← 📬 Application Layer: DNS, HTTP, Email
Subtopic 2 of 3

8.2HTTP timing: persistent and non-persistent

The idea

HTTP page-load time is counted in RTTs: 1 RTT for the TCP handshake and 1 RTT for each request-response. Persistent connections and pipelining save round trips.

Think of it like this: Phoning a shop: non-persistent is hanging up and redialling after every question; persistent is staying on the line; pipelining is asking all your questions in one breath.

Why it exists

The "minimum time to render the page in RTTs" question comes up regularly.

How it works

  1. Add the DNS lookup RTTs first.
  2. Base page: 1 RTT (handshake) + 1 RTT (request).
  3. Objects, non-persistent: 2 RTT each, divided into ⌈n/p⌉ rounds if p connections run in parallel.
  4. Objects, persistent: 1 RTT each; with pipelining all together cost 1 RTT.

Worked example: 3 DNS RTTs, 10 small objects, non-persistent with 5 parallel connections versus persistent with pipelining.

Non-persistent: 3 + 2 + 2 × 2 = 9 RTT. Persistent with pipelining: 3 + 2 + 1 = 6 RTT (GATE 2023).

Now play with it

  • →Change the number of parallel connections in the calculator.
  • →Compare persistent with and without pipelining for 10 objects.
Non-persistent, sequential25 RTT
Non-persistent, p parallel9 RTT
Persistent, no pipelining15 RTT
Persistent with pipelining6 RTT

GATE 2023: 3 DNS RTTs, 1 RTT for the TCP handshake and 1 for the base page, then 10 objects. Parallel non-persistent: 3 + 2 + 2 × 2 = 9. Persistent with pipelining: 3 + 2 + 1 = 6.

Check yourself

A web page has a base file and 5 images. With non-persistent HTTP and no parallelism, how many TCP connections are needed?

Remember: DNS + 2 RTT base. Non-persistent 2 RTT per object. Persistent 1 RTT, pipelined 1 RTT for all.
Exam notes (the dense version)

An RTT is the round trip time. A TCP handshake costs 1 RTT, and a request-response costs 1 RTT (ignoring transmission time). Non-persistent HTTP opens a new connection per object: 2 RTT per object (1 handshake + 1 request). Persistent HTTP keeps the connection: after one handshake each object costs 1 RTT, and with pipelining all requests go out together, so n objects cost about 1 RTT. With p parallel connections, objects are fetched in ⌈n/p⌉ rounds. Total time = DNS lookup + 2 RTT for the base page + time for the embedded objects. HTTP is stateless; cookies add state.

Previous year questions: Application Layer: DNS, HTTP, Email (9)

PYQs are grouped by topic, so this is the full set for Application Layer: DNS, HTTP, Email; try the ones that match this subtopic. Open all 9 on the topic page.