Use casesIt’s up, but it’s bad
Voice and video degrade at the same time every day.
A predictable daily dip is almost always one scheduled job colliding with business hours. It is also one of the few problems you can fix for free, by moving the job or shaping it.
Recognize it
You’re here if
- The same site complains at the same hour on most days.
- Link utilization sits in the nineties at the dip.
- Jitter and loss spike together, then recover.
Do this
5 steps, in order
Prove it is a pattern
Pick the site and the Utilization metric. Set the window to 7 days.
Open Explorer- Look for
- A repeating daily shape with the same peak hour.
- What it means
- A repeating shape means a schedule. A one-off means an incident.
Find who is talking
Drill from the site to its link and check which applications carry the load at the peak.
Open Dashboards- Look for
- One bulk application, such as a database replication or file transfer, taking most of the capacity.
- What it means
- That is the job to reschedule or shape.
Check the headroom trend
Open Predictive and look for this link.
Open Predictive- Look for
- A forecast crossing a threshold, or none.
- What it means
- If peak demand is structural and growing, QoS buys time and a bandwidth order is the real fix.
Fix it for free first
Review the recommended QoS policy that puts voice and video ahead of bulk, and the replication reschedule.
Open Remediation- Look for
- Reversible actions with a rollback plan.
- What it means
- Move the bulk job to off-peak. Enforce priority for real-time classes.
Verify at the next peak
Come back after the next business-hours peak and compare jitter and MOS with the same hour last week.
Open Explorer- Look for
- Jitter back near its baseline and MOS recovered.
- What it means
- Not fixed until the peak has passed once.
Then decide
What you found decides what you do
IfMoving the job and shaping fixes the dip
ThenDone. Add the job’s schedule to the change calendar.
IfUtilization is still high with the job moved
ThenDemand has outgrown the circuit. Plan a bandwidth increase. Get warned before a breach →
Watch out
Where this goes wrong
- Do not add bandwidth first. If a bulk job is the cause, you will fill the new capacity too.
The bulk job is off the peak or shaped, real-time classes have priority, and you have checked the next peak.
Keep going
Often next
Get warned before a breach
Read a forecast, act on its ETA, and check the track record before you trust it.
~10 minNetwork lead, NOC engineerIt’s up, but it’s badCalls are choppy, links are green
Find the silent brownout that never trips an outage alarm, and bring the carrier evidence.
~10 minNOC engineer, Network lead, Service ownerRun the NOCApprove or hold a fix
Read the recommendation, the rollback plan and the blast radius, then decide.
~10 minNetwork lead, NOC engineer, Security analyst