Over the last couple of years I’ve been doing talks about running Oracle databases in the cloud. In both my talks and articles I refer to the Licensing Oracle Software in the Cloud Computing Environment document. This morning I was reading a post on a mailing list and someone mentioned the document had been updated on 23-Jan-2017 and contained two rather interesting changes.
The document now explicitly mentions the difference between how vCPUs are licensed on different cloud providers. On AWS a vCPU is one Intel hyper thread, so you need 2 vCPUs to make a real core. Azure does not use hyper threading on their boxes, so 1 vCPU equals a real core. The previous version of the document did not make this clear, so it read like you were paying per thread on AWS, even though people who used cloud-savvy partners understood this issue and paid correctly (vCPUs/2 on AWS).
The document now says,
“When counting Oracle Processor license requirements in Authorized Cloud Environments, the Oracle Processor Core Factor Table is not applicable.”
Just digest that for a moment. The intel core factor is 0.5, so an 8 core physical box requires 4 cores of licensing. Now on the cloud, an 8 core VM (16 vCPUs on AWS or 8 vCPUs on Azure) requires 8 cores of licensing.
On the 23-Jan-2017 the intel core factor was removed from the cloud licensing calculation and overnight your cloud licensing costs doubled! WTF! 🙁
Update: For the new AWS bare-metal service, the core factor *should* still apply.
The same person also pointed out that in a MOS Note (Doc ID 2174134.1), last updated on 20-Aug-2016, Oracle pulled support for the Oracle Multitenant option from AWS EC2. WTF on a bike! 🙁 I assume they mean both non-CDB and Single Tennant (Lone-PDB) are still supported.
The reaction to this is going to be really bad! It’s getting really hard to remain an Oracle fanboy these days!
If you have been to one of my cloud talks over the last couple of years and you are basing your opinion on something you’ve heard me say, please remember things change. For those people I presented to at the UKOUG Database SIG on Tuesday, I’m sorry, but I was 2 days out of date on one slide. I’ve updated my slides and articles to reflect this change!
This is all so completely depressing!
PS. I’m not saying this policy document overrides your contracts. Just saying this is the most recent policy document produced by Oracle!
PPS. You might want to take a look at page 19 and the addendum on page 23 of this copy of the NoCOUG Journal.
39 thoughts on “Oracle’s Cloud Licensing Change : Be Warned!”
Whatever happened to contractual arrangements?
Normally, buyer and seller agree on a price and that’s it.
How can Oracle ask for more money based on a non-contractual document that the buyer has never signed?
(Oh by the way, I just doubled the rent on the property I am letting…)
Stew: I imagine existing stuff, as long as you have it down on paper, will continue until the next review cycle. It will be new requirements and services where the licensing cost is baked in that will be hit.
Unfortunetly, this is not how it works. These non-contractual papers are used as a basis for Oracle audits and they will ask for the money once you get audited. At that point, management has to decide if they want to obey or file a charge (after Oracle terminates the contract). Not a lot of companies are willing to sue Oracle, so while all this is just on a piece of paper that states that it is not part of the contract, the effect is that it does apply in the real world.
I feel the upcoming urge to climb a crane with a resist banner…
Marcus: I think you’ve misunderstood what I mean by, “have it down on paper”. 🙂 Of course I mean contractual. Whenever you want to do anything with Oracle (or any other company for that matter), but especially when it involves virtualization or cloud, you must have it all down on paper (in contractual form) before you start doing anything.
If you have a contracted price, it will not be affected by this policy change. Caveats apply.
It is never a good idea to assume you know what the rules are and wing it.
It sounds as if Oracle have never heard of PostgreSQL, or Azure, or any of the other reasonably easy migration paths. It’s this sort of action that should make people really nervous about getting too involved in PL/SQL, or other techniques that make a migration difficult.
Do what we do. Move your customers to PostgreSql with EDB.
It is indeed becoming difficult to remain an Oracle fanboy. Though the tech is good, their business practices are increasingly draconian.
More than once I have considered moving to the open source side.
I agree that Oracle is shooting themselves in the foot… or possibly parts north of there. I recently gave a Webinar presentation on an Oracle-based solution for a corporate application requirement to a team from our CIOs office. After I was finished, it became obvious that I might have well presented my idea to an empty room.
A few weeks later I found out that Oracle has become a curse word at the CIO level. A little over a year ago, the company I work for split into two completely independent organizations. When this happened, Oracle forced our half of the company to purchase new licenses for all of our Oracle products despite having clear documentation that the instances brought over to our side during the split had already been licensed (For all I know Oracle may well have forced the *other* half of the organization to repurchase licenses as well). Our CIO went ballistic and mandated that anything in our company that could be moved off of Oracle should be moved ASAP and that new database requirements should consider Oracle only as a last resort.
