Subnetting Explained: How IP Subnet Masks Divide a Network
2026-09-05 · undefined min
Every device on a network needs an IP address, but those addresses are not handed out one by one. Networks are split into blocks called subnets, and the tool that defines the boundary of each block is the subnet mask. Understanding subnetting is what turns a confusing string of numbers into a clear map of your network.
In this guide we break down how a subnet mask works, what CIDR notation means in practice, how to count the usable hosts in a subnet, and how large networks are carved up or merged together. By the end you will be able to read a subnet like a network engineer.
1981
IPv4 and the birth of classful addressing
RFC 791 defines IPv4 and introduces the classful scheme, dividing addresses into fixed A, B and C classes with rigid network boundaries.
1993
CIDR replaces the class system
RFC 1519 introduces Classless Inter-Domain Routing, allowing arbitrary prefix lengths and ending the wasteful fixed-class approach.
1995
Variable-length subnet masks
VLSM lets a single network use different mask sizes, making address allocation far more flexible and efficient.
Today
Careful subnetting under IPv4 scarcity
With IPv4 exhausted, precise subnetting is essential for conserving addresses, while IPv6 brings a return to generous address space.
What is a subnet mask?
A subnet mask is a 32-bit number that sits next to an IP address and decides which part identifies the network and which part identifies the host. Bits set to 1 mark the network portion, while bits set to 0 mark the range of hosts available within that network. Two devices that share the same network part are on the same subnet.
The mask is written like an IP address, for example 255.255.255.0, and it is the router's guide to how far traffic can travel before it must be forwarded elsewhere. Devices on the same subnet can talk directly, while traffic to another subnet is handed to a router. Changing the mask reshapes the whole network.
From masks to CIDR notation
CIDR notation expresses the same information as a mask, but far more compactly. Instead of 255.255.255.0 you write /24, where the number is the count of 1-bits in the mask. This is why a /24, a /16 and a /8 correspond to different network sizes, from a small office to the public internet.
The slash number tells you how many bits are fixed for the network. A /24 leaves 8 bits for hosts, a /16 leaves 16, and a /8 leaves 24. More network bits mean fewer hosts, so choosing a prefix is a trade-off between the number of subnets you need and the hosts each one must hold.
How a /24 subnet mask splits an address into a network part and a host part.
How many hosts can a subnet hold?
To count usable hosts, you first work out the number of host bits by subtracting the prefix length from 32, then raise two to that power. A /24 has 8 host bits, so it covers 2^8, or 256, total addresses. The formula 2^hostbits gives the whole block, but not every address can be assigned to a device.
The very first address is the network identifier and the very last is the broadcast address, so the usable hosts are 2^hostbits minus 2. A /24 therefore holds 254 usable hosts. For a point-to-point link a /30 is the common choice, leaving exactly two addresses so each end gets one.
Supernetting: merging networks back together
Supernetting is the opposite of subnetting. Instead of carving one block into smaller pieces, it merges several contiguous subnets into one larger block by borrowing bits back from the host portion. This is how providers summarize many customer routes into a single advertised network, keeping the global routing table small and efficient.
The practical benefit is route summarization: fewer routes mean smaller tables and faster forwarding across the internet. A classic example is aggregating two /24 networks into a single /23, which covers both with one entry. Used well, subnetting and supernetting let you design networks that scale cleanly without wasting address space.
The takeaway
Subnetting is one of the most useful skills for anyone working with networks. Once you can read a mask and a CIDR prefix, you understand how IP space is organized, how many devices a subnet supports, and how large networks are summarized into manageable routes.
Your public IP address and network location, detected automatically.
Loads automatically when you open NetChecks — no input needed. Use the Refresh button to re-check after switching networks or reconnecting your VPN.
Your browser
All-in-One Scan
Runs every relevant check against one IP or hostname in a single pass: DNS, whois, ping, traceroute, a well-known-ports scan (1-1024), HTTP headers and the SSL certificate.
Enter a domain or IP address and run it to check DNS, whois, ping, traceroute, common ports, HTTP headers, and the SSL certificate all at once.
Most checks run in parallel - typically finishes in about 30 seconds, longer if the target is slow or unreachable.
