TCP Slow Start

TCP slow start is the opening phase of a TCP connection's congestion control. The sender begins with a small congestion window, commonly 10 segments in modern systems, and grows it by about one segment for every acknowledged segment, which roughly doubles it each round trip.

Why it matters

A fresh connection cannot use full bandwidth right away, even on a fast link. Small web resources may finish before the window has grown, so their speed is limited by latency instead of bandwidth. This is why the first 14 KB or so of an HTML response matters: ten segments of about 1,460 bytes fit in the first flight, so a small critical response can arrive in one round trip.

Example

Initial window: 10 segments (about 14.6 KB)
RTT 1: send 10 segments  (~14.6 KB)
RTT 2: send 20 segments  (~29 KB)
RTT 3: send 40 segments  (~58 KB)

A 100 KB file therefore needs several round trips, not one. After an idle period, some stacks reset the window and slow start repeats.

Common mistakes

• Expecting a faster plan to fix small-file downloads; round trips dominate.
• Opening many short-lived connections instead of reusing one.
• Forgetting that TLS handshake bytes use up part of the first window.

Related

• TCP
• Keep-alive
• Latency
• HTTP/3 and QUIC

यह पेज सामान्य जानकारी है, जिसे हमारी संपादकीय टीम ने लिखा है और दस्तावेज़ीकरण बदलने पर अपडेट किया जाता है। सर्वर और प्रदाता अलग-अलग होते हैं, इसलिए बदलाव पहले सुरक्षित तरीके से जाँचें। देखें: संपादकीय नीति और अस्वीकरण.