1.

Preparation and goal definition

- Clarify your needs: target users (Mainland China/Hong Kong, Macao and Taiwan/others), bandwidth scale (Mbps/Gbps), maximum allowed delay and packet loss rate; - Determine the test period (it is recommended that at least 72 hours include peak and low peaks); - Prepare testing resources: VPS or PC in mainland China (for backhaul testing), local testing machine, and Vietnam or international VPS (for forward testing).

2.

Understand CN2 and its types

- Understand that CN2 is China Telecom’s high-quality backbone network. Commonly used are CN2 GIA and CN2 GT. GIA has lower latency and higher priority; - Ask the service provider about the type of CN2 used and ask for a routing diagram to confirm whether it is a real direct connection to CN2 and not ordinary international bandwidth.

3.

Select the list of candidate service providers

- List 5-8 companies based on word-of-mouth, forums, third-party ratings, and upstream operator relationships; - Each company is required to provide AS number, outbound POP location, CN2 exit node, bandwidth model, whether it supports BGP multi-line, trial period/refund policy.

4.

Prepare test account and environment setup

- Apply for a trial or test account from each company for at least 7 days (you can request speed limit but real routing); - Prepare a public network VPS in mainland China (preferably under a telecommunications network), and prepare a test VPS in Vietnam or abroad; - Install the necessary tools: mtr/traceroute/ping/iperf3/speedtest-cli/tcping, and ensure that you can remotely control and record logs.

5.

Actual measurement steps of delay and jitter

- Execute the Vietnamese test IP on the Chinese VPS: ping -D -i 0.2 -c 1000 (record each RTT), and mtr -r -c 100 to obtain the hop count and packet loss; - Calculate average latency, median, 99th percentile latency and jitter (stddev), and compare data between working hours and non-working hours; - Require the service provider to provide peak time period data and explain the reasons for abnormal fluctuations.

6.

Packet loss and link stability test

- Use mtr to run continuously for 2-3 hours to observe packet loss per hop, and use tcptraceroute or tcping to test the TCP port (80/443/22), because ICMP filtering may not be accurate; - Repeat the test during peak periods and record the end-to-end packet loss rate (<=0.5% is good); - If intermittent packet loss occurs, check whether it is a problem with the local export, Vietnam access, or the last hop in China.

7.

Throughput and Concurrency Performance Test

- Use iperf3 for two-way testing: from China VPS to Vietnam iperf3 -c -P 10 -t 60 to record bandwidth and retransmission rate; - Test under different time periods and different concurrency numbers to evaluate the linear scalability and peak capacity of the link; - If possible, require the service provider to guarantee dedicated bandwidth or provide bandwidth slicing instructions before testing.

8.

Routing and BGP information verification

- Use traceroute to view the autonomous systems (AS) that actually pass through; use bgp.he.net or major route viewers to query the upstream relationship of the peer AS; - Require the service provider to provide BGP notification strategies, whether there are fixed routes or community values, and whether multiple backup routes are supported; - Use China Telecom, China Unicom Looking Glass (or 3rd party LG) to view backhaul routing consistency from different points in China.

9.

Long-term monitoring scripts and automated assessment

- It is recommended to deploy a ping+mtr scheduled script (for example, once every 5 minutes) and report the results to Influx/Grafana or CSV for at least 72 hours; - Use smokeping or Prometheus to record latency trends and packet loss bursts to facilitate identification of intermittent faults; - Automatically compare data from multiple service providers, generate visual reports, and score based on delay, packet loss, jitter, and throughput.

10.

Business terms, safety and after-sales evaluation

- Check SLA terms (bandwidth, availability, failure response time, compensation mechanism); - Understand whether DDoS protection is included and whether traffic cleaning and burst traffic are supported; - Test customer service response: Submit a work order or fault report, and record the first response time and problem resolution cycle.

11.

Decision Matrix and Negotiation Points

- Establish a scoring table: delay (30%), packet loss (25%), bandwidth stability (20%), after-sales (15%), price (10%), and score based on test data; - Give priority to suppliers that meet latency and packet loss requirements and have guaranteed after-sales service. Prices can be used as final adjustments; - When signing the contract, it is clear that the refund terms, SLA indicators and technical contacts will be retained during the test period.

12.

Final inspection before deployment and launch

- Conduct a 24-hour full test before going online to confirm that BGP policies, ACLs, firewalls, and application ports are performing normally; - Configure monitoring alarms (delay/packet loss thresholds) and rehearse the failover process; - Save all test logs and conversation records for future claims or technical communication.

13.

FAQ 1: How to confirm that the other party is really using CN2 instead of an ordinary international link?

- You should ask the other party to provide the exit AS and CN2 exit node information, and use traceroute/mtr to check whether the path passes through China Telecom CN2's AS (usually it can be identified from the AS path and BGP information). You can also ask the other party to provide a routing diagram and Looking Glass test screenshot to prove it.

14.

FAQ 2: What should I do if ICMP is filtered during testing? How to ensure the validity of the test?

- Prioritize the use of TCP testing (tcping, tcptraceroute, iperf3 over TCP), and explain to the service provider the need to open test ports or provide bandwidth test endpoints when requesting a trial; at the same time, try to use telecom export VPS on the Chinese side to ensure that the return journey truly reflects user experience.

15.

FAQ 3: If the short-term performance of the test is good, but there are fluctuations after going online, how should I protect my rights?

- State the SLA and refund/compensation terms in the contract in advance, and save all test data and work order records; when problems arise, submit a work order first and ask for technical traceback routing logs. If the service provider cannot solve the problem according to the SLA, it can claim compensation according to the contract or terminate the contract in advance.