What is Latency and Why It's Not the Same as Ping
Latency is the round-trip time of a packet from source to destination and back to the source. Its unit is milliseconds (ms). When you say "my ping is 40," you're actually talking about RTT (Round-Trip Time), not one-way latency. One-way latency is usually half of RTT, but this rule breaks on asymmetric paths; the outbound path goes through one carrier and the return path through another.
To measure it, use ping with the -c flag for the packet count:
ping -c 10 8.8.8.8The output includes min/avg/max. Look at the avg value, not min. Minimum latency is always misleading; the average reflects the reality of the path.
Latency of 1 to 20 milliseconds is normal on a local network. Above 100 milliseconds on an international path means traffic is going through a congested route or fiber optic cable with significant distance. The speed of light in fiber is about 5 microseconds per kilometer; so Tehran to Frankfurt, roughly 4,000 kilometers, has a physical minimum of 40 milliseconds RTT. Any number significantly lower than that means a more direct path or incomplete measurement.
This is where people make mistakes: confusing high latency with low bandwidth. Downloading a 100 MB file with 200 ms latency and 100 Mbps bandwidth takes about 8 seconds. The same file with 5 ms latency and 1 Mbps bandwidth takes 800 seconds. Two completely different problems with two different solutions.
Bandwidth and Throughput: The Number the Vendor Tells You Differs from What You Get
Bandwidth is the maximum theoretical capacity of a path, measured in bits per second (bps). Throughput is the amount that actually gets transferred. The difference between the two is protocol overhead: TCP performs a three-way handshake, every packet has a header, and ACKs come back.
A rule of thumb: actual TCP throughput on a link is typically 80 to 90 percent of the nominal bandwidth. If your service is 10 Mbps and your download runs at 1.2 MB/s (9.6 Mbps), everything is as it should be. For precise measurement, use iperf3:
iperf3 -c server_ip -t 30The -t flag specifies the test duration. Thirty seconds is a fair minimum; shorter tests are affected by TCP slow start and show lower numbers.
Take units seriously. Megabits (Mb) differ from megabytes (MB); the factor is 8. Providers almost always quote megabits, while download tools show megabytes. If your download speed is 12.5 MB/s, that means you have 100 Mbps.
Why Hops Matter and How They Relate to Latency
A hop is the passage of a packet through a router. Each hop adds latency; typically 1 to 5 milliseconds on urban paths, and more on international routes. The traceroute tool (or tracert on Windows) shows the path:
traceroute -n example.comThe -n flag skips reverse DNS lookups and makes the output faster. Each line is one hop. Three asterisks (* * *) mean that router doesn't respond to ICMP; this isn't necessarily a problem—some routers deliberately don't respond.
The number of hops isn't a problem by itself. The problem is when a specific hop's latency suddenly jumps from 5 to 150 milliseconds. That point is the bottleneck. Look for the first hop with a sudden jump, not the last hop.
MTU: The Number That, If You Get Wrong, Packets Silently Get Lost
MTU (Maximum Transmission Unit) is the maximum size of a frame at the data link layer, measured in bytes. The standard value on Ethernet is 1500 bytes. On PPPoE (which many ADSL and fiber connections in Iran use), this number becomes 1492; the 8-byte PPPoE header subtracts from it.
If the path MTU is smaller than your interface MTU, two things can happen: large packets get fragmented, or if the DF (Don't Fragment) flag is set, they get dropped and an ICMP "Fragmentation Needed" message is returned. Many firewalls block this message, and the result is connections that open pages but downloads or uploads stall.
To find the right MTU, use ping with the -M do flag (on Linux) and -s for the payload size:
ping -M do -s 1472 -c 5 8.8.8.81472 bytes of payload + 28 bytes of IP/ICMP header = 1500 bytes. If you get a response, the path MTU is 1500. If not, decrease the number until you get a response. This is where people make mistakes: they set MTU to 1400 "just to be safe." The result is wasting 6 percent of bandwidth and more latency per packet. Measure MTU precisely, not carelessly.
NAT and Its Difference from Proxy: Both Change IPs, But Not in the Same Way
NAT (Network Address Translation) changes the source or destination address of a packet at the router. Its most common form is NAT overload or PAT, which hides thousands of internal devices behind a single public IP. The NAT table on the router maintains the mapping of internal IP:port to public IP:port.
A proxy, however, works at a higher layer. A proxy establishes a new connection to the destination on your behalf; the destination sees the proxy's IP, not yours. NAT only rewrites the address, but the client itself establishes the connection.
The practical difference is this: NAT doesn't work for protocols other than TCP/UDP without special modules (such as nf_conntrack_pptp for PPTP). A proxy can understand application-layer protocols like HTTP, cache them, and filter them. That's why NAT is sufficient on a home router, but organizations deploy proxies.
If you work with NAT and a service from outside needs to reach an internal device, Port Forwarding is required. On Linux with nftables:
nft add rule nat prerouting dnat to 192.168.1.100:8080This rule forwards incoming traffic on port 80 to port 8080 on the internal device. Without this line, the incoming packet reaches a destination nobody is waiting for and gets dropped.
DNS: The First Thing That Breaks and the Last Thing People Check
DNS (Domain Name System) converts domain names to IP addresses. An A record is for IPv4, AAAA for IPv6, MX for mail servers, and CNAME for aliases. When you say "the site won't open," check DNS first.
The dig tool gives the most precise output:
dig example.com A +shortThe +short flag shows only the final answer. If the output is empty, the record doesn't exist, or the TTL has expired and the DNS server has no answer. TTL (Time To Live) is the cache validity period of a record, in seconds. A short TTL (300 seconds) is good for rapid changes but puts more load on the DNS server.
This is where people make mistakes: after changing a DNS record, they get an answer with dig, but the browser still shows the old page. They forget the browser and operating system DNS cache. dig goes directly to the DNS server and bypasses the local cache; the browser uses its own cache. Two different paths, two different results.
Frequently Asked Questions
What's the difference between ping and latency?
Ping is both a tool and a colloquial term for latency. The ping output shows the RTT value, which is the complete round trip. One-way latency is usually half of RTT, but this proportion doesn't hold on asymmetric paths.
What is the appropriate MTU for Iran's internet?
If your connection is PPPoE (most ADSL and home fiber), it's 1492 bytes. If it's a direct Ethernet connection, it's 1500. You can find the exact path value with the ping -M do -s 1472 command on Linux. An MTU lower than necessary wastes bandwidth.
Why does traceroute show asterisks?
An asterisk means that router didn't respond to the ICMP packet. Some routers don't respond for security reasons or rate-limit ICMP traffic. If subsequent hops respond, the problem isn't with the unresponsive router.
Does NAT slow down the internet?
NAT itself doesn't add noticeable latency; address rewriting is on the order of microseconds. However, maintaining the NAT table on a weak router with thousands of simultaneous connections can saturate the CPU and cause latency. The problem is the hardware, not NAT.
Before you submit a ticket to support, have the ping and traceroute outputs ready. Those two outputs alone have done half the troubleshooting. If you're looking for an environment to run these commands and test real-world scenarios, ServerNet's Linux hosting is a suitable platform for practice and measurement. See the rest of the guides in the documentation and knowledge base.