This document discusses the support of varying Maximum Transmit Units (MTUs) on Catalyst 4000/4500 series switches with Supervisor III and IV.
Standard Ethernet frame MTU is 1500 bytes. This does not include the Ethernet header and Cyclic Redundancy Check (CRC) trailer, which is 18 bytes in length, to make the total Ethernet frame size of 1518. In this document, MTU size or packet size refers only to Ethernet payload. Ethernet frame size refers to the whole Ethernet frame, including the header and the trailer. Baby giant frames refer to Ethernet frame size up to 1600 bytes, and jumbo frame refers to Ethernet frame size up to 9216 bytes.
Baby and Jumbo Frame Support on Catalyst 4000 Switches
Catalyst 4000 Switch Software | Baby-Giant | Jumbo Frame | Max Frame size |
---|---|---|---|
CatOS (1) | Not Supported | Not Supported | 1522 |
IOS (2) | Yes (12.1(12c)EW) | Yes (12.1(13)EW) | 9216 |
(1) Catalyst 4000s running CatOS refers to Supervisor I and II based Catalyst 4000 switches, Catalyst 2948G, Catalyst 2980G, and Catalyst 4912G. Refer to the Baby Giant and Jumbo Support in Supervisor I and II section of this document for further details.
(2) Catalyst 4000s running IOS refers to Supervisor III or IV based Catalyst 4000/4500 switches. Please refer to following sections for understanding the feature support and caveats.
For more information on document conventions, see the Cisco Technical Tips Conventions.
There are no specific prerequisites for this document.
The information in this document is based on the software and hardware versions below.
Catalyst 4500 with Supervisor IV Engine
Cisco IOS ®12.1(13)EW
Supervisor I and II based Catalyst 4000/4500 switches, which includes the WS-C2948G, WS-C2980G, and the WS-C4912G fixed-configuration switches, do not support baby giant or jumbo frames due to an Application-Specific Integrated Circuit (ASIC) limitation.
A possible workaround is to force the switchport to accept an extra four bytes of data by configuring it as a trunk.
When a port is enabled for 802.1q trunking (Inter-Switch Link (ISL) encapsulation is not supported on Supervisor I and II based switches), the switch will automatically assume that there is an extra four bytes of data appended on, incrementing the frame size of the Layer 2 (L2) packet. Therefore, for implementations that require exactly only one tag to be carried (either 802.1q or Multiprotocol Label Switching (MPLS), but not both), it is possible to force the switchport to accept an extra four bytes of data by configuring it as a trunk port.
For example, if a port needs to carry an MPLS label, configure the port as an 802.1q trunk by changing the native VLAN to be the one desired to carry the traffic.
Baby giant refers to Ethernet frames up to 1600 bytes on the Catalyst 4000/4500 platform, or packet sizes of (MTU size) of 1552 bytes (without any header or trailer bytes). The table below lists sample protocols that can use the baby giant feature, and the configuration that is required.
Protocol/Application | Number of Header Bytes | Total Frame size | Command |
---|---|---|---|
802.1q trunking | 4 | 1500 + 4 + 18 = 1522 | No MTU commands required. |
QinQ pass-through (802.1q inside 802.1q, useful for ISPs to segregate customer traffic) | 4 + 4 | 1500 + 8 + 18 = 1526 | system mtu 1504 |
MPLS VPN pass-through (two 4-byte labels) | 4 + 4 | 1500 + 8 + 18 = 1526 | system mtu 1508 |
UTI/L2TPV3 pass-through (encapsulate one Ethernet packet in another Ethernet packet with a tunneling header. Useful to transport any payload such as IP/IPX and so on over an IP backbone). | 18+ 20+12 | 1500 + 50 + 18 = 1568 | system mtu 1550 |
Jumbo frames refer to Ethernet packets of up to 9000 bytes in size. Supervisor III and IV can handle packets up to a maximum size of 9198 bytes. This value includes the 802.1q tag or ISL VLAN tag, but does not include the Ethernet header and CRC trailer. Thus, the maximum Ethernet frame size, including the Ethernet header/trailer, is 9198 + 18 = 9216 bytes.
Note: There is a mismatch in the maximum supportable packet size between Catalyst 4000s and Catalyst 6000s. Catalyst 6000s can support packets whose packet size is as large as 9216 bytes, which means it supports a maximum total Ethernet frame size of 9216 + 18 = 9234 bytes.
Jumbo frames are used in situations where certain applications would benefit from using a large frame size (for example, Network File System (NFS)) for better throughput.
Baby giant support has been available since Cisco IOS software release 12.1(12c)EW for the Supervisor III and IV.
The baby giant feature is supported on all modules on the Catalyst 4000/4500 with the exception of the following two modules:
WS-X4418-GB module (ports 3-18 only)
WS-X4412-2GB-TX. (ports 1-12 only)
The frames are dropped if forwarded to these ports.
To enable baby giants, issue the system mtu global config command, as shown below.
