Optical Technologies: SONETSDH, DWDMCWDM, and RPR802.17
Optical Technologies: SONET/SDH, DWDM/CWDM, and RPR/802.17Any transmission technology that supports fiber optic cabling can be considered an optical technology. However, the leading communications equipment vendors have long used the phrase optical technologies to refer specifically to a subset of transmission technologies that operate at Layer 1 of the OSI reference model (see Chapter 2, "OSI Reference Model Versus Other Network Models") and support very long distances. SONET, SDH, DWDM, and CWDM fall into this category. Operating at OSI Layer 1 is an important characteristic because it permits a broad range of network protocols that operate at OSI Layer 2 or higher to be transported on optical technologies. Resilient packet ring (RPR) technologies combine optical transmission with LAN transport. The only standardized RPR technology is IEEE 802.17, which operates at OSI Layer 2 but is closely bound to OSI Layer 1 optical operations. SONET/SDHSONET is the transmission standard for long-haul and metropolitan carrier networks in North America. There are approximately 135,000 metropolitan area SONET rings deployed in North America today. SDH is the equivalent standard used throughout the rest of the world. Both are time division multiplexing (TDM) schemes designed to operate on fiber optic cables, and both provide highly reliable transport services over very long distances. Most storage replication solutions traverse SONET or SDH circuits, though the underlying SONET/SDH infrastructure may be hidden by a network protocol such as Point-to-Point Protocol (PPP) or Asynchronous Transfer Mode (ATM). SONET and SDH are often collectively called SONET/SDH because they are nearly identical. This overview discusses only SONET on the basis that SDH is not significantly different in the context of storage networking. SONET evolved (among other reasons) as a means of extending the North American Digital Signal Services (DS) hierarchy, which is also TDM oriented. The DS hierarchy uses one 64 kilobits per second (kbps) timeslot as the base unit. This base unit is known as digital signal 0 (DS-0). The base unit derives from the original pulse code modulation (PCM) method of encoding analog human voice signals into digital format. Twenty-four DS-0s can be multiplexed into a DS-1 (also known as T1) with some framing overhead. Likewise, DS-1s can be multiplexed into a DS-2, DS-2s into a DS-3 (also known as T3) and DS-3s into a DS-4. Additional framing overhead is incurred at each level in the hierarchy. There are several limitations to this hierarchy. One of the main problems is nonlinear framing overhead, wherein the percentage of bandwidth wasted on protocol overhead increases as bandwidth increases. This can be somewhat mitigated by concatenation techniques, but cannot be completely avoided. Other problems such as electromagnetic interference (EMI) and grounding requirements stem from the electrical nature of the transmission facilities. SONET was designed to overcome these and other DS hierarchy limitations while providing a multi-gigabit wide area data transport infrastructure. SONET can transport many data network protocols in addition to voice traffic originating from the DS hierarchy. The base unit in SONET is the synchronous transport signal level 1 (STS-1), which refers to the electrical signal generated within the SONET equipment. STS-1 operates at 51.84 Mbps. Each STS level has an associated optical carrier (OC) level, which refers to the optical waveform generated by the optical transceiver. STS-1 and OC-1 both refer to 51.84 Mbps transmission, but OC terminology is more common. OC-3 operates at three times the transmission rate of OC-1. Most SONET implementations are OC-3 or higher. This supports transport of common LAN protocols (such as Fast Ethernet) and enables aggregation of lower-speed interfaces (such as DS-3) for more efficient use of long distance fiber infrastructure. The most popular SONET transmission rates are OC-3 (155.52 Mbps), OC-12 (622.08 Mbps), OC-48 (2.49 Gbps), and OC-192 (9.95 Gbps). Many other OC signal levels are defined, but the communications industry generally produces SONET products based on OC multiples of four. The OC-3 frame structure is a modified version of the OC-1 frame. In fact, framing differences exist between all SONET transmission rates. These differences permit consistent framing overhead at all transmission rates. SONET has a framing overhead penalty of 3.45 percent regardless of the transmission rate. At low rates, this overhead seems excessive. For example, a T1 operates at 1.544 Mbps and has approximately 0.52 percent framing overhead. However, at higher rates, the efficiencies of SONET framing become clear. For example, the framing overhead of a DS-3 is 3.36 percent. Considering that the DS-3 transmission rate of 44.736 Mbps is approximately 14 percent slower than the OC-1 transmission rate, it is obvious that the DS framing technique cannot efficiently scale to gigabit speeds. Table 1-1 summarizes common DS hierarchy and SONET transmission rates and framing overhead.
