Amazon and Oracle: Fluffy white bear-traps?
It really is buyer beware with this service
Published 11:10, 08 February 11
Amazon’s Oracle offering illustrates the commercial dangers of software licensing in the cloud. Organisations should be wary of unbudgeted costs and Oracle license disputes.
As expected Amazon this week announced the release of Oracle 11g services as part of its Relational Database Services (RDS) portfolio. This is potentially quite useful to the market and is a good indicator of future trends. We’ve talked about cloud computing and Oracle before but this is one of the first instances of it becoming a reality.
For Oracle it helps fend off cloud based competition from “open-source” vendors; for Amazon it underlines the enterprise strength and potential of its offerings; and for Oracle clients it creates a meaningful way of trying “platform as a service” scenarios without introducing unnecessary data silo complexity. In other words, Oracle data can be migrated to and from the RDS platform with “relative” ease.
What’s on offer?
It was confirmed with Amazon that Oracle licensing options would include: -
- Transfer of existing Oracle licenses
- Buy new Oracle licenses (no surprise)
- Rent per hour
The option to buy more licenses from Amazon (reserved DB instance model) is effectively Oracle’s generic term licensing model whereby the appropriate licenses are, for a one off payment, available for 12 or 36 months.
Oracle term licensing has been available for many years but is not particularly popular with enterprise clients as it is typically either not flexible enough or cost effective in comparison to Oracle perpetual licensing.
Amazon says it will be significantly cheaper than the rent per hour model (on -demand database instance model). It’s fair to assume that the hourly rental model will involve relatively high licensing costs. As the Oracle On-demand team will be providing the tech support for Amazon Web Services (AWS) the Amazon model for Oracle license “rental” as above should be simple to administer.
This brings us to the Amazon “Bring Your Own License” (BYOL) model. This is where existing Oracle clients can use existing Oracle license grants and apply them to the Amazon RDS service with no additional cost.
This is where the potential bear-traps exist. Although Amazon has not specified it in any way, we would expect normal Oracle licensing practices to apply. Hence it will be the obligation of the client to prove licensing compliance across its estate, and this will include Amazon RDS services for Oracle.
I guess Amazon will be obliged to inform Oracle as to which clients use its Oracle services including when, which products, volumes and duration. Oracle License Management Services (LMS) will then periodically audit the client.
The client will then need to demonstrate that it had the appropriate licenses surplus at that moment. If this cannot be demonstrated I’d guess that the client gets a bill based on Amazon/Oracle rental style licensing.
The real issue here is that in our experience (for the past 10 years we have independently advised hundreds of clients on Oracle compliance issues) very rarely can an enterprise client demonstrate its actual Oracle deployment pattern compared to Oracle license grant without a specific project being launched. So what chance will any of us have a being able to prove that there was enough surplus of licensing to cover AWS potentially going back 2 or 3 years?
BYOL sounds attractive and simple to enterprise clients, but unless Oracle licensing control is world class, don’t be surprised if the Oracle software costs of AWS manifests itself as unplanned and unbudgeted cost down the line.
We’ll come back to this as Amazon announces more.