4507(config)#system mtu ? <1500-1552> MTU size in bytes 4507(config)#system mtu 1552 Global Ethernet MTU is set to 1552 bytes. Note: this is the Ethernet payload size, not the total Ethernet frame size, which includes the Ethernet header/trailer
Note: The baby giant configuration applies to all interfaces that support this feature. You cannot enable the baby giant configuration per interface.
Note: If you enable jumbo frame support on a specific interface, it will override the baby giant global configuration for that interface.
The show system mtu command will display the globally configured MTU, as shown below.
Switch#show system mtu Global Ethernet MTU is 1552 bytes.
The show interfaces <interface-id> mtu command gives the configured MTU. MTU reflects the existing configured baby giant or jumbo frame, as shown below.
Switch#sh interfaces fastEthernet 4/1 mtu Port Name MTU Fa4/1 1552
The show interface <interface-id> command shows the globally configured baby giant value, as shown below.
Switch#sh int fas 4/1 FastEthernet4/1 is up, line protocol is down (notconnect) Hardware is Fast Ethernet Port, address is 0009.e845.633f (bia 0009.e845.633f) MTU 1552 bytes, BW 100000 Kbit, DLY 100 usec, reliability 255/255, txload 1/255, rxload 1/255 Encapsulation ARPA, loopback not set Keepalive set (10 sec) Auto-duplex, Auto-speed input flow-control is off, output flow-control is off ARP type: ARPA, ARP Timeout 04:00:00 Last input never, output never, output hang never Last clearing of "show interface" counters never Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 0 Queueing strategy: fifo Output queue: 0/40 (size/max) 5 minute input rate 0 bits/sec, 0 packets/sec 5 minute output rate 0 bits/sec, 0 packets/sec L3 in Switched: ucast: 0 pkt, 0 bytes - mcast: 0 pkt, 0 bytes L3 out Switched: ucast: 0 pkt, 0 bytes - mcast: 0 pkt, 0 bytes 0 packets input, 0 bytes, 0 no buffer Received 0 broadcasts (0 IP multicast) 0 runts, 0 giants, 0 throttles 0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored 0 input packets with dribble condition detected 0 packets output, 0 bytes, 0 underruns 0 output errors, 0 collisions, 0 interface resets 0 babbles, 0 late collision, 0 deferred 0 lost carrier, 0 no carrier 0 output buffer failures, 0 output buffers swapped out
The show interfaces <interface-id> counters all command provides statistics for jumbo frames, as shown below.
sup3# sh interfaces gigabitEthernet 1/1 counters all Port InBytes InUcastPkts InMcastPkts InBcastPkts Gi1/1 0 0 0 0 Port OutBytes OutUcastPkts OutMcastPkts OutBcastPkts Gi1/1 0 0 0 0 Port InPkts 64 OutPkts 64 InPkts 65-127 OutPkts 65-127 Gi1/1 0 0 0 0 Port InPkts 128-255 OutPkts 128-255 InPkts 256-511 OutPkts 256-511 Gi1/1 0 0 0 0 Port InPkts 512-1023 OutPkts 512-1023 Gi1/1 0 0 Port InPkts 1024-1518 OutPkts 1024-1518 InPkts 1519-1548 OutPkts 1519-1548 Gi1/1 0 0 0 0 Port InPkts 1549-9216 OutPkts 1549-9216 Gi1/1 0 0 <output truncated>
Baby giants are supported, however they are counted as oversized giant packets. The error counter is incremented in the output of the show interface command on the following line cards:
WS-X4504-FX-MT
WS-X4232-RJ-XX
WS-X4148-FX-MT
WS-X4148-RJ21
WS-X4148-RJ21
WS-X4232-GB-RJ (Ports 3-34)
WS-X4124-FXMT
WS-X4148-RJ
WS-X4148-RJ
WS-X4148-RJV
Jumbo frame support has been available since Cisco IOS software release 12.1(13)EW for the Supervisor III and IV.
Jumbo frames are supported only on non-blocking Gigabit ports. The following is a list of Gigabit modules and their specific ports supporting jumbo frames:
both Supervisor uplink ports on Supervisor III (WS-X4013) and Supervisor IV (WS-X4014)
WS-X4306-GB
WS-X4232-GB-RJ (ports 1-2 only )
WS-X4418-GB (ports 1-2 only)
WS-X4412-2GB-TX (ports 13-14 only)
To configure jumbo frame support, issue the mtu <mtu-size> interface configuration command, as shown below.
sup3#conf t Enter configuration commands, one per line. End with CNTL/Z. sup3(config)#interface gigabitEthernet 1/1 sup3(config-if)#mtu ? <1500-9198> MTU size in bytes sup3(config-if)#mtu 9198 sup3(config-if)#end
Jumbo frame support can be enabled on the following types of interfaces:
port-channel interface
Switched Virtual Interface (SVI)
physical interface (L2/ Layer 3 (L3))
The show interfaces <interface-id> mtu command gives the configured interface level jumbo frame configuration, as shown below.
sup3#sh interfaces gigabitEthernet 1/1 mtu Port Name MTU Gi1/1 9198
The show interface <interface-id> command provides the configured MTU for the specific interface.