The port scan step only runs once the consent checkbox above is checked - every other check runs regardless.
Ping
Send ICMP echo requests to a host to check reachability and latency.
Enter a hostname or IP address and press Ping to send ICMP echo requests and measure round-trip latency.
Learn more about Ping
What it is
Ping sends ICMP Echo Request packets to a host and measures how long it takes for ICMP Echo Reply packets to come back. It is the most basic network connectivity test there is: it answers exactly one question, "is this machine reachable, and how quickly?" The ICMP protocol (RFC 792) was designed back in 1981 specifically to carry control and diagnostic messages over IP networks, outside of any application traffic - Ping is its best-known and most universally available implementation, present on virtually every operating system and network device since their earliest versions.
How it works
Every ICMP packet carries a TTL (Time To Live) field decremented by one at each router it crosses; if it hits zero before reaching the target, the packet is dropped and an error is sent back to the sender. The round-trip time (RTT) measured in milliseconds reflects the cumulative network latency over the entire round trip, not just the last leg near the target - a point that is often misunderstood, since slow ping results can have their root cause anywhere along the path, not necessarily near the server being tested. A ping typically sends several packets in a row rather than just one, which lets you tell a one-off latency spike apart from a recurring problem and calculate a packet loss rate over the sample.
Reading the results
A stable, low RTT - a few milliseconds on a local network, 10 to 50 ms for a destination within the same country, and noticeably more for an intercontinental link - indicates a healthy connection. Even a small amount of packet loss (above 1-2%) is particularly damaging for latency-sensitive uses like VoIP or interactive remote sessions, where every dropped packet shows up as a glitch or audible cutout. Highly variable latency from one packet to the next (jitter) is often more of a problem for these same use cases than latency that is high but perfectly stable. "Request timed out" means no reply arrived within the allotted time - the host could genuinely be down, a firewall could be silently blocking ICMP, or a route could be broken somewhere along the path; "Destination unreachable" is different and more informative: an intermediate router explicitly sent back a message saying it could not forward the packet, which helps narrow down where the problem actually sits.
Common mistakes
The most common misreading is concluding a host is "down" the moment a ping fails, when in fact a great many servers and devices - especially behind a properly hardened firewall, or hosted with major cloud providers - deliberately block inbound ICMP as a matter of policy while being fully operational and reachable on their actual services (HTTP, a database, and so on). A missing ping reply is therefore only meaningful evidence of downtime when combined with other signals, such as the application itself also failing to respond. Conversely, a successful ping is no guarantee whatsoever that the application service hosted on that machine is working correctly - these are two entirely independent layers of the network stack.
When to use it
The first thing to check before escalating a ticket: does the machine respond at all, before digging any further? Confirming connectivity after a firewall rule or routing table change, to make sure the change did not break access. Establishing a baseline latency measurement before a VoIP rollout or a carrier link failover, so there is an objective point of comparison if call quality complaints come in later. A lightweight, low-overhead periodic health check for an MSP monitoring several client sites in parallel, always as a complement to - never a replacement for - deeper application-level monitoring.
Traceroute
Trace the network path (hop by hop) to a destination host.
Enter a hostname or IP address and run it to see every network hop between this server and the destination, with latency per hop.
DNS Lookup (Nslookup)
Query DNS records: A, AAAA, MX, TXT, NS, CNAME, SOA, PTR, SRV, CAA.
Enter a domain, choose a record type (A, AAAA, MX, TXT, NS, CNAME, SOA, PTR, SRV, or CAA), then look it up.
Whois
Look up registration information for a domain or IP address.
Enter a domain or IP address to look up its registration details — registrar, owner organization, and important dates.
Blacklist Check
Check whether an IP address or domain is listed on public spam/abuse blocklists (DNSBL).
Enter an IPv4 address or domain and run it to check 7 public DNSBL/RBL blocklists at once - each shows as listed, not listed, or check failed.
TCP Port Scan
Check whether TCP ports are open on a host or IP: common ports, a custom list, or the full 1-65535 range.
Enter a host or IP, pick common ports, a custom list, or the full range, then scan to see which TCP ports respond.
