Enterprise Network Time Servers

 Network Time Servers

What is a Network Time Server?
The general descriptive term 'network time server' can apply to any instance of the Network Time Protocol (NTP) server software running on a computer/server platform. The product term, 'Network Time Server' generally applies to a rack mounted network appliance (rack or tray mount), that acquires time from an external source, maintains time in its internal local clock, and supplies time to a connected network using the network time protocol NTP.

»  Back to Top

 Why should I synchronize my network?
Many operational processes (business and IT related) that occur within the network depend on accurate network timing. Further, more and more applications that support eBusiness and other enterprises are dependent on conducting their functions over the Internet. In summary:

Network operations typically require the most accuracy, on the order of 1 to 10 milliseconds. Network synchronization provides a solid infrastructure and helps avoid costly downtime problems.

Applications typically require time stamping accuracy on the order of 1 millisecond to 1 second. Sometimes it is difficult to appreciate why one millisecond can make much difference. Usually it is because it is necessary to establish the order of events.

For a complete discussion see the white papers, The Importance of Network Time Synchronization and Five Dangers of Poor Network Timekeeping.

»  Back to Top

 What's important in selecting a Network Time Server?
As in selecting any product or service there can be a long list of attractive features and Microsemi network time servers are no different. However, the list usually can be reduced to a few critical features that make a world of difference in installing and using the server, as well as some that may seem important but are not.

»  Back to Top

GPS for Time & Frequency I thought GPS was for position. Does it also supply time? Most definitely. Since the first days of longitude determination in navigation, time is the REQUIRED 4th dimension. Although GPS was originally developed for military applications, it is now the #1 source of precision time in the world. Each GPS satellite has on board multiple atomic clocks. Virtually all of our power, telecommunication, and networking infrastructure depends on accurate GPS timing for synchronization purposes. »  Back to Top  How accurate is GPS for network timing? When tracking GPS, the S600/S650 SyncServers are accurate to less than 15 nanoseconds to UTC (USNO). Time out of these units via NTP delivers less than 5 microseconds to UTC with 1-sigma variation of 15 microseconds (typical). Network factors reduce client synchronization accuracy to 0.5 to 3 ms typically. You can not get any better time synchronization in a local network than you can from a GPS synchronized local time server. »  Back to Top  What do I need to track GPS satellites? Microsemi NTP time servers are supplied with a GPS antenna, mounting kit, and various lengths of standard coaxial cable. The roof antenna (all weather) is mounted such that it has a view of most of the sky. The GPS satellite constellation consists of a minimum of 24 satellites with 12 hour inclined orbits. That means that there should always be plenty of satellites within the hemisphere to satisfactorily operate the system. Mount the antenna, string the cable, connect the cable to the back of the Network Time Server and wait a couple of minutes for it to track satellites and obtain a time solution (note: if only part of the sky is in view, it may take longer to get the time). For window mounted antenna installations it is best to use a window that faces the equator or has the best view of the sky. Generally more satellites will be in view toward the equator than away from it, east or west facing windows will also work. Polar facing windows will also work but in general are not preferred. Windows that have the best view of the sky are preferred over those that don't, regardless of orientation. Attach the antenna above the window sill versus at the top of the window. This will improve the upward visibility from the antenna to the sky. Note that some window glazing treatments may reduce or block the GPS signals preventing the time server from acquiring the time. »  Back to Top  Will GPS work inside a building? The GPS signal is a relatively low strength signal (~ -132dBm) at an extremely high frequency of 1,575 Megahertz. In order for the antenna to receive such a signal, it is necessary to have a relatively unobstructed view of the sky to view satellites within the GPS constellation (clouds are not a concern). You can track satellites with a window mounted GPS antenna. The primary concern with reception in a building window is that you effectively block off at least a 180° azimuth viewing angle for available satellites. The time server can accommodate this by deriving the time from as little as one intermittent satellite signal. However, the solar reflective coating found on some buildings today effectively reflects the GPS signal and may reduce the signal strength to a level the time server cannot track. »  Back to Top

 Network Time Protocol

What is the Network Time Protocol?
The Network Time Protocol (NTP) is a UDP protocol for IP Networks. It was designed to synchronize the clock on client machines with the clock on network time servers. But NTP is just the protocol. The implementation of NTP requires separate client and server applications. Microsemi's time servers are excellent examples of server implementation of NTP. The client application runs on workstations such as Windows or Linux. Using NTP packets, the client and server exchange time stamp data, ultimately setting the clock on the client machine very accurately to that of the time server.

»  Back to Top

 Is NTP an open source protocol?
Yes. It was developed at the University of Delaware by Dr. David Mills, under contract to DARPA. Version 1 was distributed in 1985. The current version can be found here: http://www.ntp.org/

»  Back to Top

 How does NTP work?

Superficially, NTP is a software daemon operating in a client mode, server mode, or both.

The purpose of NTP is to reveal the offset of the client's local clock relative to a time server's local clock. The client sends a time request packet (UDP) to the server which is time stamped and returned. The NTP client computes the local clock offset from the time server and makes an adjustment. But network latencies, the need to prioritize multiple servers, and the requirement to be self-healing leads to a fairly complicated, yet robust algorithm.

