Sizing and Capacity¶
What drives CPU, memory and disk use on the MetaMiner server host, and how to set the server memory.
What drives load¶
| Factor | Affects | Why |
|---|---|---|
| Number of objects in each CMS | Server memory, client memory, MMDB size | The server builds and the client holds a snapshot of each environment. |
| Number of Web Intelligence documents, universes and Crystal reports | CPU, MMDB size, first-pull time | Deep processing opens and parses each document. |
| Number of environments | Memory, CPU | Each environment has its own pull workers and SDK session. |
| Audit history length | MMDB size, first audit pull time | The first audit pull reads up to 550 days back by default; the MMA recommends 2 years. |
| Document and universe version history | Disk | Stored on the versions path when version history is turned on. |
| Backups, BIAR files and promotions | Disk | Stored on the archive path. |
| Concurrent client users | Server CPU | Each user's actions run on the server. |
Server memory¶
The MetaMiner Server heap is set in the MetaMiner Administrator, Preferences > Settings... > System. The value is written to the Windows service definition and applies at the next server start. The embedded database uses memory outside that heap, so leave room for it and for the operating system.
The MetaMiner Administrator warns you at Start Server if free memory is low compared with the configured heap.
See Memory and JVM.
Disk¶
| Location | Grows with | Default |
|---|---|---|
server\data\mmdb\ |
Metadata, audit history | Inside the installation folder |
| Versions path | Web Intelligence and universe version history | Set in Backup settings |
| Archive path | Backups, promotions, alert archives | Set in Backup settings |
server\logs\ |
Log retention | Inside the installation folder |
| Java temporary folder | Document processing, exports | server\javatemp, set on the System settings page |
Use a separate volume or share for backups
If you use backup, versioning or promotion, point the archive and versions paths to a volume with room to grow, or to a UNC share the service account can write to.
Sizing table¶
TODO: replace this table with current measured guidance
The table below is the previously published guidance. It predates Java 17, PostgreSQL 16 and the current Web Intelligence processing. Re-measure with representative landscapes (object counts, document counts, audit volume) and replace it. Consider sizing by number of CMS objects and documents rather than concurrent BusinessObjects users.
Previous published guidance (by peak concurrent BusinessObjects users)
"Concurrent users" here means the peak number of distinct users logged on to BusinessObjects at the same time (CMC > Sessions at the busiest hour).
| Concurrent users | RAM (GB) | CPU cores | Disk (GB) |
|---|---|---|---|
| 0-50 | 4 | 2 | 60 |
| 51-100 | 4 | 4 | 80 |
| 101-150 | 8 | 4 | 120 |
| 151-300 | 10 | 4 | 200 |
| 300+ | 10 | 8 | 400 |