How to measure Availability Complexity Operation & Maintenance Costs etc

Question

Dear All
We have multiple configurations for attaining High Availability, Maximum Availability with and / or across data centers for databases. Some employ Active - Active Clustering Solution while Some use Active - Passive Solution. Some employ only OS clusters like Sun Clusters,Veritas clusters. Often as I noticed, a combination of OS cluster say Veritas Cluster as well as Oracle Cluster (Clusterware + RAC) in their database solutions.
In such situations how does one come up with Quantitative analysis of each methodology? What are the inputs one would need and how does these calculations come up? While searching on internet, I have seen at times ACTIVE - ACTIVE Database clusters using VCFS and Oracle Clusters RAC say the availability is some 99.xx% while other approaches are accordingly rated. How does one come up to such quantitative decisions? Could you please advice.
I was asked to come up with such numbers in terms of below mentioned items for a few approaches I suggested.
Availability of the Database
Complexity of Architecture and understanding.
Set up costs.
Operational Costs.
Cost of Maintenance
For your understanding I've suggested below approaches to attain High availability with Production Data Centre.
Approach - I : I will have Active - Passive Database running on node a and b respectively. DB1 will be active on node a and DB2 will be active on node b. In the event of failure they will fail over to the alternative.
Approach - I : I will have Active - Passive Database running on node a and b respectively. DB1 and DB2 will be active on node a and node b will always running waiting for a failover event. In the event of failure they will fail over to node b.
Approach - III : I will have DB1 and DB2 active on Node a. We could configure Dataguard (Logical or a physical standby) on node b in the production data centre.
Approach IV : I will have a two node cluster running RAC and will have ACTIVE- ACTIVE database DB1 and DB2 running on them.
Personally I like the Approach IV. However, for business reasons, we need to make a decision based on quantitative analysis done.
If one of you have already done one such excercise, could you please share yoru experiences here?
Many Many thanks for all your guidence in this direction.
Regards!
Sarat

Answer 1

Sarat Chandra C wrote:
Dear All
Approach - I : I will have Active - Passive Database running on node a and b respectively. DB1 will be active on node a and DB2 will be active on node b. In the event of failure they will fail over to the alternative.
Non Availability of the Database: Time taken to detect failure+ OS failover + starting of the database
Complexity of Architecture and understanding: From oracle side there is no complexity as hardware cluster takes care about this
Set up costs: If management has already decided to have one server for each database, then no extra cost except the cost of license of the hardware cluster. Else cost of extra server.
Operational Costs: (None) (Excluding the dba/sa task to verify the failover)
Cost of Maintenance: (None) (Excluding the dba/sa task to verify the failover)
Approach - I : I will have Active - Passive Database running on node a and b respectively. DB1 and DB2 will be active on node a and node b will always running waiting for a failover event. In the event of failure they will fail over to node b.
Non Availability of the Database: Time taken to detect failure+failover+starting of the database*2 (Since both the database needed to be failover)
Complexity of Architecture and understanding: From oracle side there is no complexity as hardware cluster takes care about this
Set up costs: If management has already decided to have one server for each database, then no extra cost except the cost of license of the hardware cluster. But you would be underutilizing your one server and over utilizing your another server
Operational Costs: (None) (Excluding the dba/sa task to verify the failover)
Cost of Maintenance: (None) (Excluding the dba/sa task to verify the failover)
Approach - III : I will have DB1 and DB2 active on Node a. We could configure Dataguard (Logical or a physical standby) on node b in the production data centre.
Non Availability of the Database: Since there would be a manual intervantion here. Down time would be (Time taken to detect failure+ time taken to activate standby database)*2 (Since both the database needed to be failover)
Complexity of Architecture and understanding: Complex since data gaurd need to be setup.
Set up costs: License for extra two database and for data gaurd.
Operational Costs: (None)
Cost of Maintenance: Verification that log files are continuously shipped and that there is no lag.
Also note if the database is not in maximum protection mode, chances are that you would loose data. If they are in maximum protection mode, then there would be a performance impact to your production server.
Approach IV : I will have a two node cluster running RAC and will have ACTIVE- ACTIVE database DB1 and DB2 running on them.
Non Availability of the Database: None till the time both cluster goes down.
Complexity of Architecture and understanding: Complex since you would be using RAC
Set up costs: Am not sure about licenscing part of RAC
Operational Costs: None
Cost of Maintenance: Nearly double then using single node Database.
Personally I like the Approach IV. However, for business reasons, we need to make a decision based on quantitative analysis done.
If one of you have already done one such excercise, could you please share yoru experiences here?
Many Many thanks for all your guidence in this direction.
Regards!
SaratI have not come up with the exact number since it would depends on the infrastructure and hardware.
Moreover Non availibility of database was only due to the fact hardware is not available it does not include the non availibility of database due to the issue with database. If there is issue with database, then even the passive node would not be able to handle it except only inthe case of data guard.
Regards
Anurag