|
Home
>Technology > Full Story
Synchronization
for QoS in networks
By
A.K. Vanwasi
Synchronization
distribution is essential for maintaining quality-of-service
of optical and wireless networks. Here is a description of
universal coordinated time (UTC) and its distribution in digital
networks and the Internet
Network
synchronization requires precise timing standards and distribution
of timing information
Round-the-clock
information flow is the invisible glue that binds the vast
expanse of humanity today, and coordinates the functioning
of the various constituents of society and essential services
such as transport, energy, finance and so on.
The
telecommunication infrastructure connects people to people,
people to machine and machine-to-machine any time, any-where.
It comprises
a digital transmission network, switching nodes and synchronization
sub-network. The transmission network comprises a wired network,
optical fiber network, terrestrial
wireless network, and satellite network etc.
Network
synchronization is a key factor for ensuring quality-of-service
of the telecommunication network. Network synchronization requires
precise timing standards and distribution of the timing information.
The world's first practical precise timing sources were crystal
oscillators invented in the 1920s, and Atomic clocks came in
only in the 1940s. Currently, there are clocks that provide
accuracy of 10 ps/day.
Accurate
timing information is distributed globally through 'Global Positioning
System' (GPS) and communication links. GPS is a dual system
that acts as a navigation system and time-transfer system. GPS
provides stable timing information close to one ns/day.
The
task of network synchronization would have been much easier
had all the clocks and the propagation delay between two nodes
been stable. Thus, with an initial adjustment, all clocks could
run at the same rate and the network would operate without timing
faults forever. However, available devices are not so perfect
and path delays do vary. Thus, network synchronization maintains
acceptable service quality at a reasonable cost.
Synchronization
sub-network
It
deals with the distribution of time and frequency over a network
of timeservers spread over a geographical area. It comprises
of a network of timing servers. The goal is to align the time
and frequency scales of all timing servers using communication
links.
The
synchronization network must be reliable and survivable even
under unstable network conditions and where connectivity may
be lost for many days. Thus, there must be redundant timeservers
and diverse transmission paths. It must also be possible to
dynamically reconfigure the sub-network.
Full
plesiochrony
It
is a no-synchronization strategy and does not involve any synchronization
distribution. Here, each network node is equipped with an independent
clock. Thus, it is also known as synchronization anarchy.
The
full-plesiochronous strategy demands independent accurate clocks
at all nodes. This method is used in Frequency Division Multiplexing'
(FDM) and 'Plesiochronous Digital Hierarchy' (PDH) networks.
Hierarchical
master-slave synchronization
Here,
timing reference is distributed from master clock (primary time
server) to all other slave clocks (secondary time servers) of
the network
directly or indirectly. It is also known as despotism. It is
generally considered unethical but it certainly ensures very
tight control of slave clocks. Here, all clocks are synchronized
with the master (primary) clock.
To
ensure that the synchronization network does not fail, several
protection mechanisms such as dual master clocks, multiple secondary
timing clocks and routes are available. Thus, such a synchronization
network is widely used to provide excellent timing performance
and reliability at a reasonable cost.
Here,
the accuracy of each timing server is defined by a number called
the 'stratum' (level). The primary timing server is assigned
stratum 1. Successive secondary servers are assigned a level
higher than the preceding level.
Mutual
synchronization
Also
known as democratic synchronization, it is based on direct mutual
control among the clocks so that the output frequency of each
is the result of the 'suggestion' of the other peer clocks.
Such a pure democracy is appealing due to the absence of master
and slaves.
However,
maintaining mutual synchronization is a very complex and expensive
task. Normally, UTC is generated via mutual cooperative working
of several national clocks.
Global
Positioning System (GPS)
GPS
is a reliable source of time and time-transfer system.
A
versatile global tool, GPS is used to distribute time to an
arbitrarily large number of users located anywhere on the earth.
It also synchronizes clocks over long distances with a higher
degree of precision and accuracy.
GPS
system comprises of a constellation of 24 satellites launched
by the US Department of Defence. Each satellite contains atomic
clock that provides accurate and precise time within 10 ns of
UTC.
The
satellites transmit spread-spectrum signals on two frequency
bands: L1 (1575.42MHZ) and L2 (1223.6MHz). The signals are modulated
by two pseudo-random noise codes: P (Precision) code and C/A
(Coarse/Acquisition) code. For civilian applications, C/A codes
in the L1 band is used.
