Performance
Monitoring → Performance answers one question: what slows the JD Edwards platform down now, and what to look at first? The page reads Server Manager, the database and the cross reference — live, read-only, in about seven seconds — and turns the figures into a short list of advice, the heaviest first. Each line says what was measured, what it means and what to do.
At a glance
The figures
Nine figures head the page: business function calls, seconds spent in them, ms per call, timeouts, requests waiting, % Java heap, highest, % cache hits, lowest, % database cache hits and sessions blocked. A figure turns amber or red past its threshold.
Read again, at the top right, reads everything once more.
The advice
The lines are grouped by weight — To act on first, To review, To know — and each one names its area.
| Area | What is flagged |
|---|---|
| Business functions | A function that timed out, failed on the system side, takes most of the time measured, is slow at each call, or had one long wait. The line lists the applications and reports whose event rules call it, from the cross reference. |
| Enterprise server | A server step such as init-remote-env (opening the user's environment at sign-in), a kernel with requests waiting, a process that keeps transactions open, zombie processes. |
| Web and AIS | A Java heap close to its maximum. |
| Caches | A JDBj cache that answers too rarely. |
| Database | Blocked sessions, unusable indexes, a low cache hit ratio, statements parsed from scratch, a statement slow at each execution (its schema, its tables, who sends it, its text), redo logs that switch too often, missing or stale statistics, CPU held back by the Resource Manager, and the three largest waits. |
An area read with nothing to flag is named under the list (Nothing to flag: …): silence is not "not read". A source that did not answer is said in a warning; the rest of the page is still read.
Details
Most lines carry a Details button, at their right: it reads, live, the rows behind the figure and shows them under the line, in a grid with search, filters, columns and export. Hide closes them.
| Line | What Details lists |
|---|---|
| Tables never analysed, or with stale statistics | Each table: owner, table, its description, rows, last analysis, state of its statistics. |
| Unusable indexes | Each index, its table and its type. |
| Blocked sessions | The sessions waiting, and who blocks them: session, user, program, machine, wait. |
| A slow statement | Its full text and the plan the database runs it with. |
| Low database cache hit ratio | The statements that read the disk most. |
| Statements parsed from scratch | The statements sent with their values written in the text. |
| Redo logs switching too often | The switches by hour. |
| A business function | Who calls it: the applications and reports whose event rules name it, up to 200 places. |
| Zombie processes, kernel with requests waiting, open transactions | The processes of the enterprise server named — the zombies themselves when Server Manager says which they are, otherwise every process of the server. |
| Java heap, server step, cache | The nodes, the business function calls or the caches of the server named. |
A database detail is read from an Oracle database only; on another database the line says so. Like the rest of the page, Details only reads: no AWR, no ASH.
How to read the page
The box at the bottom of the page states its limits:
- every figure is the present moment or a counter since the servers started: there is no history yet, so the page says what is slow, not since when;
- the thresholds: a function is flagged from a share of the time measured or a time per call, a Java heap from 80 %, a cache under 80 % of hits, a statement from a time per execution, the database cache under 90 %;
- the timeouts are the business functions' own counters — the enterprise server's log files are not reachable through Server Manager;
- the database is read from its own counters, without the Diagnostics Pack: no AWR, no ASH.
Access
The page shows to the roles that hold its menu entry and the permissions plugin:nomajde.performance:*; Details needs nothing more. It reads Server Manager through the jde_server_manager pool (see Server Manager) and the JD Edwards database.
The assistant is told what the page shows — the heaviest advice — and offers its own questions: What should I look at first, and why? See AI assistant.
Tips & best practices
- Start with To act on first: one business function that takes a large share of the time is where users feel an improvement most.
- Open the callers of a slow function in the cross reference to know which screens and reports are affected.
- After a restart, the counters start from zero: read the page again once the platform has worked for a while.
- Details first, for the rows behind a line; then Platform analysis (servers, processes, functions, INI) or Database audit for the whole picture.