Proxy Speed Explained: Latency, Throughput and What Slows a Proxy Down

· 6 min read

"Fast" is the most common promise in proxy marketing and the least explained. A proxy can have plenty of bandwidth and still feel slow, or answer instantly and stall on a large download. The two are different measures, and knowing which one limits you tells you what to fix.

Latency and throughput

  • Latency is how long a message takes to get somewhere and back, usually given as the round trip time in milliseconds. It decides how quickly a page starts loading and how many requests per second a single connection can make.
  • Throughput is how much data moves per second once a transfer is underway. It decides how long large responses, images and downloads take.

Most automated work, from scraping to account management, makes many small requests. For that kind of traffic, latency matters far more than bandwidth.

Where the time goes

A new HTTPS request needs several round trips before the first byte of the response comes back: one for the TCP handshake, at least one for TLS, and one for the request itself. Through a proxy, there are two legs to every trip, from you to the proxy and from the proxy to the site, plus a short exchange with the proxy at the start.

Distance sets a floor under all of it. Light in fibre covers about 200 kilometres per millisecond, so a round trip between London and New York can never take less than about 55 milliseconds, and in practice takes closer to 70, no matter how good the network is. Multiply that by the round trips of a new connection, and the location of the proxy starts to matter more than anything else.

Why some proxies are slow

  • Peer-to-peer routing. Rotating residential networks relay traffic through other people's devices. Home connections have slow uploads, phones go to sleep, and each request may take a different, longer path. Speed depends on whoever is online at the moment.
  • Crowded servers. A proxy shared by too many users, or on a cheap uplink, slows down at busy hours.
  • Distance. A proxy far from both you and the sites you visit adds two long legs to every round trip.
  • New connections for every request. Opening a fresh connection each time repeats the handshakes, and through a proxy that cost is paid twice.

Getting the most out of a proxy

  • Reuse connections. Keep a session or client open across requests instead of creating one per request. Examples are in using a proxy with curl, Python and Node.js.
  • Run your code near the proxy. A scraper on a server in the same region as the proxy cuts the first leg of every trip to almost nothing.
  • Pick a location close to your targets. For US sites, a proxy on the East Coast sits near much of the infrastructure they run on.
  • Work in parallel. When latency is the limit, more concurrent connections do more work in the same time, within what the site will tolerate.
  • Let the proxy resolve DNS. It saves a lookup on your side and avoids leaking the sites you visit, as explained in HTTP vs SOCKS5 proxies.

Measuring it yourself

curl can break a request down into its stages:

curl -o /dev/null -s -x http://USER:PASS@HOST:PORT \
  -w "connect %{time_connect}s  tls %{time_appconnect}s  first byte %{time_starttransfer}s  total %{time_total}s\n" \
  https://example.com

Run it a few times through the proxy and compare the connect time, the time to the first byte and the total, then repeat it at different hours to see whether the proxy holds up when it is busy. Your usage metrics show the other half: how much traffic you move and when.

Our setup

Leastslow IPs run on servers with gigabit connectivity, not on home devices, with our gateway in Ashburn, Virginia, one of the major internet hubs in the world and close to a large share of US hosting. There is no peer in the middle and no one else on your IP, so the speed you measure on a quiet morning is the speed you get at a busy hour. Pricing is on the homepage.