Funtion of cluster software on OPS(RAC)?

Question

I Want to Know the function or reason of need to use cluster software on OPS(RAC)
As I Know, if I use oracle with HA Architecture,
It's true that I must use cluster software to get following function
1.health check
2.IP and resource take over...etc
But In parallel db environment,
Because there is TAF(tnsnames.ora) Function, There is no need to instance takeover.
Is it true?
If it is true, Cluster software's software has no need to ip take over and
because database is shared by multinode, there is no need to volume take over.
(whether using raw device or cluster file system)
Summary of what I want to know.
===============================
Q1)the reason that cluster software need on ops(rac)
Q2)is there need to take over instance in ops(rac)?
if it need, what is the case?
Q3)the difference of cluster software between HA enviroment and RAC(OPS) environment.
Q4)is there any function that don't need with ops ,but need when used with HA environment?
Q5)if we use cluster file system solution than have buffer cache manage fuction
(sync for consistency) on two node without parallel server option,
is there any problem on processing same database?
if any, what is the problem?

Answer 1

Summary of what I want to know.
===============================
Q1)the reason that cluster software need on ops(rac)
>
The cluster software is required to facilitate internode communication between nodes in the cluster through the cluster interconnect devices. Remember, 9i RAC does not use buffer pinging for block passing it utilizes cache fusion
Q2)is there need to take over instance in ops(rac)?
if it need, what is the case?
>
sorry I don't understand the question.
Q3)the difference of cluster software between HA enviroment and RAC(OPS) environment.
>
This is dependant on the OS/Clustering S/W you are using. For instance, 9i RAC on Windows 2000 requires that you do not install M$ clustering at all and use the clustering tools provided by Oracle. On the other hand, 9i RAC onCompaq Tru 64 Unix and Open VMS rely on the native Clustering of these OS's.
Q4)is there any function that don't need with ops ,but need when used with HA environment?
Q5)if we use cluster file system solution than have buffer cache manage fuction
(sync for consistency) on two node without parallel server option,
is there any problem on processing same database?
if any, what is the problem?
>
Technically, you can use Oracle in a cluster without 9i RAC under the following conditions:
You use a clustered file system, or RAW Devices.
Only one Instance per node can access the database at any time. In other words it could be used for failover from one node to the other, but it can't be used for scaling.
For more detailed info, you should read all of the relevent Oracle white papers on 9i RAC, and if you can find them, the release notes for 9i RAC on your OS.