For a signal to traverse great distances, the signal must be amplified periodically. For optical signals, there are two general categories of amplifiers: in-fiber optical amplifiers (IOA) and semiconductor optical amplifiers (SOA). The problem with amplification is that noise gets amplified along with the signal. Part of the beauty of digital signal processing is that a signal can be repeated instead of amplified. Noise is eliminated at each repeater, which preserves the signal-to-noise ratio (SNR) and allows the signal to traverse extended distances. However, to take advantage of signal repeating techniques, an optical signal must be converted to an electrical signal, digitally interpreted, regenerated, and then converted back to an optical signal for retransmission. An electro-optical repeater (EOR), which is a type of SOA, is required for this. SONET uses EORs to extend the distance between SONET equipment installations. A connection between two SONET equipment installations is called a line in SONET terminology and a span in common terminology. Using EORs, a SONET span can cover any distance. However, some data network protocols have timeout values that limit the practical distance of a SONET span. The connection between two EORs is called a section in SONET terminology and a link in common terminology. Each SONET link is typically 50 km or less. DWDM/CWDMWDM refers to the process of multiplexing optical signals onto a single fiber. Each optical signal is called a lambda (l). It typically falls into the 15001600 nanometer (nm) range. This range is called the WDM window. WDM allows existing networks to scale in bandwidth without requiring additional fiber pairs. This can reduce the recurring cost of operations for metropolitan- and wide-area networks significantly by deferring fiber installation costs. WDM can also enable solutions otherwise impossible to implement in situations where additional fiber installation is not possible. Wavelength and frequency are bound by the following formula:
where c stands for constant and refers to the speed of light in a vacuum; therefore wavelength cannot be changed without also changing frequency. Because of this, many people confuse WDM with frequency division multiplexing (FDM). Two factors distinguish WDM from FDM. First, FDM generally describes older multiplexing systems that process electrical signals. WDM refers to newer multiplexing systems that process optical signals. Second, each frequency multiplexed in an FDM system represents a single transmission source. By contrast, one of the primary WDM applications is the multiplexing of SONET signals, each of which may carry multiple transmissions from multiple sources via TDM. So, WDM combines TDM and FDM techniques to achieve higher bandwidth utilization. DWDM refers to closely spaced wavelengths; the closer the spacing, the higher the number of channels (bandwidth) per fiber. The International Telecommunication Union (ITU) G.694.1 standard establishes nominal wavelength spacing for DWDM systems. Spacing options are specified via a frequency grid ranging from 12.5 gigahertz (GHz), which equates to approximately 0.1 nm, to 100 GHz, which is approximately 0.8 nm. Many DWDM systems historically have supported only 100 GHz spacing (or a multiple of 100 GHz) because of technical challenges associated with closer spacing. Newer DWDM systems support spacing closer than 100 GHz. Current products typically support transmission rates of 2.5-10 Gbps, and the 40-Gbps market is expected to emerge in 2006. You can use two methods to transmit through a DWDM system. One of the methods is transparent. This means that the DWDM system will accept any client signal without special protocol mappings or frame encapsulation techniques. Using this method, a client device is connected to a transparent interface in the DWDM equipment. The DWDM devices accept the client's optical signal and shift the wavelength into the WDM window. The shifted optical signal is then multiplexed with other shifted signals onto a DWDM trunk. Some DWDM-transparent interfaces can accept a broad range of optical signals, whereas others can accept only a narrow range. Some DWDM-transparent interfaces are protocol aware, meaning that the interface understands the client protocol and can monitor the client signal. When using the transparent method, the entire end-to-end DWDM infrastructure is invisible to the client. All link-level operations are conducted end-to-end through the DWDM infrastructure. Using the second method, a client device is connected to a native interface in the DWDM equipment. For example, a Fibre Channel switch port is connected to a Fibre Channel port on a line card in a DWDM chassis. The DWDM device terminates the incoming client signal by supporting the client's protocol and actively participating as an end node on the client's network. For example, a Fibre Channel port in a DWDM device would exchange low-level Fibre Channel signals with a Fibre Channel switch and would appear as a bridge port (B_Port) to the Fibre Channel switch. This non-transparent DWDM transport service has the benefit of localizing some or all link-level operations on