A
GPS receiver generates a replica of the satellite's pseudo-random
noise codes. For successful demodulation of GPS codes, exact
matching of locally generated and received code is essential.
The
received signals at GPS antenna are usually low (10 to power
of minus 15W). For successful reception of signal, clear line-of-slight
to the orbiting transmitter is essential.
Normally,
for navigation purposes, a GPS receiver is required to acquire
the signals from 4 GPS transmitters. However, for time keeping
applications, the location of the timekeeping center is known
beforehand and thus, tracking of one satellite signal is enough.
GPS
synchronizes clocks over large distances using the GPS 'Common
Views' (CV) technique. Here, two timekeeping stations simultaneously
observe the same satellite. Each of these two stations must
record the difference between his local clock time reference
and GPS time at the same instant using the same satellite and
using a GPS receiver known as GPS 'Time Transfer Unit' (TTU).
A GPS TTU is a special GPS receiver programmed to compute and
display items of interest to the timing community.
Here,
user 'A' (Ta-GPS) and user 'B' observe (Tb-GPS). Thus, by computing
the difference between the two sets of numbers we get the modulus
ta-tb. This observed time difference could be compared against
precisely known time difference between the two locations. This
helps in applying necessary time correction. Presently, using
coarse acquisition technique, it is possible to distribute time
within 10-25n sec level and time synchronization in the range
of 1-15 n/sec.
GLONASS
is the Russian 'Global Navigation Satellites System'. It comprises
of 24 satellites orbiting in three
orbits at an altitude of 19100 Km. It provides comparable time
synchronization
service.
Synchronization
in PDH systems
In
the beginning, digital transmission was confined to isolated
T1/E1 PCM links. These links were used to connect analog switching
machines. The individual PCM terminals of a link operated with
independent clocks having stability of ± 50 PPM.
To
share the transport medium, the hierarchy of plesiochronous
time division multiplexers was developed. This enabled transmission
of thousands of 64 Kbps digital channels over the shared medium.
In
PDH hierarchy, individual tributaries have plesiochronous clocks.
Thus for 8448 Kbps (E2) digital multiplexer, each individual
E1 tributary has clock variation of ± 50 PPM. To accommodate,
these clock variations, each input tributary is translated to
a common higher bitrate (2112Kb/s) and then multiplexed. This
process is known as bit-justification. Here, additional bits
are added to an individual tributary so that output bitrates
of tributaries are same.
This
bit-justification technique allows the transparent transfer
of timing contents of a digital signal across a transmission
chain. In this chain, clocks are asynchronous.
A
PDH network deploys master-slave synchronization technique for
transfer of timing in a PDH network.
Digital
Switching Operations
Digital
switching is based on moving octets (speech samples) from one
time-slot to another time-slot and from one incoming channel
to another output channel.
Time-slot
interchange is achieved by writing the frames in a memory and
then reading the time-slots in the desired order. Further, space-division
switching is used for transferring an octet from one incoming
channel to another channel.
However,
for successful switching, it is required that all the incoming
channels should be aligned/synchronized at the bit and frame
level.
The
frame and bit alignment is achieved by elastic store. Here,
incoming data bits are written at clock rate 'fw' and subsequently
read at clock rate 'fr'.
The
elastic store absorbs zero-mean frequently fluctuations between
the write and read clock within a given bound, due to buffer
limits. In practice, a frequency offset (fw-fr) between the
write and read clocks will make the buffer empty or overflow
sooner or later.
This
results in either missing or replication of an octet of data
over some time period. This phenomenon is known as slip. Its
limits are defined by ITU-T for plain old telephony (POT) applications.
Few slips over a time period were tolerable.
Contrary
to POT, the data transmission cannot tolerate slips, which results
in either loss or repetition of transported data octet. This
requirement necessitated the synchronization of digital switches.
The synchronization of two digital switching exchanges is achieved
by using master slave synchronization. Here, an exchange transmits
the master clock. Such a clock is known as Standalone Synchronization
Equipment (SASE). In ANSI terminology, it is known as 'Building
Integrated Timing Supply' (BITS)
This
master clock is recovered transparently at the remote-end and
used to synchronize the slave clock. This ensures data transmission
and reception at the same clock rate.