»  Back to Top

 Does NTP supply the time to a computer or server?
Technically no. The purpose of NTP is to reveal client clock offset; not deliver time. The operating system (OS) delivers time. Processes within the NTP application program compute the offset to adjust the client clock. Since the hardware implementation of computer clocks and the protocol to control them varies from computer to computer; it is necessary to download computer specific clock control software. If NTP is bundled with your computer OS, this has already been done. If you are downloading and installing NTP on a non-NTP aware OS; you will need to read the install instructions carefully regarding this point. This really only applies to installing NTP UNIX. The various third party NTP clients have clock control built in.

»  Back to Top

 How accurate is NTP?
It depends on how many hops occur between client and server, and other network latency inducing factors. Over Wide Area Networks, WANs, 1 to 4 milliseconds to UTC at the client is typical. Within a Local Area Network, LAN, 0.5 to 1 millisecond is typical at the client. However, when operating with the SyncServer^® models with GPS, the internal clock accuracy is less than 15 nanoseconds and the NTP time stamp accuracy at the NTP port is less than 5 microseconds. NTP client machines have been reported to be accurate in the 100's of microseconds or better realm.

»  Back to Top

 What are the main differences between SNTP and NTP clients?
SNTP is a Simple Network Time Protocol. It is based on RFC 1361/2030: it gets its time from the specified time servers of the machine on which it is installed. This protocol cannot be configured to obtain time from an alternate time server if the primary server is down. This could be called a short version of NTP client software.

NTP, Network Time Protocol, can be configured to obtain and distribute the time on the network. The algorithm can be configured to obtain time from an alternate source in case the original time server fails or gets out of synchronization.

»  Back to Top

 Where can I get NTP daemons?
Determine if your operating system has NTP built into it. If it does, then it is a simple matter of either enabling the NTP capability or taking advantage of it.

Microsemi offers the Domain Time II synchronization software for Windows systems.

»  Back to Top

 How many clients can I serve?
Microsemi's NTP time servers have demonstrated the ability to handle in excess of a million clients and still maintain microsecond level time stamp accuracy.

While you probably won't have this many clients, it is important that this level of performance be available to handle peak NTP packet request loading conditions that can occur, and still maintain desired synchronization accuracy and packet throughput.

»  Back to Top

 Do I have to run *NIX computers to use NTP?
No. NTP client software is needed on machines wishing to synchronize to the time server. Microsemi offers Domain Time II software for time client synchronization, monitoring and management. Many operating systems include an NTP client built in. Contact Microsemi for assistance.

»  Back to Top

 I am already running NTP on a Linux general purpose server. Why do I need your network appliance?
In brief, it comes down to the accuracy, reliability, security and ease of use of the dedicated network time server. If time is important to your network operations or accurate timestamps are considered an important element of your business operation, you can't afford to leave time synchronization in the hands of a generic server or a clock outside of your control on the Internet. There is a risk of invalid time and denied access to the time either deliberately by the operator of the remote server or a network disruption.

Further, these network time servers have convenient front panel time displays and controls. They have remote management tools, like SNMP and an easy-to-use HTML interface you can operate from a conventional browser, anywhere on the net. The beauty of the many monitoring aspects of the Microsemi dedicated time servers is the reliability and reporting features.

»  Back to Top

 How do I synchronize my network?
There isn't a quick answer to this. Microsemi offers many resources to assist in this process:

1. Break your organization down into Timing Dependent Entities (TDEs): departments, groups, subnets, and/or functions having discernibly unique synchronization requirements.
2. Determine the timing requirements for each TDE: legal, operational, or mandated.
3. Select the worst case scenario.
4. Select Reference Clock Sources: i.e. UTC(NIST), UTC(USNO), GPS, and synchronization methodology to satisfy the worst case scenario.
5. Establish hierarchical network timing topology including segmenting the network to take advantage of the multiple NTP ports on the SyncServers:
   
   

   Stratum 0	Reference Clock Sources (like GPS)
   Stratum 1	Primary Time Servers - the SyncServers.
   Stratum 2	Secondary Time Servers; generally application servers, NOS servers, and/or routers.
   Stratum 3	Workstations, servers, Controlled Timed Devices (CTDs).

6. Install NTP client/server application software:
7. Develop a methodology for synchronizing non-NTP synchronized CTDs.

»  Back to Top

 What is Stratum?
The term "stratum" is referenced in both the network world and the telecommunications world.

NTP uses a hierarchical structure in which Stratum 0 is the USNO clock. Stratum 1 is a radio receiver that receives the time from Stratum 0 (via GPS). Stratum 2 is a client that receives the time over a network connection from a Stratum 1 clock. Stratum 3 is a client that gets the time from Stratum 2... and so on to a theoretical Stratum 15.

In the telecommunications world, stratum refers to the holdover performance of an oscillator in the event of loss of synchronization. Stratum 1, Stratum 2, Stratum 3, and Stratum 4 are the most typical.

»  Back to Top