Field-testing issues with fiber-based Gigabit Ethernet systems
With more test-limit sets on the horizon, a field-tester's versatility becomes paramount.
Fanny Mlinarsky / Agilent Technologies
With more test-limit sets on the horizon, a field-tester's versatility becomes paramount.
Today, fiber-optic installations are quickly growing in number and bandwidth to alleviate the throughput bottlenecks on backbone networks where traffic from multiple workstations aggregates. These newest developments in high-speed Ethernet transmission over fiber-optic media have forced the industry as a whole to take a closer look at field-testing issues associated with emerging standards.
The two-year-old Gigabit Ethernet standard from the Institute of Electrical and Electronics Engineers (IEEE-New York City) for local-area-network (LAN) backbone signaling no longer sufficiently satisfies the exploding demand for bandwidth. The number of LAN users is increasing so quickly that backbones are once again ready for a tenfold boost in throughput, and as a result, last March, the IEEE officially approved the development of a 10-Gigabit Ethernet standard for backbone communications. The IEEE 802.3ae working group is responsible for the 10-Gigabit Ethernet specification.
Ethernet data rates have traditionally been a power of 10, increasing exponentially from one generation of products to the next. The 10-Gbit/sec Ethernet standard is the latest and fastest generation of the Ethernet family. With the release of this new standard, LAN backbone data rates will increase from 1 Gbit/sec to 10 Gbits/sec.
In a typical LAN, backbone connections will migrate from 1-Gbit/sec speed to 10 Gbits/sec. It is not clear whether work-area connections must be upgraded above 100 Mbits/sec at this time.
The bandwidth shortage is more acute in the wide-area-network (WAN) environment than in LAN backbones, as more and more World Wide Web traffic flows outside the LAN and over the Internet. For this reason, the IEEE is planning to develop a 10-Gbit/sec Ethernet physical layer (PHY) specification that can operate in the LAN or WAN mode. In fact, the IEEE 802.3ae working group has already made some progress toward the standard.
The emerging standard
The working group is in the process of narrowing down the numerous proposals for 10-Gbit/sec signaling schemes. Among the committee's official objectives is to provide physical layer specifications that support link distances of:
- At least 100 meters over installed multimode fiber.
- At least 300 meters over new, high-bandwidth multimode fiber.
- At minimum of 2, 10, or 40 km over singlemode fiber, depending on signaling scheme.
Other official objectives include supporting fiber media selected from the second edition of the ISO/IEC-11801 standard and defining two families of PHYs-a LAN PHY operation at a data rate of 10 Gbits/sec and a WAN PHY operating at the OC-192 data rate of 9.95328 Gbits/sec.
The target cost of a 10-Gbit/sec Ethernet PHY is three times that of a 1-Gbit/sec 1000Base-X PHY, with the tenfold increase in data rate. Expected completion date of the standard is March 2002.
To meet the distance objectives, the group may standardize several PHY variations; some of these PHYs may be optimized for multimode fiber operation and some for long-distance singlemode transmission. Three of the group's proposals appear to be the most promising.
A 1,310-nm wide-wavelength-division-multiplexing (WWDM) scheme supports 300 meters over installed 62.5- and 50-micron fiber and requires the use of an offset patch cord as does 1000Base-LX. It supports at least 10 km over singlemode fiber.
An 850-nm vertical-cavity surface-emitting laser (VCSEL) scheme supports 300 meters over new, 2200-MHz-km 50-micron fiber, but less than 100 meters over installed 62.5-micron fiber. It does not support singlemode fiber.
A 1,310-nm distributed-feedback (DFB) laser scheme supports 40 km in its "cooled" version, while its "uncooled" version supports 10 km. This scheme is a candidate for supporting dual-data-rate communications-10 Gbits/sec for LAN environments and OC-192 rates in the WAN.
Organizers expect that at least one of these PHY schemes will support both LAN and WAN data rates. This LAN/WAN PHY will have the ability to operate at exactly 10 Mbits/sec on an Ethernet LAN or at a multiple of the 51.84-Mbit/sec optical carrier-1 (OC-1) data rate. For example, the OC-192 data rate is exactly 192 times the OC-1 data rate. The most likely candidate for the dual-data-rate PHY is the 1,310-nm DFB laser proposal.