Synchronization
in SDH transmission
ITU-T
defined SDH standard in the year 1988. Of late, SDH is progressively
replacing PDH in backbone networks worldwide.
The
frames and signals of the SDH hierarchical levels are named
'Synchronous Transport Module-N' (STM-N) for N=1,4,16,64. Similarly,
for the American standard SONET, corresponding hierarchical
levels are named 'Synchronous Transport Signal-N' (STS-N) for
N=1,3,12,48,192.
SDH
hierarchy deploys byte multiplexing. Traditional PDH signals
are mapped into 'Virtual Containers' (VCs). A VC is individually
and independently accessible within a SDH frame through pointer
information. The pointer information is directly associated
with payloads. A pointer is located in a determined position
within the multiplexing element.
VCs
are allowed to shift independently in the output STM-N frame.
However, their position is always tracked by the respective
pointers. These pointers are incremented/ decremented according
to specified rules. Such a mechanism is known as 'pointer-justification'
and is analogous to bit-justification in PDH networks.
In
SDH system, jitter affects the output tributary severely. Besides
the jitter, pointer-justification and transmission line related
jitter results in large shifts in VCs in STM-N frame. Though,
it is possible to recover the PDH payload, but it is difficult
to have timing transferred transparently.
Synchronization
scheme of two digital switching exchanges is based on the SASE/BITS
concept and hierarchical master-slave synchronization strategy.
(see Figure-5)
The
SASE in the first digital switching exchange synchronizes both
the digital exchange and SDH multiplexer equipment using 2.048
MHz clock.
At
the receiving end, the reference signal is extracted from the
input STM-N signal and is used to synchronize the SASE clock.
The SASE signal is now used to synchronize both SDH demux and
digital switching exchange.
Synchronization
in ATMs
The
'A' in ATM stands for Asynchronous. This makes one believe that
ATM equipment can work in a system of independent and non-synchronized
clocks. In practice, the asynchronous word does not refer to
the equipment clock operation or to the physical level of information
transfer. The word asynchronous only implies that the digitized
information can be given to ATM equipment independently. Here,
it is not necessary to input digital information in predefined
time-slots as in primary PCM.
Thus,
like PDH and SDH equipment, there is provision of a synchronization
port in an ATM equipment The ATM equipment can accept external
E1/T1 reference clocks. This ensures interworking of ATM equipment
in a SDH/PDH network.
Cellular
Wireless Network
The
cellular wireless networks are structured in cells. A cell is
a limited geographical area served by a dedicated base transceiver.
A number of cells are so arranged that the seamless coverage
of a specific area is possible. Thus, a wireless network must
be capable of maintaining communication service to a mobile
user.
In
general, the smooth hand-over required in cellular network is
achieved in the following ways:
1.
In some networks, radio carrier frequencies
of all the base stations
in a region shall be within
certain accuracy limits.
2.
Other networks require that signal frames at the radio interface
(antenna) of all base transceivers should be time aligned within
certain accuracy.
Synchronization
of computers over the Internet
Normally,
every PC has an internal clock. However, this clock is inaccurate
and cannot be used for accurate timekeeping. Thus, for demanding
timing applications, it is necessary to synchronize widely distributed
PC clocks to a network timeserver.
The
'Network Time Protocol' (NTP) is an Internet standard protocol.
It is designed to distribute accurate and reliable time information
to systems operating in diverse and widely distributed environment.
NTP
is defined by RFC-1305. Current version of NTP is version-3.
NTP is built on IP and typically operates over 'User Datagram
Protocol' (UDP) but is also adaptable to other connectionless
transport protocol.
In
NTP, hierarchical master-slave strategy is deployed to distribute
reference clock. Here, one or more primary timeservers synchronize
directly to external reference source such as a GPS receiver.
Further, secondary timeservers are synchronized to primary timeservers.
NTP
servers can operate in broadcast and client-server mode.
Broadcast
mode
In
the broadcast mode, an NTP server broadcasts time to a group
of hosts on a periodic basis. The clients then determine the
time based on an assumed delay between them and the time server.
This mode is suitable for use with a large number of hosts where
modest accuracy reduced network traffic is desired.
