Performance shims and workarounds for Produce Pro and DBC legacy systems.

Legacy ERP systems often contain inefficient patterns that can't be fixed without vendor involvement. This repository contains drop-in shims and wrappers that intercept problematic operations and optimize them without modifying application code.

Problem: Produce Pro's Linux User Management screen (WSUINQ) calls last <user> -n 1 for each user individually. With 150+ users and a 17MB wtmp file:

• 150 users × 0.46 seconds = 70+ seconds just for last commands
• Total screen load time: ~7 minutes

Solution: Cache the full last output on first call, grep from cache for subsequent calls.

Result: Screen load time reduced from 7 minutes to 10 seconds (42x improvement)

Installation:

cp bin/last /usr/local/bin/last
chmod +x /usr/local/bin/last

Uninstallation:

rm /usr/local/bin/last

How it works:

1. First call runs /usr/bin/last and caches output to /tmp/.last_cache
2. Cache refreshes every 60 seconds
3. Subsequent calls grep from cache instead of scanning wtmp
4. Appends dummy footer lines so | head -n -2 works correctly

• readcache - LD_PRELOAD library for userspace ISAM read caching

rm /tmp/.last_cache

which last
# Should show: /usr/local/bin/last

# Test directly
/usr/local/bin/last testuser -n 1
/usr/bin/last testuser -n 1

/usr/bin/last <args>

When adding new shims:

1. Document the problem (what's slow, why)
2. Document the solution (how the shim works)
3. Include installation/uninstallation instructions
4. Include performance measurements (before/after)
5. Ensure easy rollback

MIT License - See LICENSE file

EW White Consulting (EWCS)
https://www.zfsexpress.com