jordan Meyer of Rittman Mead consulting. He did social media analysis with R.  His R source code is available at

https://github.com/jordanmeyer/oracle-sna-hol

I’ve mentioned lab 128 before but I have to mentioned it again. Every DBA should know about this tool. It’s the fastest way to connect to an Oracle database and get some hard core performance information nicely visualized graphically.

Lab 128 starts up in seconds, like 1 or 2 seconds. Its super fast and light weight but powerful.  My older version only took 10M of RAM. The newest version is about 47M but still as fast as ever.

Lab128 can be used to monitor 15-20 instances. Lab 128 keeps  13-14 hours of performance data in memory, and last 60 days on the disk. Monitoring that many instances for that long takes around 800-900 MB of RAM and well less than  10% of the CPU on a laptop.

by Matt Hayward

One of the less obvious advantages of database virtualization is a reduction in the read I/O issued against the underlying physical storage (henceforth called “the SAN”) that ultimately stores the data for virtual databases.

In practice, Delphix prevents around 60% of all non-production database I/O* from ever being issued to the SAN with the Delphix cache.

If you’ve used Tanel Poder’s snapper script then you probably want to know about the latest release, and webinar coming up this Wednesday.

I’ll be presenting twice at RMOUG next week on the subject of how to manage creating copies of Oracle databases and creating copies through thin provision cloning  as well as database virtualization.

• Feb 12, 2013 11:15 room 402
• Feb 12, 2013   1:15 room 407

The first talk covers the technologies in the industry

The second talk concentrates primarily on Delphix database virtualization but will hit on the other technologies as well.

If you are going, I highly recommend the guidebook app for the conference. It’s awesome. I wish Oracle had this at Open World. It makes browsing presentations and building a schedule a piece of cake. You can install it by either of these methods

Ask Jonathan about Delphix at

http://jonathanlewis.wordpress.com/2013/02/06/delphix/

Jonathan Lewis has graciously accepted an offer to come out to sunny California next month  and spend a few days at Delphix !  Jonathan will be putting  Delphix through it’s paces.  I’m super excited to have Jonathan  test Delphix and then learn about what he discovers. We are planning on co-presenting a webinar on the findings as well as Jonathan will be blogging is findings.

If you have questions for Jonathan about Delphix or scenarios you’d like him to test out then ask Jonathan to  test them out by commenting on  his blog post  on Delphix:

I was doing some work for a client, and LOB processing seemed incredibly slow.  After some poking around, it appears that LOB processing is quite sensitive when it comes to flashback mode in the database. Test bench is a simple table with three clobs. 4kb will be insert into each clob in 5000 individual inserts. The "wait" option is applied to the commit to ensure that the standard PLSQL commit optimizations do not come into play.

Database is 11.2.0.1 running in flashback mode.