White Paper: New Needs for Synchronization Testing in Next Generation Networks

Transcription

1 White Paper: New Needs for Synchronization Testing in Next Generation Networks Next generation networks (NGN) combine the traditional synchronous SDH/SONET networks with packet-based (IP/Ethernet) networks to meet the demand for increased capacity at lower costs. This puts a burden on network operators and equipment manufacturers to add and validate synchronization in packet networks through the emerging Synchronous Ethernet (SyncE) and Precision Time Protocol (PTP IEEE-1588) standards. Network operators greatly benefit from synchronization testing at network interfaces to ensure quality-of-service, and equipment manufacturers benefit from a highly accurate tester/analyzer to verify performance. While packet based synchronization measurement builds confidence in the network, measurement of physical synchronization signals remains mandatory to fulfill compliance to existing and new standards. The Pendulum STA-61 is an example of field portable synchronization tester that sets a bridge between traditional network measurement and packet based measurement. Background Market drivers Mobile network operators are challenged with ever increasing demands for throughput with mobile communication devices. The demands for higher data rates in smart phones and tablets, and pressure to lower costs, are the main drivers to force backhaul networks to migrate from synchronous TDM solutions such as SDH/SONET in the fixed wireline core network, to packet-based solutions (IP/Ethernet). Even if an Ethernet network in itself does not require synchronization to transport data, the wireless access points need to be synchronized at the end nodes (base stations, BTS, Node B). The access network hence remains in charge of carrying synchronization towards base stations. LTE or 4G deployments bring new time synchronization requirements (1us typical accuracy to UTC) in addition to the previous frequency synchronization requirements (50 ppb to 16 ppb frequency accuracy). SyncE and PTP (IEEE-1588) are the two main technologies to provide packet based synchronization over these IP networks. Synchronous Ethernet (SyncE) provides frequency synchronization, through the propagation of a stable and accurate ethernet frequency carrier, all along the network, in a similar architecture to the existing SONET/SDH model. Its quality is independent of network load, but it cannot transport phase / time information. However, it requires that all network elements are consistently SyncE compliant. PTP is a master-slave time transfer protocol, based on time-stamp message exchanges. The clock recovery accuracy tends to vary upon network topology and load. This accuracy can be improved through the implementation in network elements of PTP on path support mechanisms, like Boundary Clocks or Transparent Clocks. PTP is also used for frequency synchronization, but with less predictable accuracy than SyncE. Recent studies have shown that the combination of PTP and SyncE provides better synchronization performance in the presence of network impairments. Next Generation Networks (NGN) contain a mix of traditional synchronous TDM core network and packet data based IP backhaul networks. The current trend is to use SyncE in the physical layer to assure very good frequency synchronization, and transport PTP messages to achieve time synchronization. IP networks have been in use by telecom operators for more than 5 years. And there are many predictions about very strong growth in Ethernet based backhaul networks in the coming years. As an example, see the slide below from Microsemi, presented at the WSTS conference, March In order to provide the required synchronization infrastructure, most of the network equipment vendors now propose a range of solutions that feature SyncE support, PTP support or both. However, like for any introduction of new technologies, it is essential for the telecom / network operators to gain confidence in their implementation. A strong standardization effort has been instrumental to build such confidence. But the on field validation of IP based synchronization is also a critical element, so as to make sure that these solutions meet the expectations in a variety of network conditions not only in the nominal case. 1 Test & Measurement White Paper

2 Since the break-up of earlier telco monopolies in the USA and other regions, transmission networks are sometimes managed by different operators than the backhaul networks. The synchronization is passed on from the core network to the backhaul, creating a extra demand to formally measure sync quality at the interface between networks. What happens if the synchronization is poor in traditional TDM networks? The two most commonly used properties for characterizing the synchronization clock is wander and jitter. Wander is the slow phase/frequency variations (< 10 Hz) and jitter is the fast changes (> 10 Hz). Wander is the major source of problems in a TDM network, since Jitter is easy to filter out, whereas wander propagates and accumulates in a network. It then generates transmission incidents like loss of calls, data, video frames, etc. and is therefore a source of poor communication quality and loss of revenues for operators. Wander limits are well standardized and measurement methods, based on dedicated metrics, are well known. The STA61 offers a cost-effective, multi-channel way to measure wander and check that synchronization performance meets the standards. Probes and field sync testers Measurement probes are permanently installed in the network and are used to continuously monitor the synchronization quality. These instruments normally communicate with a central server and can issue alarms if the synchronization clock is outside given limits. Probes are normally mounted in the major nodes only and do not cover the total network. They require a good external frequency reference, normally taken from GPS or from the network itself, e.g. from a local SSU. Field sync testers are portable tools mostly used for commissioning, network characterization or troubleshooting. They have their own high stability timebase. When a sync problem is detected, the field sync tester can be connected to the different network elements for quick location of the origin of the sync fault. This type of equipment provides flexibility, quick and easy operation, and is typically used by maintenance technicians. It can be used, however, also in a lab when a portable device is needed. 2 Test & Measurement White Paper

