I was having an email discussion with the folks from DBmarlin, and the question came up about the adoption of the multitenant architecture. Were people using the multitenant architecture, or sticking with the non-CDB architecture for now?
Followers of the blog know I threw my hat in with the multitenant architecture from Oracle 18.104.22.168 onward. We use Oracle 19c for everything of value now (there are a couple of dead projects still on 11.2) and all but one of those projects use PDBs. Suspecting I was not the norm here, I asked some questions on Twitter.
As usual I’m going to start with some caveats. The sample size is small. People who interact about tech on social media my not be a diverse sample. I’m going to act like these results are representative of the wider scene, but they may not be.
Here is the first question.
The fact that only 33% were on non-CDB only was a positive sign in my view. The introduction of the multitenant architecture made a lot of people nervous, and it was not without its problems in the early days. Even so, the combined total of only non-CDB and mostly non-CDB is still 50% of respondents. When you consider Oracle 21c makes the multitenant architecture mandatory, and the next long term release is 23c, there is a lot of work for people to do when they make the switch to 23c. The conversion is simple enough. It’s the testing resource that could hurt people.
Having 50% of people using mostly PDBs or all PDBs is a really good sign, and will make life much easier for them when they come to upgrade to the next long term release.
To dig a little deeper I asked this question.
So just over 65% of people said they are provisioning new Oracle databases as PDBs. That’s very positive, and makes a lot of sense going forward. Why would 35% of people stick with the non-CDB architecture for new databases? Some things I can think of include.
- They are using an older versions of the database, and don’t have the option of using the multitenant architecture.
- They want a one-size-fits-all approach to the database, and will convert everything when they are forced to.
- Vendors don’t support the multitenant architecture. I have one project where I suspect the vendor doesn’t even know the multitenant architecture exists, let alone supports it.
- Internal development teams haven’t caught up with the database version. From my experience, the only thing that was really affected by our move to PBDs was CRON jobs using OS authentication. We switched to using secure external password stores and everything was fine. I wrote an article on possible solutions to the OS authentication issue here.
If I was not working at my company, I don’t believe they would have been running on 19c with PDBs. I’ve been pushing for many years to improve the attitudes to upgrades and patching. The easy path is to do nothing…
Finally I wondered how many people were purchasing the multitenant option. Remember, from 19c onward you can run up to 3 user-defined PDBs without having to buy the multitenant option.
At 35%, I’m actually surprised how many people have purchased the multitenant option. I expected it to be a lot lower. Don’t get me wrong, I think the multitenant architecture is fine. I’ve been advocating for people to switch to it and use lone-pdb since it was introduced on Oracle 12.1. I would like to use more than 3 PDBs per 19c instance, but I can’t justify the cost for a feature that I could argue should be free in all editions.
So there you have it. A quick snapshot of what my followers are saying.
Remember, the multitenant architecture is mandatory from Oracle 21c onward, with Oracle 23c being the next long term support release, so you are going to have to get comfortable with this stuff if you want to remain in support long term.
If you want any help getting to grips with the multitenant architecture I have a load of articles and videos.