Maximum transmission unit

In computer networking, the maximum transmission unit (MTU) of a communications protocol of a layer is the size (in bytes or octets) of the largest protocol data unit that the layer can pass onwards. MTU parameters usually appear in association with a communications interface (NIC, serial port, etc.). Standards (Ethernet, for example) can fix the size of an MTU; or systems (such as point-to-point serial links) may decide MTU at connect time.

A larger MTU brings greater efficiency because each network packet carries more user data while protocol overheads, such as headers or underlying per-packet delays, remain fixed; the resulting higher efficiency means an improvement in bulk protocol throughput. A larger MTU also means processing of fewer packets for the same amount of data. In some systems, per-packet-processing can be a critical performance limitation.

However, this gain is not without a downside. Large packets occupy a slow link for more time than a smaller packet, causing greater delays to subsequent packets, and increasing lag and minimum latency. For example, a 1500-byte packet, the largest allowed by Ethernet at the network layer (and hence over most of the Internet), ties up a 14.4k modem for about one second.

Large packets are also problematic in the presence of communications errors. Corruption of a single bit in a packet requires that the entire packet be retransmitted. At a given bit error rate, larger packets are more likely to be corrupt. Their greater payload makes retransmissions of larger packets take longer. Despite the negative effects on retransmission duration, large packets can still have a net positive effect on end-to-end TCP performance.^[1]

Table of MTUs of common media

Note: the MTUs in this section are given as the maximum size of an IP packet that can be transmitted without fragmentation - including IP headers but excluding headers from lower levels in the protocol stack. The MTU must not be confused with the minimum datagram size that all hosts must be prepared to accept, which has a value of 576 bytes for IPv4^[2] and of 1280 bytes for IPv6.^[3] It must also not be confused with the size of the physically transmitted frame. In the case of an Ethernet frame this adds an overhead of 18 bytes, or 22 bytes with an IEEE 802.1Q tag for VLAN or quality of service.

IP (Internet protocol)

DARPA designed the Internet protocol suite to work over many different networking technologies, each of which may use packets of different size. While a host will know the MTU of its own interface and possibly that of its peers (from initial handshakes), it will not initially know the lowest MTU in a chain of links to other peers. Another potential problem is that higher-level protocols may create packets larger than a particular link supports.

To get around this issue, IPv4 allows fragmentation: dividing the datagram into pieces, each small enough to pass over the single link that is being fragmented for, using the MTU parameter configured for that interface. This fragmentation process takes place at the IP layer (OSI layer 3) and marks the packets it fragments as such, so that the IP layer of the destination host knows it should reassemble the packets into the original datagram. This method implies a number of possible drawbacks:

• All fragments of a packet must arrive for the packet to be considered received. If the network drops any fragment, the entire packet is lost.
• When the size of most or all packets exceed the MTU of a particular link that has to carry those packets, almost everything has to be fragmented. In certain cases this can cause unreasonable or unnecessary overhead. For example, various tunneling situations may exceed the MTU by very little as they add just a header's worth of data. The addition is small, but each packet now has to be sent in two fragments, the second of which carries very little payload. The same amount of payload is being moved, but every intermediate router has to do double the work in terms of header parsing and routing decisions.
• As it is normal to maximize the payload in every fragment, in general as well as when fragmenting, any further fragmentation that turns out to be necessary will increase the overhead even more.
• There is no simple method to discover the MTU of links beyond a node's direct peers.

The Internet Protocol requires that hosts must be able to process IP datagrams of at least 576 bytes (for IPv4) or 1280 bytes (for IPv6). However, this does not preclude Data Link Layers with an MTU smaller than IP's minimum MTU from conveying IP data. For example, according to IPv6's specification, if a particular Data Link Layer physically cannot deliver an IP datagram of 1280 bytes in a single frame, then the link layer must provide its own fragmentation and reassembly mechanism, separate from IP's own fragmentation mechanism, to ensure that a 1280-byte IP datagram can be delivered, intact, to the IP layer.

Path MTU Discovery

The Internet Protocol defines the "Path MTU" of an Internet transmission path as the smallest MTU of any of the IP hops of the "path" between a source and destination. Put another way, the path MTU is the largest packet size that can traverse this path without suffering fragmentation.

RFC 1191 (IPv4) and RFC 1981 (IPv6) describe "Path MTU Discovery", a technique for determining the path MTU between two IP hosts. It works by setting the DF (Don't Fragment) option in the IP headers of outgoing packets. Any device along the path whose MTU is smaller than the packet will drop such packets and send back an ICMP "Destination Unreachable (Datagram Too Big)" message containing its MTU. This information allows the source host to reduce its assumed path MTU appropriately. The process repeats until the MTU becomes small enough to traverse the entire path without fragmentation.

Unfortunately, increasing numbers of networks drop ICMP traffic (for example, to prevent denial-of-service attacks), which prevents path MTU discovery from working. One often detects such blocking in the cases where a connection works for low-volume data but hangs as soon as a host sends a large block of data. For example, with IRC a connecting client might see the initial messages up to and including the initial ping (sent by the server as an anti-spoofing measure), but get no response after that. This is because the large set of welcome messages are sent out in packets bigger than the real MTU. Also, in an IP network, the path from the source address to the destination address often gets modified dynamically, in response to various events (load-balancing, congestion, outages, etc.) - this could result in the path MTU changing (sometimes repeatedly) during a transmission, which may introduce further packet drops before the host finds a new reliable MTU.

Most Ethernet LANs use an MTU of 1500 bytes (modern LANs can use Jumbo frames, allowing for an MTU up to 9000 bytes); however, border protocols like PPPoE will reduce this. The difference between the MTU seen by end-nodes (e.g. 1500) and the Path MTU causes Path MTU Discovery to come into effect, with the possible result of making some sites behind badly configured firewalls unreachable. One can possibly work around this, depending on which part of the network one controls; for example one can change the MSS (maximum segment size) in the initial packet that sets up the TCP connection at one's firewall.

RFC 4821, Packetization Layer Path MTU Discovery, describes a Path MTU Discovery technique which responds more robustly to ICMP filtering.

MTU in other standards

The G.hn standard, developed by ITU-T, provides a high-speed (up to 1 Gigabit/s) local area network using existing home wiring (power lines, phone lines and coaxial cables). The G.hn Data Link Layer accepts data frames of up to 2¹⁴ bytes (16384 bytes). In order to avoid the problem of long data-frames taking up the medium for long periods of time, G.hn defines a procedure for segmentation that divides the data frame into smaller segments.

Disruption

The transmission of a packet on a physical network segment that is larger than the segment's MTU is known as jabber. This is almost always caused by faulty devices.^[18] Many network switches have a built-in capability to detect when a device is jabbering.^[19]

References

Further reading

• Marc Slemko (January 18, 1998). "Path MTU Discovery and Filtering ICMP". Retrieved 2007-09-02. 

External links

• Tweaking your MTU / RWin for Orange Broadband Users
• How to set the TCP MSS value using iptables
• mturoute – a console utility for debugging mtu problems
• MSS Initiative
• MTU Path   MTU discovery tool for IPv4 and IPv6 networks