3 Testing requirements for synchronization in NGN networks Traditional synchronization testing instruments on the market today are either designed specifically for SDH/SONET or are dedicated SyncE or PTP testers, but a few recently introduced instruments, like the STA-61 Sync Tester/Analyzer from Spectracom, are designed for testing all types of synchronization clocks in Next Generation Networks (NGN). These instruments can qualify the synchronization provided by all three common technologies in NGN networks; the SDH/SONET, SyncE and PTP technologies. Traditional TDM networks based on SDH/SONET In SDH/SONET the basic synchronization is propagated along the network via E1 (2.048 Mbps) or T1 (1.544 Mbps) frames. The network reference is normally a Cesium PRC (Primary Reference Clock) that is transported via the traffic signals and re-generated in all network elements. A sync tester for SDH/SONET will extract the clock from the E1 and T1 frames and accurately measure the clock stability. Also higher order data traffic like STM-1, E3, DS3, etc. can sometimes be analyzed. Typical measurement cases: - Measure Sync Equipment Clock (SEC) performance (ITU-T G.811, G.812, G.813) - Measure sync wander at any network interface (ITU-T G.823, G.824) Packet Based synchronization networks PTP (Precision Time Protocol) IEEE-1588 With PTP, a central accurate Grandmaster clock, normally a GPS-controlled Rubidium clock, transfers its time information to PTP slave clocks, which adjust their internal clock to match the time of the Grandmaster. The PTP-slaves typically provide a physical 1- pps signal or clock (E1, T1) signal, which can be directly measured by the sync testers, according to traditional metrics (MTIE, TDEV). This allows characterization of the synchronization performance of Packet Equipment Clock, as well as the compliance of synchronization signal at any network interface. Packet Based Synchronization is also subject to network impairments like asymmetry and Packet Delay Variation (PDV). PDV is related to the variation of the propagation delay of PTP packets between the Master and the Slave. Packet delay varies with the load, the topology, the path reconfiguration phenomena, etc..; in the network, and introduces timing noise that can prevent proper clock recovery. Measuring PDV and making sure it is within standard tolerance allows operators to ensure that the network provides appropriate support to PTP. Some metrics have been recently standardized by ITU-T based on a PTP packet selection process, aiming at keeping consistency between physical synchronization metrics and packet based synchronization metrics. 3 Test & Measurement White Paper

4 Typical measurement cases: - Measure Packet Equipment Clock performance (ITU-T G.8260) - Measure wander at any network interface (ITU-T G.8261) - Measure PDV on a network portion (ITU-T G ) SyncE (Synchronous Ethernet) In SyncE networks, each network node has a local clock that is locked to the incoming data traffic, in a way that is very similar to SDH/SONET. This is the difference to traditionall Ethernet networks where all clocks in switches and routers are free-running with a accuracy of 100 ppm. A SyncE network uses a very stable clock reference, for example an E1 signal derived from an SSU in the core TDM network or from GPS-disciplined atomic clocks. The reference clock is transported with the data in the SyncE network and extracted and regenerated in all nodes. That ensures a sub ppb accuracy of the clock frequency, which is more than sufficient for the frequency accuracy demands of the radio network access nodes (50 ppb). The basic clock frequency is 125 MHz in 1GB optical SyncE networks and MHz in 10 GB networks. SyncE has been specified in a very similar way to SDH/SONET. Related standards define Ethernet Equipment Clock (EEC) performances, wander limits in the network, as well as functional model (including synchronization message exchange). Synchronous Ethernet equipment and networks must therefore be tested according to these standards, at the physical layer. Typical measurement cases: - Measure Ethernet Equipment Clock (EEC) performance (ITU-T G.8262) - Measure wander at any network interface (ITU-T G. 8262) - Check EEC synchronization messaging (ITU-T G.8264) Metrics used in synchronization testing / analysis Traditional SDH/SONET networks metrics The standard metrics for measurements of Wander in traditional SDH/SONET networks are TIE, MTIE and TDEV. The basic measurement is TIE (Time Interval Error) which expresses the phase relationship between the clock under test and the theoretically ideal clock. In the real world there are no ideal clocks, but there exists various physical implementations of very good reference clocks. The ultimate clock in a network is an expensive Cesium atomic clock, which has a very high accuracy and stability. The second best clock is an atomic clock based on Rubidium. Rubidium oscillators are commonly integrated in high performance sync testers. 4 Test & Measurement White Paper