A full range scan can take up to ~100 seconds. Large scans (more than 100 ports, including custom ranges/lists) are limited to 1 per minute per visitor.
Open ports (0)
HTTP Header Inspector
Fetch the HTTP response status and headers for a URL.
Enter a URL to fetch its HTTP response status code and every response header the server sends back.
SSL / TLS Certificate Checker
Inspect a host's TLS certificate: issuer, validity dates, and days remaining.
Enter a hostname to inspect its TLS certificate — issuer, validity dates, and days remaining before it expires.
Geo-IP Lookup
Look up the geographic location and network info for an IP address. Leave empty to look up your own public IP.
Enter any IP address, or leave it empty to look up your own, to see its approximate location and network/ISP info.
Computed entirely in your browser — no data sent to the server.
Enter an IP address and CIDR prefix (e.g. 192.168.1.0/24) to instantly compute the network range, broadcast address, and usable host count.
Speed Test
Basic download/upload throughput test against this server (accuracy depends on the server's own uplink).
Press Start to measure download and upload throughput against this server. Accuracy depends on this server's own connection.
Country Code Dictionary
ISO 3166-1 alpha-2 country codes — searched entirely in your browser.
Search or browse the list of ISO 3166-1 alpha-2 country codes, looked up entirely in your browser.
Country
ISO Code
Phone Dialing Code Dictionary
International calling codes by country — searched entirely in your browser.
Search or browse international calling codes by country, looked up entirely in your browser.
Country
Dial Code
World Clock
Pick a time zone to see the current time — drag the globe to spin it.
Pick a time zone from the list, or drag the globe, to see the current time there.
Your time
--:--:--
—
—
Selected time
--:--:--
—
—UTC±00:00
Difference vs. you—
—
Drag to rotate the globe.
French Mobile Network Status
Mobile antenna sites down or under maintenance in France, by operator (Orange, Free, SFR, Bouygues Telecom), from ARCEP's public data. Snapshot updated once a day by ARCEP - not a minute-by-minute feed.
Browse mobile antenna and fibre outage data by French operator — no input needed, updated automatically from ARCEP's public data.
Source: ARCEP, "Sites indisponibles" dataset, published under Licence Ouverte / Etalab 2.0 - commercial reuse explicitly allowed, unlike the IODA/CAIDA data this tab used before. The Normal/Watch/Alert badge is an in-house estimate (today's outage count vs. the median of prior days), not an official ARCEP classification. Source links below.
Most affected departments
Number of sites currently down or under maintenance, by department. Click an operator above to filter.
Fibre (FTTH) network quality by operator: reported-outage rate and connection-failure rate, from ARCEP's public data. Monthly indicators, 6-month rolling average - not a live feed like the mobile section.
Source: ARCEP, "Qualité des réseaux en fibre optique" dataset, published under Licence Ouverte / Etalab 2.0 - commercial reuse explicitly allowed. Source links below.
By operator (parent group)
Averages over the last 6 available months, per infrastructure-operator parent group.
NetChecks is provided for legitimate network diagnostics and educational purposes only, on systems, domains, and IP addresses that you own or are explicitly authorized to test.
Running port scans, DNS/whois lookups, traceroutes, or other checks against third-party systems without authorization may violate computer misuse laws in your jurisdiction (for example the Computer Fraud and Abuse Act in the US, the "Loi Godfrain" in France, or equivalent local legislation). You are solely responsible for ensuring you have the right to test any target you enter here.
This site and its results are provided "as is", without warranty of any kind. The operator of this site accepts no liability for any direct, indirect, or consequential damages, legal consequences, or misuse arising from the use of this site.
Requests are rate-limited to prevent abuse. The operator of this instance may optionally enable request-activity logging - if so, each tool request (your IP address, timestamp, the tool used, and the target/query entered) may be recorded in a database for security, abuse-prevention, and usage-analytics purposes; this is disabled by default. By using this site you agree to these terms; if you do not agree, do not use it.
Programmatic (API) use of this site is subject to these same terms, including the authorization requirement above.
Send feedback
Report an issue, leave a comment, or suggest a new feature.
This site uses ads and, if you accept, ad-partner cookies to help fund hosting. You can decline and keep using every tool ad-free.