Dedicated internet access feels fine right up until it does not. Then calls clip, cloud apps drag, and backups start to fight for the same narrow uplink. At some point, adding another DIA circuit stops being a fix and starts being a delay. The signals below show, in simple traffic and jitter terms, when it is time to stop stacking lit circuits and start looking at leased or dark fiber instead.
Signals at a glance
| Signal | Typical threshold | Why it hurts |
|---|---|---|
| Peak jitter p95 | Over 8 to 10 ms | Video and voice stutter and retransmits grow |
| Packet loss at peak | Over 0.5 percent | Meetings drop and backups slow down |
| Backup window overrun | Jobs spill past window | Recovery points get stale and risk increases |
| Large file transfer time | 1 hour plus for 100 GB | Builds, media, and analytics stall |
| Outage hours per year | More than 4 to 6 hours | Lost labor and missed orders snowball |
| Two circuits on same entrance | Single lateral | One backhoe can take both links down |
| Cloud on ramp distance | Added 20 to 50 km | Latency increases and jitter spikes |
| Camera and POS bursts | Frequent timeouts | Store ops slow and help desk tickets rise |
| Market or event peaks | Recurring slowdowns | Revenue events collide with network limits |
| Growth trend | 25 percent yearly | Today’s DIA headroom disappears next quarter |
1️⃣ Peak jitter p95 is above comfort
If most of your traffic looks fine but the 95th percentile of jitter spikes in busy hours, real time apps become unreliable. A private wave or dark route gives you a fixed path and fewer conversions.
How to check
- Graph p95 and p99 jitter for 30 days
- Note busy hours and match to tickets
- Correlate with BGP path changes if visible
Targets
- Keep p95 under 8 ms for voice and video
- Watch p99 for spikes from route changes
2️⃣ Packet loss crosses 0.5 percent at peak
Small loss rates cause big slowdowns when protocols back off. A dedicated optical path reduces congestion loss and keeps retransmits down.
How to check
- Measure loss with active probes every minute
- Compare peak hour to overnight
- Look for microbursts around backups or opens
Targets
- Aim for 0.1 percent or less
- Cap per-queue bursts with QoS until you upgrade
3️⃣ Backups miss their window
If the largest daily job never finishes on time, the recovery point is stale. A private wave or dark uplink removes the upload choke and keeps RPO tight.
Backup window estimator
Time needed
—
Meets window
—
Tip
Schedule very large jobs off peak
4️⃣ Large file moves block the day
If a 100 GB package takes an hour or more, teams wait on builds and media posts. Private capacity shortens the path and adds headroom.
How to check
- Time a 100 GB transfer in business hours
- Repeat late at night
- Compare DIA to a direct wave or colo on ramp
Targets
- Keep 100 GB under 20 to 30 minutes for busy teams
- Use a private path for build artifacts
5️⃣ Outage hours exceed tolerance
If you lose more than a few hours a year, the hidden cost beats the added fee for a private path with diverse entrances.
Downtime cost quick math
Annual impact
$0
Note
One clean cut can exceed this
Tip
Use diverse entrances on private links
6️⃣ Two circuits share the same lateral
If the primary and backup use the same entrance or trench, a single construction incident can take both links down. A leased dark strand on a different path fixes the single point of failure.
How to check
- Ask for first handhole photos for each circuit
- Trace laterals to different streets if possible
- Require route maps with distance in kilometers
Targets
- Different curb entrances and building rooms
- Different bridges and rail crossings where relevant
7️⃣ Your cloud on ramp is not local
If traffic detours to a distant on ramp, you add distance and jitter. A private wave to the right carrier hotel removes those extra kilometers.
How to check
- Measure latency from site to on ramp MMR
- Compare providers that have on ramps in your metro
- Map both primary and backup paths
Targets
- Keep added distance minimal
- Pick an on ramp you can reach with a short private hop
8️⃣ Camera, POS, or scanner bursts cause timeouts
Short traffic spikes can exceed what DIA uplink will absorb. A private link with higher headroom and clear QoS keeps operations steady.
How to check
- Look for bursty queues at shift changes
- Check WMS or POS timeout logs
- Correlate to uplink saturation graphs
Targets
- Keep uplink below 70 percent at peaks
- Reserve QoS for transactions until you upgrade
9️⃣ Big events overwhelm the link every time
If earnings days, product drops, or market opens always feel choppy, the path is the bottleneck. A dedicated optical route keeps latency and jitter predictable during surges.
How to check
- Overlay performance graphs on event calendars
- Track meeting error rate during launches
- Compare with test traffic at night
Targets
- Flat latency through the event window
- No drop in video bitrate during peaks
🔟 Growth trend says you are outgrowing DIA
If traffic grows faster than your DIA upgrades, headroom shrinks until every busy day feels tight. Leasing dark fiber or a private wave lets you raise capacity in bigger steps without redesigning the edge every year.
How to check
- Graph peak utilization for the last 12 to 24 months
- Note any big step changes like new apps or sites
- Compare peak percent to your comfort limit, often 60 to 70 percent
Targets
- Plan upgrades before peaks live above 70 percent for hours
- Use dark or waves when you are upgrading every one to two years
- Leave room for one surprise project without hitting the ceiling
Capacity growth helper
Predicted peak at end of period
—
Over comfort limit
—
Hint
Plan next step now
If predicted peak crosses your limit inside two years, it is time to model dark or a private wave so you can scale in larger steps.