Client
server mode
In
NTP client server mode, a client periodically polls the server
for timing information in traditional client-server fashion.
The synchronization involves the following steps:
-
The client sends an NTP packet to the server. The NTP packet
contains the timestamp at the time of leaving.
-
The server sends the response NTP packet to the client.
The NTP packet contains original timestamp of client and
timestamps corresponding to arrival and dispatch at the
server.
-
The Client takes fourth timestamp when it receives the server's
NTP
packet.
-
Now, these measurements enable a client to calculate the
round-trip delay of packet exchange and the clock offset
between the client and the NTP server.
- The
client uses the clock offset information to slew/adjust
its clock into agreement with the NTP severs. The slewing
operation typically advances or retards the clock by 0.1
per cent. In case, if the offset is too long, then the local
clock is jammed with a new time.
It
is impressive to note that the synchronization accuracy of the
order of 1 milisec is typically reachable even with the inferior
clocks found in most computers. Today, using GPS-based reference
for primary NTP server, it is possible to achieve synchronization
of the order of 1micro sec, among computers operating in several
power stations, in western USA and Canada.
Network
synchronization in India
National
synchronization network is based on the hierarchical master-slave
synchronization technique. The synchronization network supplies
reference clock to all the digital switching exchanges.
The
scheme envisages Main National Reference Clock Center (MNRC)
at Mumbai and 'Back-up National Reference Clock Center' (BNRC)
at Delhi. Each center has 3 Cesium beam clocks. However, the
output of only one clock is being used at a time. If the reference
clock at MNRC deteriorates, then it will be detected at MNRC
by continuous comparison with other clocks. The MNRC itself
can switch to a better clock.
The
BNRC works as a standby primary reference clock. Both main and
backup reference clock centers have connections via dedicated
links to each other. This enables mutual monitoring of clocks
by each other.
Level
2 reference clock centers have provision of dedicated links
from both the MNRC and BNRC. This ensures continuous availability
of the reference clock. Similarly, each lower layer clock center
has at least three dedicated links for the clock reception from
immediate higher-level clock centers.
Stability
of various levels of clocks:
Level
1: Better than 1x (10 to power of minus 11)
(for
the life of cesium beam tube)
Level
2: 1x (10 to power of minus 10)/day
Level
3: 2x (10 to power of minus 10)/day
Level
4: 1x (10 to power of minus 8)/day
Conclusion
Precise
time standard and time-transfer is at the heart of modern society.
Precise timing is essential for information flow, navigation
and position-location systems. Conti-nuous efforts are being
made to improve accuracy of clocks and devise better techniques
to distribute the time and frequency.
Coordinated
Universal Time (UTC): There
is no universal agreement for a natural phenomenon i.e. Earth
spin, Earth orbit, Moon/Sun rotation, that can be used as
unique true 'time'. Thus, ITU by international agreement has
defined 'Coordinated Universal Time' (UTC) as world's official
time. Currently, the best accuracy for the determination of
the second for UTC is equivalent to ± 1sec. in 10 million
years.
Second:
It is defined as the duration of 9192631770 periods of the radiation
corresponding to the transition between the two hyperfine levels
of the ground state of the Cesium-133 atom.
Clock:
It is a timing source. Normally, it is a two-part device. The
first part provides equally spaced periodic events or intervals
from some oscillating device. The second part counts the number
of events and displays them.
Primary
reference clock (source):
It is usually a calibrated atomic clock.
Frequency
stability:
The stability of a clock is how well it can maintain a constant
frequency. It measures the changes in frequency from one time
interval to the next. A particular time interval such as day/month/year
is chosen. It is known as averaging time. The frequency stability
is then ascertained for different values of averaging time.
Time
accuracy:
It defines how well a clock duration compares with UTC.
Precision:
It pertains to how precisely time can be resolved in a particular
time keeping system.
Offset:
The offset of two clocks is the time difference between them.
Skew:
It is the frequency difference between two clocks.
Reliability:
For a time keeping system, it is the fraction of time, it can
be kept operating and connected in the network without considerations
of stability and accuracy.
Timing
server:
A timing server must deliver continuous local time based on
UTC. It should be able to accommodate leap seconds in the UTC
time scale. The timing server must provide accurate and precise
time, despite relatively large delay variation in network path
and environmental conditions.