5 MTIE expresses the Maximum TIE variation over different observation windows and is a standard metric in various telecom recommendations (ITU, ANSI, ETSI). A high MTIE means a high frequency offset, which may lead to severe synchronization problems, like buffer overflow/underflow or frequent pointer adjustments in the network nodes. Another telecom standard metric is TDEV (Time Deviation) which expresses the rms variations of the phase over various observation intervals. TDEV may reveal resonance/modulation frequencies or interfering frequencies. ITU-T standards specify the synchronization performance to be met: by different types of equipment clocks (ITU-T G.811 for PRC, G.812 for SSU, G.813 for SEC) at any network interface (ITU-T G.823, G.824) Corresponding measurements are based on MTIE and TDEV masks: as long as both MTIE and TDEV measurements are within (below) the masks, the synchronization performance is okay. In addition to the standard telecom recommendation metrics for MTIE and TDEV, the following metrics are useful to characterize frequency stability: Allan deviation (ADEV) and frequency deviation (FDEV). Metrics for next generation networks As a physical synchronization signal, propagated along the network in a quite similar way to SDH/SONET synchronization, SyncE performance measurement is done in a similar way to SDH/SONET. Synchronization measurements are therefore based on MTIE and TDEV calculated on the clock signal extracted from the Ethernet link carrying SyncE, and comparison to appropriate masks, and are: Ethernet Equipment Clock (EEC) performance (G.8262) wander at any network interface (G. 8262) In addition, conformance of synchronization message exchange (ESMC) to G.8264 must be also checked per: EEC synchronization messaging (G.8264) Synchronization measurement cases in NGN Synchronous Ethernet Measurement of SyncE extracted clock at network interfaces (1000bT, optical) against G.8261 MTIE and TDEV masks can be associated with measurement of EEC recovered clock. In addition, displaying received ESMC message allows to check its value and its conformance to G Measurement of wander generated by an EEC is done by generating a wander free clock on the Ethernet link towards the device under test, and measuring the wander of the outgoing Ethernet link. The SyncE extracted clock is then measured against G.8262 MTIE and TDEV masks. This is a key measurement for all network devices like SyncE switches, routers, gateways, that must be done in lab qualification but also in the field under real conditions. It is also important to check how the ESMC message is handled by the EEC. To this end, a useful test consists of sending a controlled sequence of ESMC messages to the devicee under test, and checking that it reacts appropriately by changing its disciplining mode and by modifying the emitted EMSC, in accordance with G Test & Measurement White Paper

6 IEEE frequency transfer A common way to measure the end-to-end IEEE1588 frequency transfer is to measure the stability of the recovered clock, behind the IEEE1588 slave, through a physical clock (E1, 1 pps, ) provided by the slave, against G.8261 MTIE, TDEV masks. IEEE1588 time transfer Though standardization on IEEE1588 is very recent, a similar approach can be used to measure the end-to-end time transfer performance, based on the measurement of the time accuracy of the recovered clock. This requires thatt the measurement device shares the same timescale than the PTP master, which is obtained thanks to GPS reception and synchronization. An absolute time comparison can be then performed between the measurement device and the device under test. IEEE1588 packet delay variation End-to-end measurement doesn t allow separating the error coming from the network Packet Delay Variation (PDV) and from the slave clock recovery performance. As PDV is strongly dependant on network conditions, it is important to measure PDV and check it is below an acceptable level. PDV is measured between a PTP grandmaster and a PTP slave precisely located in the network (transparent mode or passive probe) or between a PTP grandmaster and the measurement device whichh acts as a PTP pseudo-slave (active probe). Raw PDV can be represented as a function of time or as a packet delay distribution. Floor Packet Count and Floor 6 Test & Measurement White Paper

7 Packet Percentage measurements must be compared to a threshold (defined in G.8261), which determines the compliance in terms of packet delay variability. Packet selection can be applied in order to obtain a more stationary set of samples, which can then be analyzed using TDEV. It is also interesting to correlate the observed PDV and the recovered clock from a close (from a network perspective) slave, using the physical clock output by the slave. IEEE1588 Boundary clock (BC) measurement As boundary clocks play an important role in the PTP network by regenerating the clock and avoiding PDV propagation, it s important to qualify the BC performance not only in lab, but also on the field, in presence of real traffic. Such measurement relies on the measurement of the wander generated by the BC under test, when it s disciplined to a perfect (wander free) PTP grandmaster. The measurement device must act as a PTP grandmaster on one side and as a PTP pseudoslave on the other in order to measure the wander introduced by the BC. Such an approach also allows to measurement of the holdover performance of the BC when the PTP grandmaster is disconnected. 7 Test & Measurement White Paper

