23.03.2017 Views

wilamowski-b-m-irwin-j-d-industrial-communication-systems-2011

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

18-6 Industrial Communication Systems<br />

and exiting, thus, enabling a system to calculate the processing time of a packet in devices. This is<br />

required for switches, which do not have a predictable forwarding for packets or in ring and line<br />

structures as implemented in PROFINET. This addition was introduced when it turned out that<br />

the concept of boundary clocks of version 1 suffers from difficulties caused by cascaded control<br />

loops, which may exhibit stability problems after several hops.<br />

• In principle, the synchronization accuracy with IEEE 1588v2 can get better than 1.ns. Compared to<br />

version 1, this was made possible by the increased length of several fields in the protocol allowing for<br />

a higher resolution of parameters and timestamps. This extension reflects the increased demand for<br />

accuracy in clock synchronization. Related to it and, among others, motivated by telecom applications<br />

is the increased synchronization rate of the new version. Finally, fields to compensate asymmetry<br />

have been included, which can be used to compensate inherently, different delays for send<br />

and receive channels. However, also measures outside of the scope of the protocol can be made (e.g.,<br />

usage of calibrated cables or knowledge of the channel as it is the case in powerline).<br />

• Optional support of unicast messages. Generally speaking, IEEE 1588 assumes that one master is<br />

elected and via that master several slaves are synchronized. Of course, one would take, if available,<br />

multicast messages for synchronization, as the time to be distributed is the same for all<br />

nodes attached to this master. However, in some cases, for example, if the <strong>communication</strong> technology<br />

does not support multicast or different timescales need to be used, unicast is needed. This<br />

will be possible in version 2.<br />

• Profiles. Within the scope of standardization, profiles are in general needed to apply a special,<br />

reduced set of values and ranges to a more general standard in order to apply it for a special use case.<br />

In the case of IEEE 1588, profiles are needed in order to reflect the growing number of application<br />

areas. For example, in the new version, a special profile for telecom applications, PROFINET, etc.,<br />

can be defined. For all values not defined in a specific profile, the default values of the standard apply.<br />

• Alternative best master clock algorithms. Especially, the telecom industry demands long holdover<br />

times after a master failure. This is reflected by the possibility for alternate best master clock<br />

(BMC) algorithms.<br />

• Experimental specifications. As it can already be predicted that new features such as security<br />

(which is currently not supported at all) or cumulative frequency offsets to the master are needed,<br />

some experimental, non-normative clauses have been added. For some of those clauses, further<br />

research and case studies are needed; however, the experimental annex eases the start for such<br />

investigations [Treytl2007,Gaderer2006a].<br />

18.8 Network Time Protocol<br />

With the introduction of distributed applications in the Internet, a new requirement arose: the synchronization<br />

between the participating computers. For example, simple applications like the synchronization<br />

of files on two servers, like the well-known rsync application, use timestamps of files to distinguish<br />

between older and newer. This problem was solved with the probably oldest application layer protocol<br />

of the Internet still in use: The network time protocol (NTP). The open protocol standard NTP is a<br />

specification for a TCP/IP-based clock synchronization scheme for network devices. The main work on<br />

NTP was contributed by David L. Mills. The standardization was done via request for comment (RFC)<br />

documents—the key contribution can be found in [Mills1991a]. NTP is a very carefully designed protocol<br />

almost to every extent investigated. It covers, in opposite to IEEE 1588, all aspects like the control of<br />

the clock and respective filter algorithms [Gurewitz2003,Mills1998b,Mills1998a] with a wide range<br />

of application studies [Richards2006].<br />

The latest version of NTP is 4; nevertheless, in the meanwhile, a simplified version called SNTP has<br />

been specified and is in use.<br />

© <strong>2011</strong> by Taylor and Francis Group, LLC

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!