How to Test a VPS Before Buying #
Short version: do not trust a provider’s latency screenshot. Test the route yourself during the day and at peak time, on the ISPs your visitors use. Return routing matters as much as outbound routing.
What you need #
- The provider’s test IP or Looking Glass URL.
- A machine on your network or the target audience’s network.
- ping, mtr and traceroute; WinMTR on Windows.
- Test points on different ISPs if possible.
Metrics that matter #
- Average latency: distance and route, but not stability.
- Packet loss: more important than latency; anything above 1% at peak is a warning.
- Jitter: unstable latency hurts real-time applications.
- Return path: how traffic comes back from the data center.
- Throughput: test the actual port speed, not the label.
Commands #
ping -c 100 TEST_IP
mtr -rwzbc 100 TEST_IP
traceroute TEST_IP
On Windows:
tracert TEST_IP
winmtr TEST_IP
Run 50-100 packets, not four.
Test at the right time #
- Daytime: baseline latency and route.
- 8-11pm: international transit and host uplink congestion.
- Weekend: some routes behave differently.
- Repeat over 2-3 days before deciding.
Check the return path #
Look for:
- A China backbone node (CN2, CMI, 9929, etc.)
- A detour through the US or Europe
- Packet loss concentrated at one hop
- Latency jumping from 40ms to 200ms+
Using Looking Glass #
Looking Glass lets the provider run ping/traceroute from the data center. It is useful, but it tests the opposite direction; combine it with your own tests.
15 providers with a confirmed public test IP or Looking Glass, covering 679 listed plans.
| Provider | Type | Entry / test IP | Locations | Plans |
|---|---|---|---|---|
| GreenCloud | Test IP | Open Test IPs are listed on the data center page. | Los Angeles, San Jose, Phoenix, Las Vegas and more | 234 |
| JustHost | Looking Glass | Open Public Looking Glass entry. | Multiple locations | 99 |
| RackNerd | Test IP / Looking Glass | Open Test IPs are listed on each data center page; Looking Glass at lg-lax03.racknerd.com. | Los Angeles, San Jose, Seattle, New York and more | 85 |
| AlphaVPS | Looking Glass | Open Public Looking Glass for latency testing. | Los Angeles and more | 76 |
| ByteVirt | Looking Glass | Open Public Looking Glass for each location. | US, Japan, Singapore and more | 36 |
| SmartHost | Test IP | 23.228.222.50 Southern California45.35.132.2 Texas199.43.206.3 Nevada66.11.121.252 ColoradoPing the test IPs below before ordering. | California, Texas, Nevada, Colorado | 30 |
| ZgoVPS | Looking Glass | Open Public Looking Glass; the entry URL may change. | Los Angeles, Netherlands, Osaka | 25 |
| ServaRica | Test network | Open Test Our Network page. | Canada and more | 24 |
| VPSDime | Looking Glass | Open Public Looking Glass. | Multiple US locations | 18 |
| V.PS | Looking Glass | Open Speed Test page with Looking Glass. | Hong Kong, Japan, Singapore, US, Europe and more | 15 |
| ColoCrossing | Looking Glass | Open Los Angeles Looking Glass. | Los Angeles and more | 10 |
| HostHatch | Looking Glass | Open Public Looking Glass by location. | US, Europe, Japan, Singapore, Hong Kong and more | 10 |
| EthernetServers | Test IP | Open Test IPs are listed on the Network page. | New Jersey, Los Angeles, Jacksonville | 9 |
| Sharktech | Looking Glass | Open Public Looking Glass. | Los Angeles, Las Vegas and more | 7 |
| BuyVM | Test IP | 209.141.56.135 Las Vegas198.98.53.31 New Jersey104.244.76.191 LuxembourgPing the test IPs below before ordering. | Las Vegas, New Jersey, Luxembourg | 1 |
Data verified 2026-11-12. Test IPs and Looking Glass entries can change; check the provider page before ordering. Providers not listed are not necessarily without test IPs, only not yet confirmed.
Test record template #
| Item | Daytime | Peak | Notes |
|---|---|---|---|
| Average latency | |||
| Max latency | |||
| Packet loss | |||
| Return path detour | |||
| Download speed |
FAQ #
Is low ping enough? #
No. Low latency with high loss or jitter still feels bad. Use mtr and real application tests.
What if the provider has no test IP? #
Start with a small monthly plan, or use third-party data as a lead. Test yourself when possible.
Why do my results differ from someone else’s? #
ISP, region, time of day and local network all matter. Test from the target audience’s network.