Measured Nov 18, 2026
Datacenter proxy concurrency: 50, 200 and 500 sessions at once, measured
We opened 50, 200 and 500 simultaneous sessions through one QuanticData datacenter account, 20 seconds per level, one request per second per session. At 500 sessions at once, 9,545 of 9,546 requests were answered, and the median latency stayed at 235 to 237 ms at every level. Concurrency is not billed: you pay per GB or per day.
cloud server IPs · 1 Gbps+ · rotating & static
Measured at 500 sessions at once · Per GB from $0.50 · Per-day plans from $10
- 300+ cryptocurrencies
Pay as you go · No subscription · No KYC
City, state & ISP targeting
Per-GB plans target 200+ countries by country, state and city.
1 Gbps+ speed
<50ms average latency, no throttling
rotatingstatic
Rotating & static
fresh IP per request, or one consistent exit
HTTP · HTTPS · SOCKS5
HTTP, HTTPS and SOCKS5, with a username and password or an IP allowlist.
How many concurrent sessions can a datacenter proxy handle?
At 50: 50 sessions open, 100% of 994 requests answered, median 237 ms. At 200: 199 sessions open, 99.97% of 3,759 requests answered, median 235 ms. At 500: 500 sessions open, 99.99% of 9,546 requests answered, median 235 ms. Latency did not grow with the load: the median stayed between 235 and 237 ms at every level, so throughput grew with the number of sessions. Each session got its own exit IP (50, 196, 490 distinct addresses).
50 sessions
100% of 994 requests
median 237 ms · p95 369 ms
200 sessions
99.97% of 3,759 requests
median 235 ms · p95 491 ms
500 sessions
99.99% of 9,546 requests
median 235 ms · p95 452 ms
| Sessions opened | Sessions established | Requests | Answered | Latency median / p95 / p99 | Session setup median / p95 | Distinct exit IPs |
|---|---|---|---|---|---|---|
| 50 | 50 | 994 | 100% (994) | 237 / 369 / 581 ms | 562 / 1,057 ms | 50 |
| 200 | 199 | 3,759 | 99.97% (3,758) | 235 / 491 / 1,027 ms | 583 / 2,305 ms | 196 |
| 500 | 500 | 9,546 | 99.99% (9,545) | 235 / 452 / 735 ms | 590 / 1,531 ms | 490 |
Source: summary.csv, our own measurement on Nov 18, 2026.
What the test shows about concurrent sessions
From one machine, 500 simultaneous sessions through one datacenter account were served without slower answers: the median at 500 sessions was the same as at 50. That is the level we measured, and it held. If your job runs more sessions than that from one machine, run the same test on your own workload before you buy volume: the trial and the per-day plan are made for it.
Concurrency is not the same as permission from the target. A rotating gateway spreads sessions over many IPs, so a site that limits one IP sees each address only a few times; a site that limits an account or a header pattern still sees one client. The datacenter vs residential test lists the sites where a datacenter address returned the full page.
How we tested concurrency
- Target:
https://www.cloudflare.com/cdn-cgi/trace, a diagnostic endpoint that returns a few hundred bytes, including the IP address it sees. It runs on a network built for this kind of traffic, so the target is not the bottleneck, and it lets us count the exit IPs. - Proxy: QuanticData datacenter, rotating, United States exits, per-GB plan, one account and one set of credentials for every session.
- Levels: 50, 200, 500 sessions, one level after the other with a 15-second pause, 20 seconds each. All the sessions of a level were opened within the first second.
- A session: one CONNECT tunnel through the proxy and one TLS connection to the target, kept open; it sends one request per second and waits for the answer before the next one.
- Answered: HTTP 200 with the IP line in the body, within 30 seconds.
Limits of this test
One test machine (Windows 11, 32 logical cores, 95 GB of RAM, Node.js v22.23.2) on one consumer internet line in Europe, connecting to US exits; the client used at most 10.3% of one core. 20 seconds per level is short: it shows whether a level opens and is served, not how it holds for hours. One target with a tiny response, one country of exits, one time of day (Nov 18, 2026).
Data
summary.csv has one row per level: start time, sessions established, requests sent and answered, session setup and latency percentiles, and the count of distinct exit IPs. The exit IPs themselves are left out.
FAQ
Questions buyers ask about datacenter proxies, answered from the measurements on this page.
Something else? Ask usHow many concurrent sessions can a datacenter proxy handle?
On QuanticData we measured 50, 200 and 500 simultaneous sessions from one machine on Nov 18, 2026. At 500 sessions, 99.99% of requests were answered, with the same median latency as at 50.
How many sessions did one account hold at once in the test?
500 at once, from one account and one machine, opened within one second and served for 20 seconds with a median latency of 235 ms. That is what we measured; for a larger job, test your own workload on the trial or a per-day plan.
What is a concurrent session on a proxy?
One open connection through the proxy at the same moment as others: a browser tab, a scraper thread or a worker. Each session in this test was one tunnel through the gateway with its own exit IP, sending one request per second.
Does more concurrency slow each request down?
Not in our test: the median latency was 237 ms at 50 sessions and 235 ms at 500, so throughput grew with the number of sessions.
Do I get a different IP for each session?
On the rotating datacenter gateway, yes: each new tunnel takes a fresh exit IP. We counted 490 distinct exit IPs for 500 sessions at 500. A static session keeps one IP instead.
Is concurrency billed?
No. Per-GB plans bill the traffic, from $0.70/GB at 100 GB down to $0.50/GB at 500 GB and 1 TB; unlimited-bandwidth plans bill the time, $10 for 24 hours, $50 for 7 days, $150 for 30 days. The number of sessions is not part of the price.
Run your own job at the concurrency you need
Datacenter proxies per GB from $0.50, or unlimited bandwidth from $10 for 24 hours. No subscription, no KYC.