I’m sure some salesperson really slam dunked their quota for that one quarter, but they really poisoned the well for future years.
Oracles mentality here is simple, squeeze the lemon dry. When the market revolts and migrates to a cheaper solution (e.g. mysql) buy the competition out and then squeeze that.
Before Oracle, hostile takeovers weren’t practiced in the tech industry but Oracle changed that (e.g. PeopleSoft). There are some cases where this will come back and bite them (e.g. Workday) but I have very little doubt that they will buy MongoDB and a lot of other infrastructure level providers to take over additional markets despite kickback.
Whilst this is altogether terrible and sad news, it should come as #NoSuprise. Most DBA’s I know *love* Oracle for its features (nitpicking aside). However, all of my clients *hate* Oracle corporation. It is seen as a necessary evil. Oracle’s pricing model is broken. Every time a client performs a hardware refresh (~4 years) their Oracle costs go up simply because of Intel’s increased core count – not because of valuable new features. Many clients do not need the advanced features and would be unable to sustain the additional complexity these require. Ask yourself, how many Oracle customers do you know, whom to be honest, are really only using Oracle 10 features – regardless of the version installed? Small to medium sized organizations often run Oracle because application software has specified it as a requirement. For many shops, the licensing changes of SE2 caused them to look elsewhere and this removal of the core factor table from cloud calculations will have similar results.
If all of this causes you to say “But the answer is the cloud”. I can only partly agree. An argument might be made that the cloud provides more exact resource allocation and metering with the potential to pay only for what is truly used. Possibly, but it also distracts organizations from the question of value, causing then to do bad math. It is similar to focusing only the monthly payments for a car and not the full cost of the vehicle. I would also make a counter argument that most software is not sufficiently resource predictable and an organization’s daily business needs even less so. Cloud costs may be greater or fluctuate more than anticipated. By way of analogy, ask yourself if last month’s cell phone usage and the resulting bill were fair, its cost/value just right?
For small shops the cloud is a good thing, these organizations do themselves a great disservice when operating their own IT infrastructure, often very, very badly. I would contend that large shops who have already invested in significant infrastructure did so to meet stringent business SLAs that the cloud as yet cannot meet and perhaps never will. This leaves the vast middle ground where yes, many organizations will move to someone’s cloud or use cloud-like features. But consider that whenever a platform migration occurs software is often updated or rewritten to take advantage of the new environment’s capabilities. Younger developers, fresh to the challenges of IT, are now often more versed in nosql and alternative solutions than they are with our trusty and beloved relational model and SQL.
What do you think the future will hold, given that organizations are increasingly questioning every component of their infrastructure stack while Oracle gets more and more expensive? Add to this, the fact that developers now question the very requirements of ACID. Mongo, Postgres, Casandra, Redis and AWS Aurora are popular for good reasons. (But I confess the phrase “optional durability” gives me great stress). We live in interesting times.
For the record, I love the Oracle database. I have used it most of my adult life and it has enabled me to pay my mortgage and have a reasonably successful career. With rare exception, I always recommend it. But I do get tired of rationalizing to customers Oracle’s expense and apologizing for Oracle’s bad behavior’s.
David, Sam, Jared, Matthew, Shai, J:
I agree with pretty much everything you all say! Thanks for the comments.
We need to form a self-help group for victims of abusive relationships… 🙁
Oracle of late are behaving like the schoolyard bully. They’re behind the curve in cloud compared to the likes of Azure and AWS and now they’re throwing their weight around to try and prevent people moving to those platforms.
We’ve already had the licence included option revoked from Azure (which although unconfirmed i’m pretty certain that was Oracle’s doing), now they double the cost to host on their rivals cloud and expect this to make people think that moving to Oracle’s own cloud is the solution to that problem. Really? If they do this with licence what would they do to cpu/storage/network costs once you are in their cloud – double that too at a moments notice?
I’m also still amazed the whole oracle on vmware licence farce has not made it into court. Surely it’s anticompetitive and Oracle should be facing something similar to happened to Microsoft with the whole internet explorer ruling a few years ago. This new licence issue just further adds to that.
Nobody likes a bully and i think this will backfire badly, we’ve already changed our primary port to SQL Server based on recent behaviour (after 20 years of it being Oracle) and we’re looking at cases for migrating off Oracle.
I hate that this is happening – i love the product but it’s an almost impossible sell internally at the moment.
I want to re-iterate what Sam said “do what we do, migrate to postgres.” We’ve migrated a few DBs to Postgres and the mandate going forward is “any new DB will go on Postgres”. Postgres works great and v9.6.1 added some nice features for PQ. And we get good support from EDB.