IEEE 802.3ae is also considering a proposal for an 850-nm VCSEL requiring the use of the new high-bandwidth 50-micron fiber. This new 2,200-MHz-km fiber is optimized for 850-nm operation but offers the same 500-MHz-km performance at 1,310 nm as the installed base of 62.5- and 50-micron multimode fiber. The specification for the 50-micron fiber likely will be incorporated into the draft ISO-11801 second edition-the cabling specification that 802.3ae will reference.
The most broadly applicable proposal is the 1,310-nm WWDM proposal. This scheme is similar to the 1000Base-LX flavor of the currently deployed Gigabit Ethernet, in that it supports both multimode and singlemode fiber and requires an offset patch cord for multimode operation. The WWDM PHY supports at least 300 meters over all types of multimode fiber and supports 10 km over singlemode fiber.
The key factors influencing the decision on the signaling schemes are the distance that the selected schemes would support over different fiber-optic cable types and the complexity of implementation. The 850-nm VCSEL proposal is best-suited for localized server-farm-type environments where the new 50-micron fiber can be installed. One disadvantage of this scheme is that it supports less than 100 meters of the currently installed multimode fiber and does not support singlemode fiber at all. The best candidate for general-purpose LAN backbone deployment may be the 1,310-nm WWDM proposal that supports both multimode and singlemode fiber with distances of 300 meters and 10 km, respectively.
While it may be too early to speculate on what PHY proposals will be selected for the final IEEE 802.3ae standard, it is clear that several schemes, optimized for different types of environments, may be standardized. All of this translates into increased complexity when it comes to field-testing.
Existing and emerging testing standards
The IEEE typically references ISO and TIA cabling standards for field-testing requirements. However, while these specifications cover field-measurement methodology, the IEEE still specifies the loss and length limits for each application.
The TIA/EIA-568B.3 and ISO-11801 specifications include generic loss limits based on wavelength and fiber type.
Those who install and verify fiber-optic systems would do well to make use of the fiber-optic network test limits programmed into some field testers.
A field tester can evaluate the measured fiber losses against the generic limits, provided the test technician specifies the length of fiber and the number of connectors or splices. However, testing to these generic limits does not guarantee the applications would work. It is important to select a field tester that can automatically produce pass/fail limits for different networks.
Already, the IEEE has specified seven different sets of length and loss limits for the existing variants of Gigabit Ethernet. The IEEE 802.3ae 10-Gbit/sec Ethernet standard likely will require at least as many different sets of limits, which will add considerably to field-testing complexity. With so many different limits, it becomes virtually impossible to certify fiber-optic installations with old-fashioned loss meters and still guarantee that all the backbone technologies will work over a given installation.
The loss and length limits for different networks are a function of cable type and the transceivers' operating wavelength. Because of the vast number of different applications, and in many cases several different sets of limits for each application, the field tester should automatically keep track of the application test limits.
Also, the test report should document the pass/fail result for each network and with respect to generic TIA or ISO limits.
The emerging IEEE 802.3ae 10-Gbit/sec Ethernet standard will add considerable complexity to field-testing procedures. This specification is expected to support several different transceivers operating over five different fiber types. Today, we already have seven different sets of test limits for the IEEE 802.3z version of Gigabit Ethernet. The new 10-Gbit Ethernet standard is expected to at least double this number of field-test limits.
Given the increasing complexity of field-testing, it is more important than ever to use a tester that has the ability to automate pass/fail determination for different networks.
Fanny Mlinarsky is research and development manager with Agilent Technologies (Marlboro, MA). She manages the Agilent engineering group responsible for the development of handheld test tools for LAN installation and certification and has participated in the development of networking and cabling standards within the IEEE, ANSI, ISO/IEC, and TIA.
Information from IEEE 802.3, 1998 edition; the ISO-11801 2nd edition, draft document N 568; and the TIA/EIA-568B.3 draft SP-3894 was used in this article.