L2-to-cloud latency: real-world numbers

TL;DR

L2 tunnel latency to the Sefthy cloud: 8-15 ms on Italian 2.5 Gbps FTTH, sustained throughput 800-1500 Mbps symmetric, jitter < 2 ms. Invisible to users for 95% of line-of-business apps.

Real numbers

Measurements run on three typical setups:

Symmetric 2.5 Gbps FTTH

• Connector → Sefthy datacentre (Bari) latency: 8-12 ms;
• Connector → Sefthy datacentre (Milan) latency: 12-15 ms;
• sustained throughput: 1200-1500 Mbps;
• jitter: < 1.5 ms.

FTTC 100/30 Mbps

• latency: 18-25 ms (contention-limited);
• symmetric throughput: 30 Mbps;
• jitter: 3-8 ms.

Sufficient for backup-only. Borderline for interactive workloads (RDP, VoIP).

Outdoor 5G

• latency: 15-30 ms (coverage-dependent);
• throughput: 150-300 Mbps;
• jitter: 5-15 ms.

Suitable for backup, marginal for real-time restore.

What works well at 10-15 ms

• web-based ERPs;
• thin-client business apps;
• SMB file servers;
• IMAP/Exchange mail;
• browser-based access.

What needs verification at 10-15 ms

• VoIP: usually OK, but test with high-quality G.711;
• remote desktop: OK with standard RDP, problematic with GPU acceleration;
• HD video streaming: needs bandwidth + prioritisation.

What does NOT work at 10-15 ms

• high-frequency trading;
• real-time CAD/CAE simulation;
• professional gaming.

How to optimise

Three levers:

1. Closer Sefthy datacentre to the customer: cuts 3-7 ms.
2. QoS on customer firewall: prioritises tunnel traffic.
3. Symmetric bandwidth: latency rises with saturated upload.

Comparison with "European" cloud

Toward generic European datacentres (e.g. Dublin, Frankfurt):

• latency from Italy: 25-50 ms;
• jitter: 3-8 ms;
• throughput: ISP-dependent.

Workable for distant DR, but the gap vs Italian sovereign is 20-30 ms average. Significant for interactive apps.

FAQ

Extra tunnel latency vs a direct connection?

Typically +1-3 ms for encryption/encapsulation overhead. Negligible.

Can I measure my latency before buying?

Yes, Sefthy offers a free 30-day test with latency report.

Does dedicated bandwidth make a difference?

Yes, especially for upload-heavy workloads.

For the L2 pillar, L2 tunnel for DR. For security, L2 tunnel security.