Note: Jumbo frame interface configuration will supersede the global MTU configuration. In the output below, the system MTU is configured for 1552, however interface Gigabit Ethernet 1/1 is configured for jumbo frame support of 9198 bytes.
sup3#show interfaces gigabitEthernet 1/1 GigabitEthernet1/1 is up, line protocol is down (notconnect) Hardware is Gigabit Ethernet Port, address is 0004.9a80.a400 (bia 0004.9a80.a400) MTU 9198 bytes, BW 1000000 Kbit, DLY 10 usec, reliability 255/255, txload 1/255, rxload 1/255 Encapsulation ARPA, loopback not set Keepalive set (10 sec) Auto-duplex, Auto-speed input flow-control is off, output flow-control is off ARP type: ARPA, ARP Timeout 04:00:00 Last input never, output never, output hang never Last clearing of "show interface" counters never Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output drops: 0 Queueing strategy: fifo Output queue: 0/40 (size/max) 5 minute input rate 0 bits/sec, 0 packets/sec 5 minute output rate 0 bits/sec, 0 packets/sec 0 packets input, 0 bytes, 0 no buffer Received 0 broadcasts (0 multicast) 0 runts, 0 giants, 0 throttles 0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored 0 input packets with dribble condition detected 0 packets output, 0 bytes, 0 underruns 0 output errors, 0 collisions, 0 interface resets 0 babbles, 0 late collision, 0 deferred 0 lost carrier, 0 no carrier 0 output buffer failures, 0 output buffers swapped out
The show interfaces <interface-id> counters all command provides statistics for jumbo frames, as shown below.
sup3# sh interfaces gigabitEthernet 1/1 counters all Port InBytes InUcastPkts InMcastPkts InBcastPkts Gi1/1 0 0 0 0 Port OutBytes OutUcastPkts OutMcastPkts OutBcastPkts Gi1/1 0 0 0 0 Port InPkts 64 OutPkts 64 InPkts 65-127 OutPkts 65-127 Gi1/1 0 0 0 0 Port InPkts 128-255 OutPkts 128-255 InPkts 256-511 OutPkts 256-511 Gi1/1 0 0 0 0 Port InPkts 512-1023 OutPkts 512-1023 Gi1/1 0 0 Port InPkts 1024-1518 OutPkts 1024-1518 InPkts 1519-1548 OutPkts 1519-1548 Gi1/1 0 0 0 0 Port InPkts 1549-9216 OutPkts 1549-9216 Gi1/1 0 0 <output truncated>
The show system mtu command shows the configured baby giant value, if configured. Jumbo frame support is per-interface, as shown below.
sup3# sh system mtu Global Ethernet MTU is 1552 bytes.
Make sure all interfaces in a VLAN are configured for jumbo frames before configuring jumbo frame support on an SVI. A packet’s MTU is not checked on the ingress side of an SVI. However, it is checked on the egress side of an SVI. If the packet’s MTU is larger than the egress SVI’s MTU, the packet is fragmented by software (if the DF bit is not set), which results in poor performance. Software fragmentation only happens for L3 switching. When a packet is forwarded to an L3 port or an SVI with a smaller MTU, software fragmentation will occur.
In the output below, you can see that by issuing the show vlan mtu command, a mismatch has occurred for VLAN 1. The port Gig 4/1 in VLAN 1 can support only 1500 bytes, and therefore cannot fully support jumbo frames for that VLAN. Packets destined to these such ports, which do not support jumbo MTU, may get dropped for L2 switching. The packet continues to be forwarded if it is destined to Gig 1/1 or any non-blocking port in that VLAN.
It is recommended that the MTU of an SVI should always be smaller than the smallest MTU among all the switch ports in the VLAN. However, this is not enforced in software.
sup3# sh vlan mtu VLAN SVI_MTU MinMTU(port) MaxMTU(port) MTU_Mismatch ---- ------------- ------------- ------------ ------------ 1 9198(TooBig) 1500 (Gi4/1 ) 9198 (Gi1/1 ) Yes 2 1552 1552 1552 No 17 1552 1552 1552 No
Jumbo frames can be enabled on interfaces configured for port-channel protocols. The following are some of the guidelines or restrictions:
All the ports in a port-channel must have the same MTU.
Changing the MTU of a port-channel interface changes the MTU of all member ports.
If the MTU of a member port cannot be changed to the new value due to the member port being the blocking port, the port-channel is suspended .
A port cannot join a port-channel if it has a different MTU than the others in the existing port-channel.
If an individual member port's MTU is changed, the port is suspended.
Revision | Publish Date | Comments |
---|---|---|
1.0 |
24-Mar-2005 |
Initial Release |