当链路同时承载弹性流量与非弹性流量时，哪类会占据主导？

Great question—your initial intuition is spot-on! Let’s break this down to confirm why non-elastic traffic’s rate is the key factor here.

First, let’s clarify the two traffic types to set the stage:

• Non-elastic traffic: Think real-time services like VoIP, video calls, or live streaming. These have fixed bandwidth requirements and can’t adjust their rate based on network congestion. If they don’t get the bandwidth they need, you’ll immediately see quality issues (stuttery audio, pixelated video) because they can’t retry or slow down gracefully.
• Elastic traffic: This includes things like file downloads, HTTP web browsing, or cloud sync. Protocols like TCP build in congestion control—they’ll automatically speed up when the link is free, and throttle back when congestion hits. They’re designed to adapt to available bandwidth.

Why non-elastic traffic dictates dominance #

Your core point is correct: the rate of non-elastic traffic directly controls how much of the link it occupies, and elastic traffic will always adapt to the leftovers. Here’s why:

• Non-elastic traffic claims its required bandwidth first (either explicitly via QoS policies, or implicitly by refusing to back down). If a video call needs 40Mbps on a 100Mbps link, that 40Mbps is effectively "reserved"—elastic traffic can only use the remaining 60Mbps.
• If non-elastic traffic’s rate spikes (say, multiple video calls start using 80Mbps total), elastic traffic will immediately scale back to use the remaining 20Mbps. If non-elastic traffic drops to 10Mbps, elastic traffic will fill the gap to use the full 90Mbps available.
• In cases where non-elastic traffic exceeds the total link bandwidth (e.g., 120Mbps of video traffic on a 100Mbps link), two things happen: either non-elastic traffic starts dropping packets (ruining quality) or if QoS is enabled, it will actively preempt bandwidth from elastic traffic (since real-time services are prioritized). Elastic traffic will just slow down or retry, so it’s the one that absorbs the impact.

A quick edge case to note #

If there’s no QoS configuration, non-elastic traffic doesn’t get explicit priority—but it still acts like the "dominant" traffic because it won’t adjust. Elastic traffic will back off first when congestion occurs, since its protocols are built to handle that.

In short: You’re exactly right. Non-elastic traffic’s rate sets the baseline for bandwidth usage, and elastic traffic always adapts around it.

内容的提问来源于stack exchange，提问作者Aman Purwar