each side of the DWDM infrastructure. Non-transparent DWDM service also permits aggregation at the point of ingress into the DWDM network. For example, eight 1-Gbps Ethernet (GE) ports could be aggregated onto a single 10-Gbps lambda. The DWDM device must generate a new optical signal for each client signal that it terminates. The newly generated optical signals are in the WDM window and are multiplexed onto a DWDM trunk. Non-transparent DWDM service also supports monitoring of the client protocol signals. DWDM systems often employ IOAs. IOAs operate on the analog signal (that is, the optical waveform) carried within the fiber. IOAs generally operate on signals in the 15301570 nm range, which overlaps the WDM window. As the name suggests, amplification occurs within the fiber. A typical IOA is a box containing a length of special fiber that has been doped during manufacture with a rare earth element. The most common type of IOA is the erbium-doped fiber amplifier (EDFA). The normal fiber is spliced into the special fiber on each side of the EDFA. Contained within the EDFA is an optical carrier generator that operates at 980 nm or 1480 nm. This carrier is injected into the erbium-doped fiber, which excites the erbium. The erbium transfers its energy to optical signals in the 15301570 nm range as they pass through the fiber, thus amplifying signals in the center of the WDM window. IOAs can enable analog signals to travel longer distances than unamplified signals, but noise is amplified along with the signal. The noise accumulates and eventually reaches an SNR at which the signal is no longer recognizable. This limits the total distance per span that can be traversed using IOAs. Fortunately, advancements in optical fibers, lasers and filters (also known as graters) have made IOAs feasible for much longer distances than previously possible. Unfortunately, much of the world's metropolitan and long-haul fiber infrastructure was installed before these advancements were commercially viable. So, real-world DWDM spans often are shorter than the theoretical distances supported by EDFA technology. DWDM distances typically are grouped into three categories: inter-office (0300 km), long-haul (300600 km), and extended long-haul (6002000 km). Figure 1-9 shows a metropolitan area DWDM ring. Figure 1-9. Metropolitan Area DWDM Ring The operating principles of CWDM are essentially the same as DWDM, but the two are quite different from an implementation perspective. CWDM spaces wavelengths farther apart than DWDM. This characteristic leads to many factors (discussed in the following paragraphs) that lower CWDM costs by an order of magnitude. CWDM requires no special skill sets for deployment, operation, or support. Although some CWDM devices support non-transparent service, transparent CWDM devices are more common. Transparent CWDM involves the use of specialized gigabit interface converters (GBIC) or small form-factor pluggable GBICs (SFP). These are called colored GBICs and SFPs because each lambda represents a different color in the spectrum. The native GBIC or SFP in the client device is replaced with a colored GBIC or SFP. The electrical interface in the client passes signals to the colored GBIC/SFP in the usual manner. The colored GBIC/SFP converts the electrical signal to an optical wavelength in the WDM window instead of the optical wavelength natively associated with the client protocol (typically 850 nm or 1310 nm). The client device is connected to a transparent interface in the CWDM device, and the optical signal is multiplexed without being shifted. The colored GBIC/SFP negates the need to perform wavelength shifting in the CWDM device. The network administrator must plan the optical wavelength grid manually before procuring the colored GBICs/SFPs, and the colored GBICs/SFPs must be installed according to the wavelength plan to avoid conflicts in the CWDM device. To the extent that client devices are unaware of the CWDM system, and all link-level operations are conducted end-to-end, transparent CWDM service is essentially the same as transparent DWDM service. Transparent CWDM mux/demux equipment is typically passive (not powered). Passive devices cannot generate or repeat optical signals. Additionally, IOAs operate in a small wavelength range that overlaps only three CWDM signals. Some CWDM signals are unaffected by IOAs, so each CWDM span must terminate at a distance determined by the unamplified signals. Therefore, no benefit is realized by amplifying any of the CWDM signals. This means that all optical signal loss introduced by CWDM mux/demux equipment, splices, connectors, and the fiber must be subtracted from the launch power of the colored GBIC/SFP installed in the client. Thus, the client GBIC/SFP determines the theoretical maximum distance that can be traversed. Colored GBICs/SFPs typically are limited to 80 km in a point-to-point configuration, but may reach up to 120 km under ideal conditions. Signal monitoring typically is not implemented in CWDM devices. Tip IOAs may be used with CWDM if only three signals (1530 nm, 1550 nm, and 1570 nm) are multiplexed onto the fiber. Most CWDM devices operate in the 