DMIT LAX Eyeball vs Premium: Beijing Latency Test 2026
When choosing a DMIT Los Angeles VPS, the most common dilemma is how much difference there is between Premium (Pro, CN2 GIA line) and Eyeball (EB, home broadband optimized line). The official Looking Glass test nodes are provided, but most reviews only post a single ping screenshot. This article uses real measurement data from the evening of August 14, 2026, in Beijing, covering telecom home broadband and mobile 5G hotspot two access networks, to lay out the latency, jitter, and actual hop-by-hop routing of the two lines.
Test Environment and Method
- Test Date: 2026-08-14 (evening, including peak hours)
- Access Network 1: Beijing Telecom home broadband, exit IP
124.127.144.x - Access Network 2: Beijing Mobile 5G phone hotspot (IPv6 exit
2409:8900:3f7:*, IPv4 via CGNAT) - Test Tools: macOS built-in
ping(interval 0.25–0.3s) andtraceroute(UDP mode) - Test Target: DMIT official Looking Glass Los Angeles nodes, all located in the same data center West 7 Center. All three test IPs are taken from the official Looking Glass page lg.dmit.sh/?server=lax-eb (this is where the Eyeball and Tier 1 IPs come from):
| Line Tier | Test IP | Official Positioning |
|---|---|---|
| Premium (Pro) | 179.255.100.100 |
CN2 GIA optimized, three-network high-end return |
| Eyeball (EB) | 179.255.150.150 |
Home broadband (eyeball network) direction optimized |
| Tier 1 (T1) | 136.175.177.177 |
International Tier 1 forwarding, cost type |
Part 1: Beijing Telecom Home Broadband Real-World Test
Ping Latency Comparison (two rounds, interval about 10 minutes)
| Target | Sample | Packet Loss | min / avg / max (ms) | Jitter σ (ms) |
|---|---|---|---|---|
| Premium Round 1 | 15 | 0% | 154.2 / 154.9 / 156.3 | 0.61 |
| Eyeball Round 1 | 15 | 0% | 154.7 / 155.4 / 158.1 | 0.75 |
| Premium Round 2 | 30 | 0% | 154.0 / 161.7 / 251.9 | 23.9 |
| Eyeball Round 2 | 30 | 0% | 154.4 / 177.6 / 370.2 | 49.9 |
| Tier 1 (two rounds) | 15+30 | 100%* | — | — |
* The T1 test IP showed 100% ICMP loss from Beijing Telecom, and TCP connections to ports 8443/443/80 all failed as well — unreachable on every protocol. A control test from overseas (another instance in the same data center) told the opposite story: 0% ping loss (0.6ms) and TCP 8443/80 open. The node itself is alive and does respond to ICMP; the problem is specific to the China-bound direction, matching the signature of an IP-level block. Details below.
Two key readings:
- Baseline latency is completely equal: The min RTT for both Premium and Eyeball is around 154ms, indicating that from Beijing Telecom, the physical path length for both lines is nearly identical.
- Difference lies in jitter: In the first round (less busy period), both were stable as direct connections; in the second round entering a fluctuation period, Premium's standard deviation of 23.9ms is significantly better than Eyeball's 49.9ms, with Eyeball peaking at 370ms.
Traceroute Routing Comparison
Premium (179.255.100.100) outbound, hops 7–10 are the standard Telecom CN2 backbone (59.43.0.0/16):
3 100.123.0.1 7.0 ms ← 电信 CGNAT
6 202.97.117.134 8.8 ms ← ChinaNet 骨干
7 59.43.46.82 29.7 ms ← CN2 入口
8 59.43.138.58 28.9 ms
9 59.43.39.190 31.0 ms
10 59.43.184.158 158.6 ms ← CN2 跨太平洋
11 218.30.48.142 158.1 ms ← 电信美洲(CTA)
12 193.41.250.193 175.9 ms ← DMIT 边缘
13 193.41.250.149 159.6 ms
Eyeball (179.255.150.150) outbound—also entirely via CN2, just with different specific links:
6 202.97.117.134 9.2 ms
7 59.43.46.86 29.0 ms ← CN2 入口
8 59.43.130.218 29.9 ms
9 59.43.181.54 31.1 ms
10 59.43.246.238 155.5 ms ← CN2 跨太平洋
11 218.30.48.50 158.9 ms ← 电信美洲(CTA)
12 193.41.250.173 164.6 ms ← DMIT 边缘
15 193.41.250.193 173.8 ms
Tier 1 (136.175.177.177) outbound is completely different—uses regular 163 backbone + Arelion (formerly Telia):
6 202.97.34.78 7.7 ms ← 163 骨干
8 202.97.59.106 178.2 ms ← 163 出境（延迟直接跳变）
9 62.115.185.212 179.0 ms ← Arelion (AS1299)
10 62.115.139.16 150.1 ms
12 193.41.250.178 182.2 ms ← DMIT 边缘
Three conclusions:
- From Beijing Telecom access, both Premium and Eyeball outbound are sent into CN2, which explains why their min RTT is nearly identical;
- Return path differs: Eyeball return TTL=47, Premium return TTL=49, the hop counts are different, and the jitter gap during fluctuation periods (49.9 vs 23.9) likely comes from the return path;
- Tier 1 uses 163 outbound and shows a 178ms jump at the international exit, which is a typical 163 congestion pattern.