Primary
timing server (clock):
It is directly synchronized to a primary reference source. It
is also known as master clock.
Secondary
timing server:
A secondary timing server derives synchronization from a primary
server over network paths possibly shared with other services.
A secondary timing server can also derive synchronization from
other secondary servers.
Network:
For any network topology, the basic elements are links and nodes.
network comprises of terminals, links and nodes. (see Figure-1
A)
Terminal:
It is a source of digitized information.
Links:
The links are physical means to transport the digitized information
from a terminal to a node and between nodes. No information
processing is done in links. They simply restore the waveform
by means of regenerative repeaters.
During
transport delay, distortion and noise may cause impairments
of the signals. Further, the propagation delay over a link also
varies. The delay variation may be short-term as well as long-term.
Thus, it is necessary to properly design the synchronization
network.
The
characteristics of a link differ according to the physical means
used for signal transmission. The three main categories of transmission
media are metallic wires or cables, optical fiber and radio
transmission.
Wander:
When a signal is transported over a link, due to slow temperature
changes in medium, delay varies at a very slow rate. It is usually
below 10 Hz. Thus, wander accounts for relatively slow variation
below 10 Hz.
Jitter:
The delay variations that are faster than 10 Hz are known as
jitter. The contribution to jitter is mainly due to finite signal-to-noise
ratio obtainable at the receiving end of the clock. Jitter accumulates
over a chain of nodes.
Node:
Here, incoming digitized information is processed so that it
is forwarded correctly to another node and finally to the destination.
Information processing is based on circuit switching and packet
switching. Circuit switching establishes a continuous bi-directional
connection between subscriber terminals for the duration of
conversation or session.
Packet
switching is analogous to store and forward mode of transferring
units of information (packets). Each packet has a source and
destination address. Using the destination address, at intermediate
nodes, the packet is routed towards destination terminals.
Technical
Glossary
Spread-Spectrum
System A transmission system that transmits signals spread
over a frequency band much broader than the minimum required.
Pseudo-random
noise It is a binary code sequence. It comprises of a sequence
of 0s and 1s. The sequence has statistical properties similar
to noise.
UTC
(USNO) A GPS system transmits UTC as determined by the 'US Naval
Observatory' (USNO). The USNO is usually kept within about 10ns
of UTC.
Isochronous
digital signal It is a digital signal wherein the time interval
between significant instants have, at least on an average same
duration/s, which are integer multiples of the shortest one.
Synchronous
Signals Two digital signals are synchronous if they are isochronous
signals having the same average frequency and have controlled
phase relationship. Conversely, two digital signals are asynchronous
if they are not synchronous.
Plesiochronous
signals Two plesiochronous digital signals are isochronous,
asynchronous digital signals. Further, the respective time signals
have the same frequency nominally but actually differ within
a given tolerance range.
Justification—i)
The term justification has origin in printing industry where
it describes the process of adjusting spaces between words so
that all the lines of print are of the same length.
ii)
In digital multiplexing, bit-justification transforms plesiochronous
tributaries into synchronous tributaries so that these can be
multiplexed.
E1
PCM—It is the European standard for 'Pulse codes Modulation
(PCM). It has nominal bitrate of 2.048 Mb/s. The clock tolerance
is ±50ppm. It multiplexes 30 speech channels.
T1,
PCM—It is the American standard for PCM. It has nominal
bitrate of 1.54 Mb/s. The clock tolerance is ± 50 PPM.
It multiplexes 24 speech channels.
References
For
more information on synchronizing computers and workstations
using network and direct connect time transfer techniques'
visit http: // www.datum.com/bancomm/ettcpape.html.
Visit
the University of Delaware's Time synchronization page at
http://www.eecis. Udel.edu/~ntp/.
Visit
the following site for ITU-T G-811 compliant clocks sources
http: // www. Oscilloquartz.com
For
more information on GPS visit www.gpsworld.com http://physics/nist.govt/Genint/Time/Time.html.
http://wwwhost.cc.utexas.edu/ftp/pub/grg/gcraft/notes/gps/gps/html.
A.K.
Vanwasi is G.M. (R&D) ITI Ltd. Naini, Allahabad and can
be reached at vanvasi_nni@tiltd.co.in
|
