This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| Jumbo frames | |
|---|---|
| Name | Jumbo frames |
| Type | Ethernet frame extension |
| Introduced | 1990s |
| Max payload | typically 9000 bytes |
| Header | 14–18 bytes (Ethernet II/802.1Q) |
| Usage | data center, storage, HPC |
Jumbo frames are Ethernet frames with payload sizes larger than the standard 1500-byte Maximum Transmission Unit (MTU). They emerged in the 1990s as vendors and researchers sought to reduce CPU overhead and increase throughput for high-bandwidth applications. Jumbo frames are implemented in switches, network interface controllers, and operating systems to optimize large data transfers in environments such as data centers, high-performance computing clusters, and storage networks.
Jumbo frames refer to Ethernet frames whose payload exceeds the IEEE 802.3 MTU of 1500 bytes, commonly configured to around 9000 bytes by vendors such as Intel, Broadcom, Cisco Systems, and Juniper Networks. Early proponents included research groups from institutions like Lawrence Berkeley National Laboratory and commercial adopters including Sun Microsystems and IBM. The term is discussed in standards and working groups within organizations such as the Institute of Electrical and Electronics Engineers and the Internet Engineering Task Force, though no single formal IEEE standard universally defines the maximum size. Industry ecosystems featuring Dell Technologies, Hewlett Packard Enterprise, and NVIDIA have produced drivers and firmware supporting larger MTUs for targeted workloads.
Typical implementations set the MTU between 1501 and 9216 bytes; common vendor defaults approximate 9000 bytes as seen in products from Cisco Systems and Juniper Networks. Ethernet frame structure fundamentals derive from IEEE 802.3, with headers described in texts by authors associated with Vinton Cerf-era networking developments and institutions like Massachusetts Institute of Technology and Stanford University. When using VLAN tagging (802.1Q), additional header bytes affect maximum frame size considerations; similar interactions occur with IEEE 802.1ad and Link Aggregation Control Protocol implementations standardized through the IEEE. Protocol stacks in operating systems from Microsoft (Windows), Apple (macOS), and distributions maintained by organizations like Red Hat and Canonical must coordinate MTU settings across layers.
Advantages promoted by vendors including Intel and Broadcom include reduced per-packet CPU interrupts, lower protocol overhead, and improved throughput for bulk transfers in environments exemplified by High Performance Computing clusters and storage arrays from NetApp and EMC Corporation. Disadvantages raised by networking practitioners at entities like Amazon Web Services and Google include increased packet loss impact on retransmission cost, potential for fragmentation at intermediate devices such as legacy switches from 3Com or routers from Cisco Systems, and complexity in mixed-vendor topologies illustrated by interoperability reports from IETF meetings. Security and debugging tools from companies like Wireshark and SolarWinds may need configuration adjustment to handle larger frames.
Network administrators at organizations such as Facebook, Microsoft Azure, and Oracle Corporation typically enable jumbo frames on end hosts and switches by configuring MTU values in network interface settings, switch ports, and virtual machine interfaces in hypervisors like VMware ESXi and KVM. Configuration steps often involve setting the MTU on NICs from Intel Corporation and Broadcom Limited and ensuring jumbo-capable firmware on switches from Arista Networks or Extreme Networks. Virtual network overlays implemented by projects and companies like Open vSwitch, Docker, and Kubernetes require consistent MTU alignment between physical underlays and virtual interfaces to prevent fragmentation and packet drops.
Benchmarks published by research groups at Oak Ridge National Laboratory and commercial labs at Intel and NVIDIA demonstrate throughput improvements and CPU utilization reductions for large-block sequential I/O and bulk TCP or UDP transfers. Microbenchmarks comparing 1500-byte MTUs with ~9000-byte frames show fewer interrupts per byte and higher wire efficiency, a pattern discussed in performance studies associated with Top500 and HPC centers such as CERN. However, real-world application-level gains vary based on workload characteristics analyzed by teams at Netflix and Spotify, where small-message workloads or latency-sensitive services may not benefit and can sometimes incur penalty due to serialization delays.
Interoperability issues arise when heterogeneous networks include legacy devices from vendors like 3Com or early-generation enterprise gear that do not support larger MTUs, leading to black-holing or fragmentation between segments. Best practices recommended by standards bodies and vendors (e.g., IEEE, Cisco Systems, Juniper Networks) stress end-to-end MTU consistency, spanning routers, switches, and virtual network layers such as those provided by VMware and OpenStack. Compatibility matrices produced by companies like Hewlett Packard Enterprise and community projects including Linux Foundation emphasize testing with tools developed by Iperf maintainers and packet analyzers from Wireshark.
Common deployments include storage networks using iSCSI and NVMe over Fabrics in data centers from providers like Dell Technologies and NetApp, HPC clusters at institutions such as Lawrence Livermore National Laboratory and Argonne National Laboratory, and cloud provider backbones operated by Amazon Web Services, Google Cloud Platform, and Microsoft Azure. Other scenarios include media production networks employed by studios associated with Industrial Light & Magic and scientific data transfer between facilities like European Organization for Nuclear Research (CERN) and national laboratories using research network backbones such as ESnet and GÉANT.
Category:Ethernet Category:Computer networking Category:Data center technologies