How the Proxy Speed Test Measures Connection Time
Requests start from the current MiyaIP application server, pass through the proxy, and reach fixed public probes. The test validates the proxy host, selects or confirms the protocol, then records six lightweight connection and TTFB samples before the separate download phase.
Connection time measures the TCP setup from the test server to the proxy; it is not ICMP ping. Time to first byte includes proxy and target response delay. The default download opens four streams of 2 MiB and calculates aggregate throughput from the bytes actually transferred and the shared elapsed time.
How Speed Ratings Are Assigned
A fast rating requires all seven groups to succeed, average TTFB no greater than 300 ms, and download throughput of at least 500,000 bytes per second. An average rating requires at least five successful groups, average TTFB no greater than 800 ms, and throughput of at least 100,000 bytes per second.
A valid measurement that misses those thresholds is slow. Fewer than three usable TTFB samples or no usable download sample remains unrated rather than forcing a conclusion from incomplete data.
Compare repeatable measurements under the same conditions
The timer reports only real elapsed time, and every probe remains visible so failures are not hidden by an average. Test-node region, cross-network routing, proxy load, target congestion, and transient packet loss can all change the result.
Repeat several runs close together and compare minimum, average, maximum, failure rate, and download throughput. This server-to-proxy-to-probe measurement is not a browser Speedtest or a service-level guarantee.
Proxy Speed Metrics Defined
No single number describes a proxy path. Read setup delay, response delay, transfer capacity, and reliability together, and remember that the measurement starts at the current MiyaIP test node rather than your browser.
Metrics reported by the proxy speed test| Metric | Definition | What it helps diagnose |
|---|
| Connect Time | Elapsed TCP setup time from the MiyaIP test server to the proxy listener. | Routing distance, congestion, filtering, and listener responsiveness; it is not ICMP ping. |
| TTFB | Time from starting a proxied request until the first response byte arrives. | Combined proxy processing, target response, tunnel, and route delay. |
| Throughput | Bytes successfully downloaded across four streams divided by their shared elapsed time. | Transfer capacity for this server-to-proxy-to-fixed-probe download path. |
| Failure Rate | The share of probe groups that did not produce a usable sample. | Intermittent reachability that an average of successful requests could hide. |
Six Lightweight Probes and an 8 MiB Download
After validation and protocol selection, the test performs six lightweight requests. Every sample remains visible so a quick success cannot hide repeated timeouts. The separate download phase opens four parallel streams, each requesting 2 MiB, and calculates aggregate bytes per second from actual completed bytes and a common elapsed interval.
When all four streams complete, the download payload is about 8 MiB, plus small overhead from lightweight probes and any automatic protocol detection. Partial streams contribute only the bytes actually recorded. This is a server-side download test: it does not measure upload speed, your local Wi-Fi, browser rendering, or the distance from your device to the proxy.
Current measurement sequence1 proxy × 6 lightweight connection/TTFB probes
then 4 parallel download streams × 2 MiB
aggregate throughput = transferred bytes / shared elapsed time
rating uses successful groups + average TTFB + throughput
Fast, Average, Slow, and Unrated Thresholds
The rating is a compact description of the current probe set, not a universal quality score. Exact measurements remain more useful when two endpoints sit near a threshold.
Current proxy speed rating rules| Rating | Successful groups | Average TTFB | Download throughput |
|---|
| Fast | All 7 groups. | 300 ms or less. | At least 500,000 bytes per second. |
| Average | At least 5 of 7 groups. | 800 ms or less. | At least 100,000 bytes per second. |
| Slow | A valid measurement that misses the Fast and Average combinations. | Shown from usable samples. | Shown from a usable download sample. |
| Unrated | Fewer than 3 usable TTFB samples or no usable download sample. | Insufficient evidence. | Insufficient evidence. |
Match the Measurement to the Workload
Use the rating as a first filter, then test the behavior the application actually needs. An interactive request and a long transfer can fail for different reasons.
Starting points for common proxy scenarios| Scenario | Useful starting signal | Additional validation |
|---|
| Short API requests | Fast or Average with low failure rate and stable TTFB. | Repeat against a permitted API-sized target from the deployment region. |
| Page retrieval | Enough throughput for expected page weight plus consistent TTFB. | Include redirects and assets only if the real task downloads them. |
| Bulk downloads | Sustained throughput matters more than one fast connect sample. | Run a longer authorized transfer and watch rate limits and completed bytes. |
| Video or streaming | The 8 MiB probe can reject obviously weak paths but cannot establish streaming quality. | Use a longer sustained-throughput test with the actual bitrate, duration, buffering, and concurrency requirements. |
Why Proxy Speed Changes and How to Troubleshoot
Proxy occupancy, provider rate limits, target throttling, peering, route changes, packet loss, TLS connection reuse, test-node region, and time of day can all move the samples. A shared proxy may perform differently seconds later, while a residential route can add distance compared with a nearby data-center exit. One fast run does not establish capacity under concurrency.
First repeat the same endpoint and protocol. Next compare failure rate and individual TTFB samples, then retest from the deployment region. Confirm provider concurrency and bandwidth limits, compare another endpoint from the same plan, and check whether only the fixed target route is affected. Change one factor at a time and keep timestamps so a provider can investigate.
- Latency
- Delay for a particular operation or round trip. Here, connect time and TTFB are distinct latency measurements.
- Bandwidth
- The theoretical or provisioned capacity of a link. This page cannot directly discover every link’s configured capacity.
- Download speed
- Observed application transfer rate on this exact path during the four-stream download. It can be below available bandwidth because of loss, limits, or protocol behavior.
Is this proxy latency test measuring the route from my computer?
No. Requests start from the current MiyaIP application server, travel through the proxy, and reach fixed probes. Your local carrier, Wi-Fi, and device performance are not part of this measurement path.
Is proxy connection time the same as ICMP ping?
No. HTTP and SOCKS proxies generally do not forward ICMP. MiyaIP measures TCP connection time from its application server to the proxy and reports TTFB separately so the two timings are not conflated.
Why does proxy TTFB change between repeated tests?
Network routing, proxy load, target congestion, and transient packet loss can all change TTFB. Repeat the test under similar conditions and compare the minimum, average, maximum, and failure rate instead of relying on one run.
Why can proxy download speed differ from a browser Speedtest?
A browser Speedtest measures your device against its selected server. This tool measures the route from the current MiyaIP server through the proxy to fixed download probes, so the starting point, target, route, and exit can all differ.
Does this proxy speed test measure upload speed?
No. The current test performs server-side downloads through the proxy and does not upload a test payload. Displayed throughput applies only to this download route.
How much proxy traffic does the 8 MiB download test use?
The default download phase opens four streams of 2 MiB each, so four successful streams use about 8 MiB. The six lightweight requests and automatic protocol detection can add a small amount of traffic.
Are the proxy username, password, and host saved after a speed test?
No. Credentials are used only for the current server connection. Responses, exports, and recent history do not contain credentials, and recent history does not save the proxy host.
What is a good proxy TTFB?
It depends on the workload and test region. MiyaIP uses 300 ms in the Fast rule and 800 ms in the Average rule, but reliability, throughput, target distance, and your own deadline must be considered too.
Why is a fast proxy sometimes rated Average?
The rating combines successful probe groups, average TTFB, and download throughput. Missing any Fast requirement produces another rating even if one individual metric looks quick.
Can the 8 MiB test predict video streaming?
It can reveal an obviously weak short download path, but video needs a longer sustained-throughput test that reflects the actual bitrate, duration, buffering strategy, destination, and simultaneous streams.
