HTTP latency

Definition

Difference in latency for HTTP requests made to hosts offering both IPv4 and IPv6, for hosts among top-500 from EU-27 + USA, India, Canada, China and Japan.

Measurement method

Difference in latency for HTTP requests made to hosts offering both IPv4 and IPv6 is made as follow for each tested server:

Data collection

  • A t1 timestamp is taken when the http request is launched from a reference server to the server under test
  • A t2 timestamp is taken the http request has been fulfilled.
  • Δt=t2-t1 provides the latency
  • This measure is done over both IPv4 $(Δt_4)$ and IPv6 $(Δt_6)$ and is repeated every hour so to obtain N=10 measures.The averages ${Δt_4}↖{-}$ and ${Δt_6}↖{-}$ are calculated.

Daily analysis

  • Once a day, information is collected following the procedure above for all servers identified as having a AAAA record
  • Servers exhibiting either a null $(Δt_4)$ or a null $(Δt_4)$ are not taken into account
  • The daily ratio $Δlat (%)=100{{Δt_6}↖{-}-{Δt_4}↖{-}}/{{Δt_4}↖{-}}$ is calculated
  • Only values obtained in the limits of ±300% are kept. Others are manually investigated.

Hypothesis and uncertainties

  • Up to 5 redirections are taken into account in the latency estimate. When more than 5 redirections occur the test is stopped

Figures

latency

Analysis

The majority of measured servers exhibits comparable http requests latency time over IPv6 or IPv4. For the one where a difference is seen, half exhibit faster requests over IPv6 while the other half exhibits faster requests over IPv4. From these figures, IPv6 offers the same user experience as IPv4 for http requests.