ODC Appreciation Day : Oracle Exadata Database Machine

Those that know me well, will know about my “appreciation” of the “Oracle Exadata Database Machine“, more commonly known as “Exadata” 🙂

So this will be my contribution to ODC Appreciation Day formally known as OTN Appreciation Day, a great initiative by Tim Hall aka Oracle-Base.com.

You can see a summary of last year’s blog post here:
OTN Appreciation Day : Summary

The very first Exadata, was the V1 model, the hardware by HP and the software by Oracle.  I still remember being very excited by this in my previous employment at Auto Trader and trying very hard to convince them to get one 🙂

I, of course, became an instant fan of the brawn hardware with smart software, Oracle labelling as “Hardware and Software optimised together“.

Oracle’s partnership with HP only lasted a year with Oracle switching to Sun on the V2 model, when shortly after Oracle then brought Sun in 2010.  This is when Oracle switched from the V models to X models, with the initial models being the X2-2 (2 socket) and X2-8 (8 sockets).

I still remember this old video “Oracle Exadata. Are You Ready?” that I played at an internal Auto Trader conference which was about sharing knowledge, interesting new things, etc:

Exadata has come a long way since the initial release that was aimed at being a Data Warehouse to a Full On OLTP, Data Warehouse, mixed load, consolidation platform, etc with “record-breaking” IOPS and scan rate!

My favourite feature is the smart scan, the ability to off load data intensive SQL operations from the database servers directly into the storage servers, mitigating the need to pull lots of data from storage to database server.  Yes you can have very fast All Flash Storage, but the network to ship all this to the database server becomes the bottleneck and the compute to filter the data on the database server.  Exadata does this at the storage server meaning only the rows and columns that are directly relevant to a query are sent to the database servers.

Another one is storage indexes where the min and max values are stored of a column in 1Mb chunk in memory to allow for unnecessary I/O to be avoided when it’s known that block of data doesn’t meet the predicate condition.

I didn’t manage to convince Auto Trader, however I have since been very fortunate in my current employment at Version 1 to have worked on Exadata since 2014 from the Exadata X2-2 through to X5-2.  I do really appreciate these “Engineered Systems” for the Extreme Performance, Reliability and Availability.  The whole concept of being “Engineered” and the whole stack optimised, really works and the fact that all Exadatas are the same hardware makes you appreciate their supportability.  Even patching them with patchmgr is pretty much a doddle these days! 🙂

For more info, visit the following site:
www.Oracle.com/Exadata
https://en.wikipedia.org/wiki/Oracle_Exadata

Tuesday 10th October 2017

Happy ODC Appreciation Day! #ThanksODC #ThanksOTN 🙂

If you found this blog post useful, please like as well as follow me through my various Social Media avenues available on the sidebar and/or subscribe to this oracle blog via WordPress/e-mail.

Thanks

Zed DBA (Zahid Anwar)

Advertisements

Assess Performance using Calibrate on Exadata

For those who are fortunate to have an Oracle Exadata Database Machine, may wonder if their Exadata meets the IOPS/MBPS as per the technical specifications.  Well with the command CALIBRATE in CellCLI, you can run raw performance tests on the cells’ hard disks and flash drives, enabling you to verify the disk/drive performance:

[root@v1ex1celadm01 ~]# cellcli
CellCLI: Release 12.1.2.3.4 - Production on Tue Jun 13 19:02:05 IST 2017

Copyright (c) 2007, 2016, Oracle. All rights reserved.

CellCLI> calibrate force;
Calibration will take a few minutes...
Aggregate random read throughput across all hard disk LUNs: 1823 MBPS
Aggregate random read throughput across all flash disk LUNs: 9973 MBPS
Aggregate random read IOs per second (IOPS) across all hard disk LUNs: 3002
Calibrating hard disks (read only) ...
LUN 0_0 on drive [8:0 ] random read throughput: 152.00 MBPS, and 243 IOPS
LUN 0_1 on drive [8:1 ] random read throughput: 157.00 MBPS, and 246 IOPS
LUN 0_10 on drive [8:10 ] random read throughput: 161.00 MBPS, and 253 IOPS
LUN 0_11 on drive [8:11 ] random read throughput: 157.00 MBPS, and 251 IOPS
LUN 0_2 on drive [8:2 ] random read throughput: 157.00 MBPS, and 244 IOPS
LUN 0_3 on drive [8:3 ] random read throughput: 158.00 MBPS, and 245 IOPS
LUN 0_4 on drive [8:4 ] random read throughput: 156.00 MBPS, and 248 IOPS
LUN 0_5 on drive [8:5 ] random read throughput: 161.00 MBPS, and 250 IOPS
LUN 0_6 on drive [8:6 ] random read throughput: 159.00 MBPS, and 252 IOPS
LUN 0_7 on drive [8:7 ] random read throughput: 158.00 MBPS, and 251 IOPS
LUN 0_8 on drive [8:8 ] random read throughput: 157.00 MBPS, and 251 IOPS
LUN 0_9 on drive [8:9 ] random read throughput: 159.00 MBPS, and 254 IOPS
Calibrating flash disks (read only, note that writes will be significantly slower) ...
LUN 1_1 on drive [FLASH_1_1] random read throughput: 2,157.00 MBPS, and 280525 IOPS
LUN 2_1 on drive [FLASH_2_1] random read throughput: 2,156.00 MBPS, and 274304 IOPS
LUN 4_1 on drive [FLASH_4_1] random read throughput: 2,158.00 MBPS, and 282083 IOPS
LUN 5_1 on drive [FLASH_5_1] random read throughput: 2,160.00 MBPS, and 287786 IOPS
CALIBRATE results are within an acceptable range.
Calibration has finished.

CellCLI>

 

The CALIBRATE FORCE, allows the test to run when CELLSRV is still up, this is acceptable if there is no user workload.  It is therefore recommended to not run during normal operations.  Without the FORCE, the CELLSRV must be shut down.

PLEASE NOTE: This is a single run on a single storage cell, you will need to run on all storage cells in the Exadata Machine to get the total IOPS/MBPS.  You can use dcli to run this across all the cells 🙂

If you found this blog post useful, please like as well as follow me through my various Social Media avenues available on the sidebar and/or subscribe to this oracle blog via WordPress/e-mail.

Thanks

Zed DBA (Zahid Anwar)