OAS: Need help on solution build

Question

Hi Team,

Could you please suggest how to build solution, Architecture for QA & PROD environments (capacity planning bases on users. Hardware configuration, build security model ect.,)  for OAS (2024).

Please help me with any document/links to go through

Thank you,

NR

Gianni Ceresa · Accepted Answer

Hi,

I would definitely start by having a look at

https://docs.oracle.com/en/middleware/bi/analytics-server/install-config-oas/index.html

for an overview of the standard installation topology. If you have the need to go for the "Enterprise Deployment", you can read about it at

https://docs.oracle.com/en/middleware/bi/analytics-server/enterprise-deploy-oas/index.html

(and here you easily see how it's already a quite complex deployment, you maybe don't need it at all…).

Keep an eye on

https://docs.oracle.com/en/middleware/bi/analytics-server/administer-oas/appendix-certification-information.html

to make sure you are working with a supported OS.

I don't believe there is still a capacity planning document or reference number anywhere, they stopped releasing those many years ago because they make little sense. You could have a look at FMW requirements

https://docs.oracle.com/en/middleware/fusion-middleware/12.2.1.4/sysrs/index.html

but that's anyway just a generic random indication.

The hardware required isn't really based on users anymore but on what your users will be doing and how many of them will be doing it at the same time: 10'000 users in total with just 50 concurrent users could need less resources than 100 users in total with 10 concurrent users all training ML models at the same time, or all dumping the whole database at the same time.

Also, consider OAS 2025, it gives you an extra year of support compared to 2024 as it's the newer version.

Steve F-2 · Accepted Answer

Hi @User_LTL5P (Nayab),

Welcome to the Oracle Analytics community!

To tag on to the information Gianni provided.   There is an old capacity planning guide in the knowledge base (if I remember correctly based on OBI 11g/12c), but it may not help too much.

I observe most have a Linux baseline of 4-8CPU and 32GB memory, on fast disk/storage.

As for concurrency, also take into account the use of classic delivers (agent) load, or Publisher usage, and self-service usage of Workbooks.

Ideally, you will run iterative "real world" load testing, meaning, appropriate "ramp" up time, and "think" time, as not all users login at the same time, and click/execute queries at the same time, and agents should be scheduled appropriately, "Dashboards" designed efficiently.

As you iteratively load test, you can tune the system.

Best Practices Guide for Tuning Oracle® Analytics Server (Doc ID 2866848.2)

It includes embedded links to best practices (i.e. - ensuring your model and other aspects are designed as efficiently as possible)

This is a related link

https://blogs.oracle.com/analytics/post/oracle-analytics-best-practices-series-optimal-performance-and-usage

Another option, you may want to consider is Oracle Analytics Cloud.  The benefits there, are more frequent updates, and the sizing/hardware, etc. are all handled in the service. Other benefits, I am just listing the ones related to your question.

Emphasizing to start with OAS 2025, as mentioned.

Thanks for bringing  your question to the community!

Other comments, welcomed!