What is Oracle ACFS and ADVM

Question

Hi, I am trying to get my head around exactly what Oracle ACFS and ADVM are in relation to ASM ( the Oracle manuals are less than clear ! )
Q1. To my mind, ACFS extends the ability of ASM to manage and present file systems which are not simply provided from a single server but where a file system itself could be distributed over a cluster of servers - is that correct ?
Q2. ADVM on the otherhand is stumping me a bit. I know that ACFS sits on top of ADVM but I am trying to workout what ADVM gives you above the usual volume management capability of ASM in the first place ?
Q3. What is ASMLib for ?
Q4. I believe oracleasm is the command line interface to ASMLib - is that correct ?
any help would be warmly appreciated,
Jim

Answer 1

In regards to OCFS and ACFS. I understand that they are different Cluster File systems, both developed by Oracle but with OCFS being the older open source, general purpose Cluster File System and ACFS being the licensed one they have developed in 11gR2 for GRid Infrastructure / ASM to use.
OCFS and ACFS are separate products by different teams for different purposes. One should not compare the two as they have different pros and cons.
Q1. Is ACFS, Oracle's default for Grid Infrastructure ( ASM + Clusterware ) from 11.2 onwards ? The image you previously referred to would appear to show this ( or does the box showing the use of a 3rd party file system include OCFS as a possibility ? )
I believe that it shows that ASM is modular with defined call interfaces allowing 3rd party integration - and that this has no bearing to OCFS specifically.
Q2. The same image appears to show RAC only using ACFS. Again is this the default from 11.2 onwards ?
ACFS is recommended when using ASM and a cluster file system is needed. You do not need any additional moving parts to have a clustered file system. But this does not mean that other cluster file systems are not supported.
Q3. Prior to 11.2 what did RAC use as it clustered file system ? Was it OCFS or were there a number of options ?
In the days of OPS (Oracle Parallel Server), a cluster was build on raw devices (seen by all nodes), with the o/s providing a DLM (Distributed Lock Manager) to allow concurrent access by different processes on different servers to the same disks.
This was very clunky and had minimal management tools. I recall requesting a DBA to add another scsi disk to a tablespace and he accidentally entered the disk size to add as 2MB instead of 2GB. The disk was toast as we had no means to reclaim the rest of the storage, and removing the disk from the tablespace was not possible. OPS also had other issues like a TCP Interconnect and so on.
When RAC was introduced, Oracle addressed a lot of the OPS cons. Like requiring a 3rd party DLM. Like having the DBA directly dealing with raw scsi devices. Like using TCP for the Interconnect. Etc.
ASM was introduced to do the storage management for the database stack. Oracle RAC has an underlying component called the cluster (CRS) stack - and this specifically needs cluster storage for the OCR and Voting files (no db storage). As ASM was part of the db stack, you had to use something else for the CRS stack's storage. That something else being OCFS for example, or a raw scsi devices, etc.
With 11gr2 Oracle extended ASM to support the CRS stack and provide for storage for the CRS. Which is why one now installs ASM with CRS (and it runs as the o/s grid user), and not with the db s/w (running as the oracle user). As part of the feature set of providing CRS storage, cluster file system storage is also provided (for example, to install the db stack in a single shared cluster Oracle home).
ASM's new features simply continued the trend of reducing CRS/RAC dependency on 3rd party components - even if said 3rd party (external) component happens to be from a different division from the same company and called OCFS.
By the same token, the entire CRS/RAC s/w stack has been further modularised, allowing you to integrate your own stuff with it. You can for example use CRS to run a series of Apache web servers as a cluster, with load balancing, redundancy and failover. Or if you feel the need, build your own cluster file system driver for ASM.