The question of database resiliency comes up frequently on O-box proposals – I’ve recently blogged about it, but thought I’d try and shed a little light on the trade-offs from an O-box appliance perspective.
Firstly, let’s recap these database options:
- Real Application Clusters (RAC) is the most sophisticated version of the Oracle Database and allows multiple instances to run concurrently on the same data on different nodes (2 in the case of ODA).
- RAC One Node (RON) is a variant of RAC when the RAC cluster is still provisioned but a database is only allowed to be active on one at any given time (except during switchover during maintenance periods).
- Enterprise Edition is the traditional single-instance Oracle Database which, on ODA, is always running on ASM. With O-box it can be started on either node, but to move it from one node to the other requires a brief outage.
Secondly, there’s Oracle’s “10 day fail-over rule” which in essences says that you can install Oracle software, without requiring a licence, on a second node attached to the same storage to act as a standby. This second node can only be activated on 10 separate occasions or up to a total of 10 days per year, and obviously if the node is used for anything else the software must be licensed. See the Software Investment Guide for details.
Whilst as I pointed out in my blog the it costs nearly 3x the licence price to run RAC as compared to just Enterprise Edition, once you consider the database as part of the other Oracle licences required for an O-box SOA appliance, the difference is not so marked. E.g. for a recent O-box SOA 12c quotation the RAC options added the following:
- RAC One Node: +4%
- RAC: +41%
So as you can see, in a real-world scenario maybe RAC is not quite as expensive as it might at first appear. The key questions you need to answer are:
- How important is it to avoid unplanned outages (if availability matters above all else then you should choose RAC)?
- How easy it is for you to get agreed downtime and how sociable are those hours? E.g. if you can get downtime on most weekdays from 10pm your view will be different than if it’s only possible on a Sunday morning from 3-5am with 2 months’ notice.
Note that there are two factors that would apply to more general database systems which don’t really apply to the ODA / O-box combination:
- complexity of installation: whilst it’s definitely true that installing Grid Infrastructure (Clusterware & ASM) is more complex than just a single-instance database, ODA handles all that for you.
- management of fail-over: again O-box has this one covered – the database (whether EE, RON or RAC) and WebLogic is automatically set up to work together in a pre-tested, robust configuration.
Remember that as of today, whilst highly available, it is not possible to keep an Oracle SOA Suite platform fully up to date and yet maintain 100% availability. The same is true for the Oracle Database Appliance – sooner or later you need an Infrastructure patch which cannot be applied in a rolling manner. Therefore if an unplanned outage would cause a large financial loss or impact on your reputation then that is probably sufficient justification for RAC; otherwise RON or Enterprise Edition is likely to be more than adequate.
In a future post I will describe the typical behavior following ODA node failure on O-box for the 3 types of database configuration so stay tuned!