Jumbo Frame - Adoption

Adoption

Jumboframes or 9000-byte frames have the potential to reduce overheads and CPU cycles. Recent work has also demonstrated the positive effect that Jumboframes have on end-to-end TCP performance. The presence of Jumbo frames may have an adverse effect on network latency, especially on low bandwidth links. The frame size used by an end-to-end connection is typically limited by the lowest common denomiator. 802.5 Token Ring can use 4464-byte frames, FDDI can use 4352-byte and ATM can use 9180-byte frames. More recent technologies, such as 802.11 can use 7935-byte frames. The IEEE Ethernet standard only mandates support for 1500-byte frames.

The use of 9000 bytes as preferred size for jumbo frames arose from discussions within the Joint Engineering Team of Internet2 and the U.S. federal government networks. Their recommendation has been adopted by all other national research and education networks. In order to meet this mandatory purchasing criterion, manufacturers have in turn adopted 9000 bytes as the conventional jumbo frame size.

IETF solutions for adopting Jumbo Frames avoids the data integrity reductions through use of the Castagnoli CRC polynomial being implemented within the SCTP transport (RFC 4960), and iSCSI (RFC 3720). Selection of this polynomial was based upon work documented in the paper "32-Bit Cyclic Redundancy Codes for Internet Applications". The Castagnoli polynomial 0x11EDC6F41 achieves the Hamming Distance HD=6 beyond one Ethernet MTU (to a 16,360 bit data word length) and HD=4 to 114,663 bits, which is more than 9 times the length of an Ethernet MTU. This gives two additional bits of error detection ability at MTU-sized data words compared to the Ethernet CRC standard polynomial while not sacrificing HD=4 capability for data word sizes up to and beyond 72k bits.

By using a CRC checksum rather than simple additive checksums as contained within the UDP and TCP transports, errors generated internal to NICs can be detected as well. Both TCP and UDP have proven ineffective at detecting bus specific bit errors, since these errors with simple summations tend to be self cancelling. Testing that led to adoption of RFC 3309 compiled evidence based upon simulated error injection against real data that demonstrated as much as 2% of these errors were not being detected.

One of the major impediments toward the adoption of Jumbo Frames has been the inability to upgrade existing Ethernet infrastructure that would be needed to avoid a reduction in the ability to detect errors. CRC calculations done in software have always resulted in slower performance than that achieved when using simple additive checksums, as found with TCP and UDP. To overcome the performance penalty, Intel now offers 1Gb NIC (82576) and 10Gb NIC (X520) that off-load SCTP checksum calculations and Core i7 processors support the CRC32c instruction as part of their new SSE4 vector math instruction set.

Support of Castagnoli CRC polynomial within a general purpose transport designed to handle data chunks, and within a TCP transport designed to carry SCSI data, both provide improved error detection rates despite the use of Jumbo Frames where increase of the Ethernet MTU would have otherwise resulted in a significant reduction in error detection.