8 Additional considerations Common mode or differential wander measurements Usual measurements in TDM involve comparison of the clocks under test against a perfect timebase, to show the absolute phase variations of all signals under test, relative to the common reference clock. With packet based synchronization, it becomes interesting to track the evolution of phase offset between a master and a slave via a differential measurement mode to measure one clock against another one (even if none of them are perfect). STA61 implements both common mode and differential mode measurements. GPS for synchronization testing Wander measurement requires a very stable timebase. A rubidium atomic oscillator provides the best compromise between stability and cost. However it is still subject to a small frequency drift as time elapses (called aging, typically 5x10-11 /month) or associated with temperature changes, which makes calibration necessary (typically an annual basis). The use of a GPS receiver in the measuring instrument allows to slave the internal rubidium timebase to the GPS time, which is traceable to UTC time, and allows after an initialization phase, to compensate for the frequency drifting effects (thus eliminating the need for regular calibration). In addition, LTE/4G implementations will require an absolute time information to the base stations. Such implementation includes a GPS driven PTP grandmaster, associated with a PTP slave that recovers GPS time. Testing the slave clock implies to measure its time offset to GPS time and thus requires an embedded GPS receiver. The STA61 is offered with an embedded timing GPS receiver. Remote operation Remote operation is a very useful feature for field synchronization testers. This enables the tester to be setup and connected to a network access point by a non-skilled user, who thereafter can rely on an expert in a central location to control and start/stop the measurement, change parameters, get the result and perform post-processing analysis and report generation. Normally there are LAN ports available in the telecom network stations, so an Ethernet communication interface is very useful. The STA-61 Sync Tester / Analyzer The physical synchronization measurement is performed by a high resolution FPGA capable of time stamping the phase of 6 input channels simultaneously. The sampling rate of the phase values can be set between 100 Sa/s to 1Sa/100s. The resolution per TIE sample is 200 ps rms. The packet synchronization measurement is performed through a gigabit Ethernet clock extractor, which provides the extracted frequency to the FPGA and the value of ESMC message for display. In addition, this packet synchronization input module timestamps PTP packets and provides these timestamps to a IEEE1588 stack which in turn processes with extreme accuracy the forward delay, reverse delay, and path delay (based on the fact that the offset is null). Each module can be configured as IEEE1588 pseudo slave or as a grandmaster, based on PTP telecom profile. Forward, reverse and path delay are then processed by the STA-61 CPU unit, to display raw PDV, packet selected PDV, as well as other useful metrics like Floor Packet Count and Percentage. The unit requires at least one input module of either 2 physical sync channels or 1 packet sync (optical fiber or electrical interface). Additional modules can be added initially or later up to any combination of 3. An option provides PDV measurements for the network under test. The internal reference, which represents the theoretically exact clock, is a high-stability Rubidium atomic clock. The clock is totally wander-free in itself, and thanks to the continuous calibration via the GPS timing receiver in the STA-61G version, the small intrinsic frequency ageing over time is completely eliminated. The STA-61 is portable and lightweight and has a rugged design for field use. The size and weight allows the STA-61, including transport case, to qualify as carry-on luggage on airplanes. The communication interfaces are USB and Ethernet. A PC program WanderView TM for STA-61 takes care of remote data acquisition and analysis via USB or Ethernet. The WanderView TM software gives the user full control of the STA-61 including continuous data 8 Test & Measurement White Paper

9 streaming of measurement data, report generation and advanced post-processing and analysis. Standard metrics include TIE, MTIE, TDEV, ADEV, MADEV, FDEV, RTIE and MRTIE. By using WanderView TM, STA-61 can be used as a network probe. Screenshot examples A typical display of raw PDV : forward delay vs time, and path delay as a distribution: PDV metric : Floor Packet Percentage USA 1565 Jefferson Road, Suite 460 Rochester, NY sales@spectracomcorp.com FRANCE 3 Avenue du Canada Les Ulis, Cedex +33 (0) sales@spectracom.fr UK 6A Beechwood Chineham Park Basingstoke, Hants, RG24 8WA +44 (0) info@spectracom.co.uk April 23, WP (B) 9 Test & Measurement White Paper