Best practices / additional platform performance monitoring

[KaueC]

Hi everyone,

We built an internal Service Portal dashboard we call WBS Hub, a Control Room that gives our team a single pane of glass for the health and performance of our ServiceNow instance, plus operational backlog and integrations. It refreshes on demand (with optional auto-refresh) and every KPI is clickable, opening the underlying list with the exact same encoded query so the number always matches the source.

Today it is organized into a few areas:

1. Platform / System health (last 15 min window)

• Average response time (based on the native Semaphore Average Time indicator)
• Agent Workspace average response time
• RPA integration average response time
• Average SQL time and average Business Rule time per transaction
• Slow transactions (> 5s)
• Nodes restarting (Node Start diagnostic events) and pending/processing events (sysevent)

2. Queues & Events

• Pending events older than 15 min, Flow queue (QUEUED contexts)
• Email outbox and send-failed emails
• ECC output queue and MID Server communication errors

3. Backlog & inconsistencies (broken down by service offering / stream)

• RITMs with unhandled flow errors, closed RITMs left in flow WAITING, RITMs without service offering
• Flows QUEUED > 1 day, flows Presumed Interrupted
• Tasks without assignment group, active tasks tied to inactive RITMs, etc.

4. Service status & overall platform health

• ServiceNow platform, MID Servers, outbound email — with thresholds that flag the platform as "Healthy" or "Degraded"

We also surface some cost/operational signals on the side (ServiceNow license usage, Twilio/WhatsApp balance and consumption).

One technical note we learned along the way: on rotating tables like syslog_transaction, GlideAggregate does not return reliable AVG/MAX/SUM values, so we compute those aggregates manually via GlideRecord. COUNT remains reliable through GlideAggregate.

My question: what other platform performance monitoring would you recommend adding? We are especially interested in:

• Which metrics/tables you rely on to catch performance degradation early (beyond syslog_transaction)
• Whether you use Health Log / Instance Observer / Performance Analytics for this instead of custom queries, and how that compares
• Good thresholds/alerting patterns for response time, SQL time, semaphore waits, scheduler/event backlog
• Any signals we're missing (cache/pool saturation, long-running scheduled jobs, slow ACLs, session counts, etc.)

Thanks in advance for any tips!