Blog

le 12/09/2014
Eric Leygonie
Home page @en

ACFS on Exadata : a gateway between 10GR2 and the Cloud ?

Robert G Freeman alerted the Oracle blogs world there : Grid Infrastructure 12.1.0.2 bring us a small but quite important feature for Exadata appliance , ACFS is now supported on Exadata stack !
For those of you who doesn’t now what is ASM Cluster File System , more details in the Administration Guide.
In the overview you can read the main facts :
» ACFS is a multi-platform, scalable file system, and storage management technology that extends Oracle Automatic Storage Management (Oracle ASM) functionality to support all customer files.
Oracle ACFS supports Oracle Database files and application files, including executables, database data files, database trace files, database alert logs, application reports, BFILEs,and configuration files.
Other supported files are video, audio, text, images, engineering drawings, and other general-purpose application file data.
Oracle ACFS conforms to POSIX standards for Linux and UNIX, and to Windows standards for Windows. »

Until now ACFS on Exadata was not supported. Best of practise was to remove ressouces from the CRS stack (ora.registry.acfs) . These days were gone , MOS note 1326938.1 cleary stated that in bold :
ACFS is now supported for Grid Infrastructure 12.1.0.2 or later. See Mos Note: 888828.1 – Section « Automatic Storage Management Cluster File System (ACFS) »
Furthermore, we’ve got support for 10.2.0.4 or 10.2.0.5 database files on ACFS, extract from MOS note 888828.1 :

Automatic Storage Management Cluster File System (ACFS)

Creating AFCS file systems on Exadata storage requires the following:

Oracle Linux
Grid Infrastructure 12.1.0.2

Database files stored in ACFS on Exadata storage are subject to the following guidelines and restrictions:

Supported database versions are 10.2.0.4, 10.2.0.5, 11.2.0.4, and 12.1.
Hybrid Columnar Compression (HCC) support (for 11.2 and 12.1) requires fix for bug 19136936.
Exadata-offload features such as Smart Scan, Storage Indexes, IORM, Network RM, etc. are not supported.
Exadata Smart Flash Cache will cache read operations. Caching of write operations is expected in a later release.
No specialized cache hints are passed from the Database to the Exadata Storage layer, which means the Smart Flash Cache heuristics are based on I/O size, similar to any other block storage caching technology.
Exadata Smart Flash Logging is not supported.
Hardware Assisted Resilient Data (HARD) checks are not performed.

First time 10G release appears as supported on someway on Exadata, that could be the key gateway missing on this system. Without that, no way to consolidate all these olds 10G DBs running this criticals sofware on the Exa stack. No way to discard these old and expensive servers. Now even if performance are not at maximum due to lack of smart scan capabilities with ACFS, We could expect more than correct results on the Database Server machine full of CPU Power, memory and even flash cache. Lets have this dream !
Requirements for running ACFS on Exadata are :

You need to use Oracle Enterprise Linux
You must setup Grid Infrastrucure 12.1.0.2 or later
You have to ask a patch resquest for bug 19136936 if you forward to use HCC (Hybrid Column Compression)

The last and not least point to get this gateway running is to verify that Oracle Database RAC Sofware 10.2.0.1 with patchset 10.2.0.4 or 10.2.0.5 could be installed and relink on the Oracle Enterprise Linux release packaged and certified on the Exadata machine. As this was never certified it could be painful and laborious. We’re going to try that as soon as we can, challenge Oracle with it if neeeded and share the result with you, stay tuned.