For anyone who has had to go through the rectal exam that is an Oracle licensing audit Oracle’s behavior is not news. We’re an Oracle EBS shop – and a fairly small one at that – and it’s a complete and total nightmare dealing with them. Some large corps – notably Mars Inc, who has a nine figure annual Oracle service and support cost regimen – have vigorously fought the vagueness of Oracle’s contractual language as well as the deeply inconsistent manner they apply their per core licensing. If your company is going through a licensing audit and you have good counsel by all means it’s probably worth fighting them to back down on their draconian tendencies, even if Oracle has many, many more lawyers than you do.
There’s something interesting about Standard Edition in this ‘authorized cloud’ document for those who were still satisfied enough with the 16 threads soft limit of SE2. You can now run SE2 on max 8 vCPU AWS instance. Worse on Azure where it is 4 cores max. And that limit is not only for user sessions but includes background processes as well.
do you have the previous version of the document ?
I usually keep them locally but I don’t saved this one.
Don’t forget that your contractual Processor definition almost certainly includes a link to the core factor table and its applicability to a given processor type (not environment)!
One thing that will be an interesting challenge on AWS would be if you are running dedicated hosts. This is a feature of AWS where you can have a dedicated machine. If you licensed all physical intel CPUs on a dedicated host (https://aws.amazon.com/ec2/dedicated-hosts/), then you are licensing the physical CPUs instead of the vCPUs and using AWS as a hosting partner, not a cloud provider.
Alternatively what is the benefit to Amazon of being an authorised cloud provider ?
Google Cloud is not an authorised cloud provider. A Google vCPU is also a hyperthread on an Intel core so if you were running Oracle on google cloud, then you would have double the vCPUs.
Oracle support are probably going to say if a bug is raised that it needs to be recreated on physical hardware whether on AWS, Azure or Google so there seems to be zero benefit to being an authorised cloud provider. Maybe AWS and Azure should just demand that Oracle can not describe them as an “Authorised Cloud Provider” so that statement about cores can no longer apply.
I am so sad to read this. Our application is built on top of Oracle. After the SE2 change one got angry … After this change ….. Wow … I don’t know what to say. They (Oracle) are damn good at increasing their prices. But now you really start to hate them.
As most seem to agree, the product is very good, but convincing customers to buy this … Completely different discussion.
I guess they are now “promoting” their new cloud infrastructure (Bare Metal Cloud), but if just half of what they promised at OpenWorld was true …. It isn’t …
One other thing. Can someone please explain the SE/One and SE in the cloud? Why do these exist in the document when only SE2 and EE exists?
“thanks” for the bad news 😉
Why are you not running Oracle on Oracle Cloud today? Clearly Oracle Cloud will be optimized/developed/supported/highest performance to run Oracle software and mostly likely lowest cost so why is this big news unless you don’t want to get the best ROI out of your Oracle software? Oracle now has a very wide range of options to run any Oracle software.
Phil: Nice sales pitch. Pity it’s not true.
The current Oracle offerings in this space have some “issues” and I could not currently endorse any of them. In the future this may change (I hope so), but for now your claims are false. I know this because I’ve used them, reported the flaws and they’ve not been corrected!
Facts beat sales pitch!
Ronan: There’s a link to the old one from the article here.
I’m not willing to redistribute it. 🙂
Trevor: Regarding Google Cloud and core factor, I’m not sure that’s how it works… 🙂
Paul Bullen: Certainly, contractual stuff will override this policy document. Plus, all things are negotiable…
I’ve read and re-read this post and the Oracle docs. Understanding the removal of the factor table, how have the examples (previous and new) changed for AWS? Here’s the example from the previous:
Licensing Oracle Database Enterprise Edition on a single instance of 8 virtual cores (on Intel
multicore chips; see the processor metric definition on the price list), would require 8 * 0.5 = 4
processor licenses (each virtual core is considered equivalent to a physical core).
While the new example:
Example, for Database Enterprise Edition licensing in an Authorized Cloud Environment: Licensing
Oracle Database Enterprise Edition on a single instance of four Amazon vCPUs, where hyperthreading is enabled, would require two processor licenses. (Two Amazon vCPUs are considered equivalent to an Oracle Processor license).
So it appears the old (8 vCPU * 0.5 = 4 proc licenses) is the same as the new (4 vCPU * 0.5 = 2 proc licenses). Or was the old method 8 vCPU * 0.5 * 0.5 (the 1/2 core per thread) = 2 licenses? And I just thought I “had it” last week.