Evidence That the Tier 1 Test IP Appears to Be Blocked
The behavior of 136.175.177.177 deserves its own section, because this is not a "poor route" — it is "cannot get there at all":
| Probe | From Beijing Telecom | From overseas (instance in the same LAX data center) |
|---|---|---|
| ICMP ping | 100% loss | 0% loss, 0.6ms |
| TCP 8443 (LG service port) | Connection failed | Open |
| TCP 80 | Connection failed | Open |
| UDP traceroute | Crosses the border, reaches DMIT edge 193.41.250.x | — |
Outbound probe packets leave China normally (the TTL-exceeded replies come back from Arelion and DMIT edge routers), but no packet sourced from 136.175.177.177 ever arrives — not a single ICMP echo reply or TCP SYN-ACK. From an overseas vantage point the node is perfectly healthy. This combination of "outbound visible, return completely dead" is the classic signature of an IP-level block. This was verified from a single vantage point (Beijing Telecom); other carriers were not tested.
Part 2: Beijing Mobile 5G Hotspot Real-World Test
After switching to mobile hotspot (IPv6 exit 2409:8900:3f7:*), continuous 20 pings were performed on the Eyeball node:
| Target | Sample | Packet Loss | min / avg / max (ms) | Jitter σ (ms) |
|---|---|---|---|---|
Eyeball 179.255.150.150 |
20 | 0% | 172.1 / 211.5 / 326.1 | 45.3 |
Another DMIT instance in the same data center 179.253.232.x |
20 | 0% | 167.8 / 207.4 / 353.1 | 45.3 |
Mobile side traceroute shows that the outbound paths for both targets are identical—both use mobile backbone 221.183.x (CMI overseas), and after exit, the path is invisible to UDP probes:
6 223.104.42.18 23.9 ms ← 北京移动
7 211.136.60.13 23.1 ms
8 221.183.39.101 25.3 ms ← 移动骨干
9 221.183.184.190 44.0 ms
10 111.24.3.101 49.2 ms ← CMI
11 221.183.179.22 64.5 ms ← 出境前最后可见跳
Comparing with telecom side data is straightforward: Mobile to DMIT LAX min RTT is higher by about 18ms (172 vs 154), average latency is higher by about 50ms, and jitter consistently sits around 45ms. Mobile outbound is determined by carrier-side routing policy, and Eyeball and other instances in the same data center have no difference in outbound; differences can only appear in the return path (in this mobile side test, return TTL for both were 46 and 43 respectively, paths are indeed different, but latency statistics show no perceptible difference).
Conclusions and Purchase Recommendations
Based on 2026-08-14 Beijing dual-network real-world tests:
- Beijing Telecom users: Baseline latency for Premium and Eyeball is identical (both ~154ms, outbound via CN2), but during fluctuation periods, Premium has significantly lower jitter. For latency-sensitive applications (games, real-time audio/video), choose Premium; for budget priority and acceptable peak jitter, Eyeball is fully sufficient.
- Beijing Mobile users: Outbound via CMI, there is no difference between the two lines on outbound, latency is overall higher than telecom (avg 207–211ms, σ45ms). For mobile users, the marginal benefit of paying more for a higher line tier is not significant.
- Tier 1 tier: Its LG test IP is unreachable on every protocol from Beijing Telecom while being fully reachable from overseas — it appears to be blocked, so this test could not evaluate T1 route quality from within China. Even setting the block aside, its outbound uses 163 + Arelion with a congestion jump right at the international exit; its positioning remains a cost option for the outbound direction (US local/international interconnection), not for direct access from China.
- All tests can be reproduced on DMIT official Looking Glass, it is recommended to run a test on your own access network, during your own peak hours before ordering—this data only represents a single point in Beijing, single evening situation.
FAQ
What does Eyeball line mean?
Eyeball network (home broadband network) refers to carrier networks primarily serving home broadband users. DMIT's Eyeball tier is a line optimized for return paths in scenarios of "access from home broadband", as opposed to Tier 1 forwarding for data center/enterprise interconnection.
Is Premium the same as CN2 GIA?
The official positioning of DMIT Los Angeles Premium (Pro) tier is CN2 GIA optimized line. In this real-world test, its outbound indeed used CN2 backbone (59.43.x) throughout, consistent with the official description.
Why can't Tier 1 node be pinged?
It is not the node's fault — the IP appears to be blocked. Testing 136.175.177.177 from overseas shows everything working (0% ping loss, TCP 8443/80 open), yet from Beijing Telecom both ICMP and TCP are unreachable, with only traceroute's intermediate router replies getting through. "Fully reachable from abroad, completely dead from China" is the classic pattern of an IP-level block. Readers who want to evaluate the T1 route themselves should first confirm the test IP is reachable on their own network.
Is it worth for Mobile users to buy Premium?
In this real-world test, mobile outbound is determined by CMI and unrelated to line tier; Eyeball and other instances in the same data center have identical latency statistics. If you only have mobile access, it is recommended to test with Looking Glass first before deciding whether to pay for a higher line tier.