Bandwidth vs. Real Throughput
| Traffic Profile | Avg Packet Size | Protocol Overhead | Effective Speed | PPS Load |
|---|
VPC Peering Throughput & Cost Estimator
VPC Peering Throughput & Performance Guide
VPC Peering is a networking connection that allows you to route traffic between two Virtual Private Clouds using private IPv4 or IPv6 addresses. While cloud providers often advertise high bandwidth limits (e.g., 5Gbps, 10Gbps, or even 100Gbps), the actual throughput you achieve is strictly governed by networking physics—specifically protocol overhead, latency (RTT), and packet loss.
Understanding the Mathis Equation (TCP Performance)
The most common reason for slow transfer speeds over a fast VPC peering link is the behavior of the TCP protocol. TCP requires acknowledgement of received packets. As the distance (Latency) increases or if there is even minor packet loss, the protocol slows down to prevent congestion.
The maximum theoretical throughput of a TCP connection can be estimated using the Mathis Equation:
Where:
- MSS: Maximum Segment Size (Packet size minus headers).
- RTT: Round Trip Time (Latency in seconds).
- p: Packet loss probability (0.01% = 0.0001).
Example: A 10Gbps link with 50ms latency and 0.1% packet loss will only achieve roughly 30-40 Mbps, regardless of the 10Gbps cable capacity.
Key Factors Affecting VPC Peering
1. Protocol Overhead & MTU
Every packet sent across the network carries administrative data (headers). The standard MTU (Maximum Transmission Unit) is 1500 bytes. If you use additional encapsulation like IPsec VPNs or VXLAN, the payload space decreases, reducing efficiency.
Tip: Enabling Jumbo Frames (MTU 9000) within a cloud region can significantly increase throughput by reducing the header-to-payload ratio and lowering CPU usage (PPS).
2. Packets Per Second (PPS) Limits
Network appliances, firewalls, and NAT gateways often have limits on how many packets they can process per second. Small packets (like DNS queries or VoIP) consume high PPS but low Bandwidth. A connection might saturate its PPS limit long before it hits its Bandwidth limit.
3. Region & Distance
Inter-region peering (e.g., Europe to Asia) inherently has higher latency due to the speed of light in fiber optics. This higher RTT directly lowers the maximum TCP window size capability unless specialized optimization (WAN Optimization) is used.
How to Use This Calculator
This tool helps network engineers and cloud architects plan capacity realistically.
- Link Capacity: The provisioned bandwidth of your peering connection.
- RTT: Ping time between the two VPC instances.
- Packet Loss: Estimated loss rate. For high-quality cloud networks, use 0.0001% (0.000001).
- Overhead: Select if you are running raw traffic or tunneling via VPN (IPSec/OpenVPN), which adds significant overhead.