Gavin: The old doc was not clear, but before this update, even though it was not explicitly stated for AWS, everyone did (vCPUs/2)*0.5. On Azure it was vCPUs*0.5 as Azure does not use hyper-threading. The new version of the doc makes the vCPUs/2 for AWS extremely clear, which is good. So in that respect things have not changed. They are just clearer.
The big difference is the statement about the Core Factor not applying for Cloud Providers, so the “*0.5” is no longer part of the calculation *according to this doc*. I have examples here.
As always, contracts are king!
@ Tim. You are referring to a review from August 28, 2015? Facts from 17 months ago-Seriously? #OracleCloud has changed *significantly* in 17 months. Suggest you relook/re-evaluate your research.
while I am so sad to read this. Our application is built on top of Oracle. After the SE2 change one got angry … After this change
Hi Gavin and others.
I challenge what you are saying. The old doc mentions AWS virtual cores. One virtual core in AWS is 2 vCPU’s. You might know this: https://aws.amazon.com/ec2/virtualcores/, and then compare to the instance types.
So on the “old” license agreement you would actually get 16 vCPU’s. License would be 8 virtual cores * 0.5 = 4 processor licenses.
With the new license rules you would also need 4 processor licenses BUT you only get 8 vCPU’s. So in effect only half the power (knowing of course that the two hyperthreads does not exactly equal one core).
In our case (using almost only SE2) this means that we can now only run 8 vCPU instance types! Before we could run up to 16 vCPU’s! This is a dramatic change! You are then forced onto EE if you want more than 8 vCPU’s!
I think this is a good case for a lawsuit. Oracle is clearly favoring their own cloud. As mentioned earlier in this blog in relation to MS and Internet Explorer.
do you know if Oracle EE can be installed on AWS under the OTN Free Developer license?
Reading the Oracle OTN pdf it seems it only covers if you install it on your own desktop for a single user, right?
Can someone tell me how does The Big Red audit your Oracle IaaS? They do it with your concern or even not contacting you?
Plus, Oracle Cloud limit on The fly The core usage or it is just a limit reminder?
Phil: I am referring to my review of 17 months ago because, as I mentioned in my last reply to you, little has changed regarding the points I raised then. I know this because I am using Oracle Cloud regularly.
You don’t have to agree with my comments. You are allowed to be wrong!
Phil: Here is the “relook/re-evalute” you requested.
“This document is for educational purposes only and provides guidelines regarding Oracle’s policies in effect as of January 23rd, 2017. It may not be incorporated into any contract and does not constitute a contract or a commitment to any specific terms. Policies and this document are subject to change without notice. This document may not be reproduced in any manner without the express written permission of Oracle Corporation.”
…so ask yourself, “what did I sign?”, and “how would a court of law interpret it?” and take it from there.
It’s all very much like the Oracle upon vSphere debate, which is slowly becoming the norm in the real world.
In the last few years, whenever there was a request for migration from Oracle DB to other DB, we as ’20’ years of Oracle Fans, would oppose and give the merits of having it for the sheer Value for money. But looking at the change of such policy where you are trying to kill competition and forcing customers to migrate to Oracle Cloud, would not auger well. Customer is king and he should be the maximum choice. Else migration to competitive RDS will flourish.
All it new policy is saying is we are going back to the orginal definitions and fixing them so microsoft and amazon can no longer change thier definitions to get more of the cloud market by making it cheaper to use their service.
In the begining cloud hosting was twice as expensive as in the datacenter.
Then things outside Oracle control made cloud equal or cheaper.
Now we are back to cloud hosting was twice as expensive as in the datacenter.
Oracle has taken back control. This only really effect BYOL.
Paul: I disagree. I don’t think there is anything more I can say about the first paragraph and I’ve not come across anyone yet that would agree with you!
Re: “This only really effect BYOL”
Let’s look at what is out there…
– Azure only allows BYOL. All on-demand licencing was pulled last year, so it affects *all* Oracle on Azure.
– AWS EC2 only allows BYOL, so if you want to DIY Oracle on AWS this affects you, regardless of edition.
– AWS RDS for Oracle only allows on-demand licensing for SE1 or SE2. If you want to run EE on RDS for Oracle you can only BYOL.
– I assume the policy applies to DIY Oracle on Oracle Public Cloud Compute, which is BYOL only too.
– Oracle DBaaS allows On-Demand licensing for all editions.
So the statement, “This only really effect BYOL”, is not really relevant or helpful for the vast majority of scenarios on AWS and Azure, or even some of the Oracle Public Cloud scenarios.
Once again, I’ve not come across anyone with your views on this either.
Anyone can comment on Postgres Enterprise DB’s Oracle database compatibility?
Comments are closed.