1470-1610 nm range. The ITU G.694.2 standard specifies the wavelength grid for CWDM systems. Spacing is given in nanometers, not gigahertz. The nominal spacing is 20 nm. The sparse wavelength spacing in CWDM systems enables lower product-development costs. Providing such a wide spacing grid enables relaxation of laser tolerances, which lowers laser fabrication costs. Temperature changes in a laser can change the wavelength of a signal passing through the laser. So, lasers must be cooled in DWDM systems. The grid spacing in CWDM systems allows uncooled lasers to be used because a wavelength can change moderately without being confused with a neighbor wavelength. Uncooled lasers are less costly to fabricate. Last, optical filters can be less discerning and still be effective with the wide spacing of the CWDM grid. This lowers the cost of CWDM mux/demux equipment. RPR/802.17Cisco Systems originally developed dynamic packet transport (DPT) in 1999. The spatial reuse protocol (SRP) is the basis of DPT. Cisco submitted SRP to the IETF, and it was published as an informational RFC in 2000. SRP was submitted to the IEEE for consideration as the basis of the 802.17 specification. The IEEE decided to combine components of SRP with components of a competing proposal to create 802.17. Final IEEE ratification occurred in 2004. The IETF has formed a working group to produce an RFC for IP over 802.17. Technologies such as DPT and 802.17 are commonly known as RPRs. Ethernet and SONET interworking is possible in several ways. The IEEE 802.17 standard attempts to take interworking one step further by merging key characteristics of Ethernet and SONET. Traditional service provider technologies (that is, the DS hierarchy and SONET/SDH) are TDM-based and do not provide bandwidth efficiency for data networking. Traditional LAN transport technologies are well suited to data networking, but lack some of the resiliency features required for metropolitan area transport. Traditional LAN transport technologies also tend to be suboptimal for multiservice traffic. The 802.17 standard attempts to resolve these issues by combining aspects of each technology to provide a highly resilient, data friendly, multi-service transport mechanism that mimics LAN behavior across metropolitan areas. Previous attempts to make SONET more LAN-friendly have been less successful than anticipated. For example, ATM and packet over SONET (POS) both bring data-friendly transport mechanisms to SONET by employing row-oriented synchronous payload envelopes (SPE) that take full advantage of concatenated SONET frame formats. However, each of these protocols has drawbacks. ATM is complex and has excessive overhead, which fueled the development of POS. POS is simple and has low overhead, but supports only the point-to-point topology. The 802.17 standard attempts to preserve the low overhead of POS while supporting the ring topology typically used in metropolitan-area networks (MAN) without introducing excessive complexity. The IEEE 802.17 standard partially derives from Ethernet and employs statistical multiplexing rather than TDM. The 802.17 data frame replaces the traditional Ethernet II and IEEE 802.3 data frame formats, but is essentially an Ethernet II data frame with additional header fields. The maximum size of an 802.17 data frame is 9218 bytes, which enables efficient interoperability with Ethernet II and 802.3 LANs that support jumbo frames. The 802.17 standard also defines two new frame formats for control and fairness. The 802.17 standard supports transport of TDM traffic (for example, voice and video) via a new fairness algorithm based on queues and frame prioritization. Optional forwarding of frames with bad cyclic redundancy checks (CRC) augments voice and video playback quality. The IEEE 802.17 standard employs a dual counter-rotating ring topology and supports optical link protection switching via wrapping and steering. Wrapping provides sub-50 millisecond (ms) failover if a fiber is cut, or if an RPR node fails. Unlike SONET, the 802.17 standard suffers degraded performance during failure conditions because it uses both rings during normal operation. The price paid by SONET to provide this bandwidth guarantee is that only half of the total bandwidth is used even when all fibers are intact, and all nodes are operating normally. The IEEE 802.17 standard scales to multi-gigabit speeds. Despite the tight coupling that exists between IEEE 802.17 operations at Layer 2 of the OSI reference model and optical operations at Layer 1, the 802.17 standard is independent of the Layer 1 technology. The 802.17 standard currently provides physical reconciliation sublayers for the SONET, GE, and 10-Gbps Ethernet (10GE) Physical Layer technologies. This means the maximum distance per 802.17 span is determined by the underlying technology and is not limited by Ethernet's carrier sense multiple access with collision detection (CSMA/CD) algorithm. Of course, each span may be limited to shorter distances imposed by upper-layer protocol timeout values. One of the design goals of the 802.17 specification is to support ring circumferences up to 2000 km. |
EAN: 2147483647
Pages: 196