India server latency report: TCP round trips from 28 Indian ISP connections to seven cloud regions (25 September 2026)
Authors/Creators
Description
Median TCP round-trip times from 28 connections on consumer and business ISP networks in 20 Indian towns and cities to public AWS endpoints in seven cloud regions, measured with the Globalping open probe network on 25 September 2026, 06:03-06:04 IST, in one run.
Results (median across connections; middle half in brackets): Mumbai 35 ms (24-45), Hyderabad 37 ms (24-47), Singapore 73 ms (54-82), London 152 ms (142-162), Frankfurt 155 ms (146-166), Virginia 247 ms (234-273), California 255 ms (238-264). For 28 of 28 connections the nearest region was in India (Mumbai or Hyderabad). Singapore added a median 40 ms per round trip over the nearer Indian region.
Probes: every Globalping probe in India that Globalping classifies as a consumer or business ISP network was asked; 38 answered. 7 that sit on cloud or hosting networks (Oracle Cloud, HostDzire, Good Domain Registry, HostStack) and 3 that did not answer for every region were left out, leaving 28 connections on 14 networks in 20 places.
Targets: one public endpoint per region, AWS's regional DynamoDB hosts: dynamodb.ap-south-1, ap-south-2, ap-southeast-1, eu-central-1, eu-west-2, us-east-1 and us-west-1 (.amazonaws.com). They stand in for "a server in that city"; this is a test of distance, not of AWS or of any hosting provider, and none of them is a FasterHost server.
Measurement: each probe opened 16 TCP connections to port 443 of each endpoint and timed each handshake (one network round trip). A probe's figure is the median of its 16; a region's figure is the median across probes. Every region was measured from the same probes.
Limits: one snapshot, taken early in the Indian morning, when networks are quiet; evening peaks run higher. Probes are fixed-line, so mobile data and Wi-Fi would add to every figure. Coverage is uneven: several states have no probe, and most places have one or two.
Files: india-latency.csv and india-latency.json hold every probe's result, including the probes that were left out.
Published by FasterHost, a web hosting company in India, which has an interest in the answer; for that reason the test targets neutral cloud endpoints rather than its own servers and publishes every probe's result. Report with the city-by-city table and method: https://www.fasterhost.in/india-latency-report.php
Technical info
india-latency.csv has 266 rows, one per probe and region (38 probes x 7 regions), with columns probe_city, probe_network, probe_asn, region, endpoint, median_rtt_ms, min_rtt_ms (milliseconds; median and fastest of 16 TCP handshakes to port 443), excluded_reason (empty for the 28 probes used in the results) and measured_from / measured_to (UTC). Rows come in blocks of seven per probe, in the order Mumbai, Hyderabad, Singapore, Frankfurt, London, Virginia, California, matching the probes array in india-latency.json. 14 rows (28 cells: median_rtt_ms and min_rtt_ms) are empty, all from the 3 probes that did not answer for every region. india-latency.json holds the same run with the tool, method, probe filter, Globalping measurement IDs, region endpoints and a per-region summary (median, 25th and 75th percentile, probe count).
Files
india-latency.csv
Additional details
Related works
- Is supplement to
- Report: https://www.fasterhost.in/india-latency-report.php (URL)
Dates
- Collected
-
2026-09-2506:03-06:04 IST, one run