Performance Troubleshooting

Performance issues on IZO™+ Multi Cloud Connect are diagnosed from the TC^x performance views before anything else: utilisation and discards tell you whether the link is full, and the round-trip delay, jitter and packet-loss reports tell you whether the path is degraded. Read the data first, then decide whether the fix is a bandwidth change, a design change, or an incident.

Gather the evidence first

Metric	Where to read it	What it tells you
Utilisation (Tx/Rx), Utilisation Max	At a glance, Dashboard Utilisation tab	Whether the connection is running out of headroom.
Discards In/Out	At a glance	Packets dropped, an early sign of congestion or shaping.
Round-trip delay	Performance report (CE-to-CE)	Baseline latency and any change over time.
Jitter	Performance report	Variation in delay, important for real-time traffic.
Packet loss	Performance report	Loss on the path over the period.
Underlay uptime	Consolidated underlay report	Whether an Internet or MPLS underlay event coincides.

Sample data is collected at a 5 minute size, so compare a suspected problem window against a known-good window.

Latency higher than expected

Compare current round-trip delay against the baseline for the same connection. A step change points to a path or underlay event; a gradual rise often tracks rising utilisation.
Latency is bounded by distance and the chosen on-ramp and region. If the cloud region or resiliency model routes traffic a long way round, that is a design consideration rather than a fault.

Throughput lower than expected

Check Utilisation Max. If it is close to the provisioned bandwidth during the problem window, the link is the limit; plan a bandwidth change.
A single TCP flow will not fill a high-bandwidth, higher-latency path on its own; test with parallel flows before concluding the link is slow.
Rule out MTU and fragmentation: jumbo frames are supported below 9000 bytes where the path allows, but an Azure ExpressRoute WAN side is effectively 1500, so an oversized packet that cannot fragment will hurt throughput.

Packet loss

Correlate Discards and the packet-loss report with utilisation: loss that appears only at high utilisation is congestion, addressed by a bandwidth change or traffic management.
Loss that is independent of load, or that coincides with the consolidated underlay report, points to a path or underlay issue; raise an incident with the evidence.

What’s on the cloud side

The cloud-side gateway and any cloud-native throughput limits apply at the far end. This page covers diagnosis on the Tata Communications side; for cloud-side gateway sizing and limits, see the relevant per-cloud section.