What is MTU?

MTU (Maximum Transmission Unit) is the largest packet size in bytes that a network interface forwards in a single frame without fragmentation. On Ethernet the standard MTU is 1500 bytes for the layer-3 payload (the full frame with header and FCS is 1518 bytes). When a packet is larger than a link's MTU, it gets split into smaller pieces or dropped. A correctly sized MTU along the whole path drives performance and reliability, especially in carrier networks and when tunneling.

Standard MTU and jumbo frames

The classic Ethernet MTU is 1500 bytes. Jumbo frames carry a payload larger than 1500 bytes, most commonly up to 9000 bytes. Larger frames cut header overhead and CPU load on high-throughput flows, so they appear in storage networks, compute clusters, on backbone links and at internet exchange points. The condition is a consistent MTU across the whole L2 path. One port with a lower MTU negates the benefit or causes packet drops.

• 1500 bytes: standard Ethernet MTU for the IP payload, the default on most networks.
• 9000 bytes: typical jumbo MTU in data centers and on iSCSI/NFS storage links.
• MTU must be consistent on every L2 hop within a domain. A mismatch creates a black hole for large packets.
• Tunnel headers (GRE, IPsec, VXLAN, MPLS) take space from the payload and lower the effective MTU.

Fragmentation and Path MTU Discovery

When an IPv4 packet exceeds a link's MTU and the DF (Don't Fragment) flag is not set, the router fragments it into smaller pieces reassembled at the receiver. Fragmentation costs resources and raises the risk of losing the whole packet. Path MTU Discovery (PMTUD, RFC 1191 for IPv4 and RFC 8201 for IPv6) avoids this. The sender sets the DF flag, and a router that cannot forward the packet returns an ICMP 'Fragmentation Needed' (IPv4) or ICMP 'Packet Too Big' (IPv6) message carrying the MTU value. IPv6 routers do not fragment transit traffic, so PMTUD is mandatory there. Trouble starts when firewalls block these ICMP messages, creating a PMTUD black hole.

MSS clamping as a practical fix

MSS (Maximum Segment Size) is the maximum TCP segment size, negotiated during the three-way handshake. For a 1500 MTU with 20-byte IPv4 and 20-byte TCP headers, the MSS defaults to 1460 bytes. MSS clamping is a technique where a router on the path rewrites the MSS value in SYN packets down to one matching the smallest MTU on the route. TCP sessions then never build packets too large for a link, even when PMTUD fails due to ICMP filtering. It is standard practice on PPPoE, IPsec or GRE tunnels, where the effective MTU is below 1500.

MTU in the AS202520 SkyPass network

At AS202520 SkyPass we keep MTU consistent across the IP transit path and BGP sessions, and on backbone links and peering ports we support jumbo frames where the exchange-point infrastructure allows it. For remote access to Polish IXPs (THINX, TPIX, WRIX, POZIX, 1-IX) and when setting up peering, we help align MTU and, where needed, apply MSS clamping to avoid fragmentation and PMTUD black holes at the edge of our network in Warsaw and Wrocław.