October 11, 2007 -- Nearly 40 companies from across the 10 Gigabit Ethernet community collaborated to demonstrate what was billed as "the largest multi-vendor 10 Gigabit Ethernet unified data and storage fabric ever deployed" during a recent, week-long, open industry test event conducted by the University of New Hampshire InterOperability Laboratory (UNH-IOL).
The UNH-IOL's "10 Gigabit Ethernet Technology Summit," held Sept. 24 - 28, verified standards conformance and interoperability between 10 Gigabit Ethernet switches, cables, 10GBASE-T data ports and iSCSI storage devices, the primary components of a "unified fabric" using high-performance Ethernet for data and storage in critical high-end networks. The UNH-IOL sees the development of 10 Gig as a unifying alternative to the current practice of wiring such networks with a mix of Ethernet as well as closed and proprietary technologies.
"We succeeded in creating a neutral environment for proving-out the readiness of 10 Gigabit Ethernet technology for deployment in high-end networks and data centers," comments Bob Noseworthy, technical director of the UNH-IOL. "We saw companies from multiple areas of the community stepping up in the spirit of cooperation to show support for interoperability and to advance the cause of the industry. The consensus in the room was that technically, 10 Gig is ready for deployment in the data center today."
The tech summit's test plans encompassed a mix of typically deployed and emerging 10 Gigabit Ethernet technologies, including LRM, SFP+ and XFP optical interconnects, CX4 copper, backplane Ethernet ports, iSCSI targets and initiators, iWARP and RDMA services, 10 GBASE-T devices and 10 Gigabit Ethernet switching. Answering the standard's most stringent requirements, 10GBASE-T devices were shown to interoperate between four connectors at 100-meter lengths.
Specifically, according to a press release issued by the company, Solarflare was the only 10GBASE-T PHY company to successfully operate on the worst-case, 105-meter link in the lab defined in terms of insertion loss and alien cross talk.
According to the company, operational testing to the limits of the specification demonstrates that one implementation can work not only on an average or typical cabling link or channel, but also on the worst-case channel specified by the 10GBASE-T standard. Successful operation on worst-case channels ensures that the implementation can work on all of the supported channels in the field, both legacy cabling plants and new cabling installations. Solarflare successfully operated on all the cabling links present in the UNH-IOL test event, at distances ranging from 30 to 105 meters.
While the IEEE 802.3an standard sets 100 meters as an objective for operation on unshielded twisted-pair cabling, conformance to the standard is only achieved by meeting all of the protocol implementation conformance statements (PICS). Suppliers must demonstrate conformance to a comprehensive list of implementation statements including operation on worst-case channels specified in terms of cabling link or channel impairments.
The most significant channel impairments for 10GBASE-T are insertion loss, alien crosstalk and return loss. Much of the testing at the UNH-IOL event took place on links that had varying levels of insertion loss margin. The channel insertion loss, typically expressed in decibels (dB), represents the signal power at a receiver. The ratio of the insertion loss to alien crosstalk and return loss represents the majority of the signal-to-noise (SNR) at a receiver.
To determine an implementation's ability to conform to the 10GBASE-T standard, the insertion loss must be at the worst-case limit. This is often achieved by extending the test channel length, while maintaining conformance to the PICS parameters, such as delay propagation. At the test summit, Solarflare successfully interoperated on a 100-meter, four-connector channel with 4.0 dB of insertion loss margin, in other words 4.0 dB less loss than the worst-case specification.
With its PHYs at both ends of the link, Solarflare says it "took the test to the next level" by extending the link to reduce the insertion loss margin in the link. The company thereby successfully operated on the worst-case link among the entire UNH-IOL test configurations (105 meters, four connectors, with 0.1 dB of insertion loss margin, and in the presence of worst-case alien crosstalk).
Additionally, Panduit's TX6 10GIG Copper Cabling System was used as the benchmark test bed media at the test summit. Standards conformance and interoperability testing was performed according to the longest and shortest channel topologies defined by TIA in the draft 568B.2-AD-10 Category 6A standard. The Panduit system was configured in a 6-around-1 topology, creating a test environment more demanding for alien crosstalk conditions than traditional field installations.
The system includes jack modules, copper cable, patch cords and patch panels. The company says the each of the system's components has been developed with complementary design technologies that work together to suppress alien crosstalk, while providing exceptional performance beyond 500 MHz and surpassing the requirements of the Category 6A standard.
The 10 Gigabit fabric at the UNH-IOL test summit also successfully incorporated iSCSI targets and initiators running open source and commonly used commercial software, and successfully passed emulated IP voice and video data between servers and switches.
Of the 40 registered companies, the following 37 participated in the Summit: ADC, Agilent Technologies, Anue Systems, Broadcom Corp, Computer Associates (CA), Chelsio Communications, Dell Inc., Extreme Networks, Foundry Networks, Force10 Networks, Fujitsu Computer Products of America, Fulcrum Microsystems, Gigamon Systems, Hewlett Packard Corp, Intel Corp, Mellanox Technologies, Inc., Methode Electronics, Inc., Molex, Napatech, NetEffect Inc, Neterion Technologies, Network Critical, NetXen Inc, Nortel, Panduit Corp, Phyworks, ServerEngines, Shenick, Solarflare Communications, Solarwinds, Spirent Communications, Teak Technologies, Tehuti Networks, Teranetics, Tyco Electronics, Woven Systems and Xilinx.
Testing performed on the unified 10 Gigabit Ethernet fabric included the following:
Storage and advanced protocols: iWARP (RDMA over Ethernet) interoperability and stress testing, iSCSI IPv4 Interoperability (multi-target, multi-initiator, multi-vendor), iSCSI interoperability with multi-initiator to single initiator, establishing multi-vendor iSCSI raid arrays, etc. and iSCSI file system stressing.
Bridging: IPv4 OSPF interoperability and re-convergence testing.
Layer 2 switch testing: LACP (Link Aggregation Control Protocol) interoperability testing, RSTP (Rapid Spanning Tree) interoperability testing, MSTP (Multiple Spanning Tree) interoperability testing, Multi-vendor fabric end-to-end latency investigations.
10 Gigabit Ethernet link testing: 10GBASE-T link testing across worst case and non-worst case channels, 10GBASE-LRM link testing across legacy FDDI grade fiber, 10G signaling interoperability across Molex and Tyco backplanes.
Application traffic generation and network monitoring: emulating application traffic to observe the impact of L2 and L3 re-convergence on the applications, stressing the network fabric, network monitoring of the fabric and end-stations, 10 Gigabit Ethernet network capturing.
The lab pooled equipment and infrastructure from five different internal test groups to accommodate the technology's range, combining test equipment, products and engineering knowledge from its 10 Gigabit, Gigabit, Fast Ethernet, Bridge Functions, iWARP and iSCSI groups. The lab provides year-round collaborative testing of products for multi-vendor interoperability and conformance to standards in these and 14 other technologies.
More information about the UNH-IOL and the 10 Gig Tech Summit program may be found here.