cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
2239
Views
25
Helpful
28
Replies

NTP Help

networkricky
Level 1
Level 1

Hi All,

It would be really helpful if you could clarify, which is best practise and why??

public time server vs building own time server

Thanks

28 Replies 28

Bilal Nawaz
VIP Alumni
VIP Alumni

Hello, this is the best practice white paper on NTP

http://www.cisco.com/en/US/tech/tk869/tk769/technologies_white_paper09186a0080117070.shtml

Having your own NTP solution will be more work, to build it - design it, making it fault tolerant solution.

I've normally seen large enterprise organizations use the time source from a validated/well known NTP server on the net. I recommend using pool.ntp.org as they are a cluster of servers which are resilient at a very good stratum level or something similar.


I've seen, the core or FW's be allowed to get time from that source or something in the DMZ, and then the core or FW's can be ntp servers serving time downstream to other network devices.
Having your own - you have to do maintenance etc...maybe?
Others may have a different opinion.
Hope this helps

Sent from Cisco Technical Support iPhone App

Please rate useful posts & remember to mark any solved questions as answered. Thank you.

Leo Laohoo
Hall of Fame
Hall of Fame

Build your own.  But which one?

There are two "types" of NTP so-called "servers".

One is a plain old linux "box" and needs to go out to the internet to get NTP/SNTP synchronize.

The other is a stand-alone unit with GPS antenna.  This type syncs time with GPS and is probably THE best solution. 

networkricky
Level 1
Level 1

Thanks For your replies,

I am looking at finding the advantages/disadvantages of these 2 types of solutions i.e accuracy, delays etc

The biggest pro I get with the GPS/NTP solution is the absence of a firewall rule to go to the internet.

how about the accuracy??

how is the delay calculated? is there any exact formula ?

If you position your GPS antenna nicely, then your sync will be as accurate as any.

Hello, I found this extract from an online document. The overall document went into some mathematical detail but was a good read: http://www.eecis.udel.edu/~mills/time.html

As in the protocol specification, let T1 be the client timestamp on the request packet, T2 the server timestamp upon arrival, T3 the server timestamp on departure of the reply packet and T4 the client timestamp upon arrival. The NTP on-wire protocol calculates the clock offset
offset = [(T2 - T1) + (T3 - T4)] / 2
and roundtrip delay
delay = (T4 - T1) - (T3 - T2).
In both the offset and delay equations, the calculations require raw timestamp differences that span no more than 68 years in the future to 68 years in the past. The previous discussion in this document confirms these differences can be computed correctly regardless of whether they span between two eras, as long as the eras are adjacent.

And this:

A time difference of less than 128ms between server and client is required to maintain NTP synchronization. The typical accuracy on the Internet ranges from about 5ms to 100ms, possibly varying with network delays. A recent survey[2] suggests that 90% of the NTP servers have network delays below 100ms, and about 99% are synchronized within one second to the synchronization peer.

Reply to a question about the actual offsets achievable: "I found that 400 of the servers had offsets below 2ms, (...)"

Please see http://www.ntp.org/ntpfaq/NTP-s-algo.htm#Q-ACCURATE-CLOCK for further details.

So I guess it comes down to this, is it critical for you to get time absolutely perfect? (E.g. Stock markets, financial news etc...) or can you afford a delay/offset of up to seconds/milliseconds?
If 1st one is the case then yes your own device is extremely beneficial, if not the case, 2nd one would be good enough, so why the hassle.

No client will be able to be as accurate as the time source/server.

Hope this helps

Sent from Cisco Technical Support iPhone App

Please rate useful posts & remember to mark any solved questions as answered. Thank you.

If you choose to build your own ntp server you will have a solution that may be marginally more accurate than the one that uses a server from the Internet. And you will have the expense of provisioning your server and the management overhead of installing, monitoring, and maintaining this solution.

For some networks (mostly quite large or with very high requirements for accuracy or security) it may be worth building your own. For most of us it is a better choice to use the Internet servers.

HTH

Rick

Sent from Cisco Technical Support iPad App

HTH

Rick

riyaaz000
Level 1
Level 1

Hi ricky,

Lets refer to your original question:

Things that you should know about NTP. NTP is a protocol that uses a hierarchical levels of source clock. Meaning, there is something called STRATUM ( stratum0, stratum1. strarum2, stratum3 and so on ).

Stratum0 are devices such as atomic clock. (The drift in an atomic clocking is about 10 to the power of -9 second per day) its very accurate.

Stratum1 are computers connecting to Stratum0 (Time accuracy is acccurate)

Stratum2 are computers connecting to Stratum1 (Time accuracy is acccurate) and so on.

So most of the NTP servers that you want to use on the internet are Stratum1 or 2 or maybe 3, so they are accurate. Now for delays, that depend where you live and where the time servers are, but time update between your computer (or device) and time server are in milli seconds. It wouldn't be an issue.

Also note that when using public time servers, that doesn't mean your computer will stay tune on that server clock, but instead your computer will just ask the time server at some interval, " hey dude, whats your time" and your computer will just adjust its own clock.

Now building your own time server, for sure you wont have an atomic clock, but using GPS clock can make the deal. This does also means that you got a Stratum0 device at you hand.

Hope that helps

Regards

Riyaaz

Edwin Summers
Level 3
Level 3

Already many good replies above, but the "best" is going to depend on your specific situation.  What are your requirements for time?

Just for the purposes of simplicity, let's assume in the below text that "time server" means a device that is pulling time from the GPS system.  "Time server" can refer to many other things as noted in other posts.

Some simple things to think about:

-For most users, pulling time from a trusted Internet source is fine.  All devices in the network will be synchronized within a second (likely much better) and you spare the expense and time of setting up your own time server.  Cons are, as noted  in other posts, having to open up your firewall and access an outside server for time.  Perhaps they're not as trusted as you thought, or the Kerbals find your open port and exploit it somehow.

-Do you have a specific application/device that needs very accurate time?  Let's say millisecond or better for argument's sake.  Perhaps it needs to be synchronized at this level to a device somewhere across the country or the other side of the world.  In this case you may be better served with the expense and trouble of your own time server.

In many cases synchronizing to an outside source has been fine, in my experience, for ensuring that log files of network devices are timestamped with enough resolution to be able to directly compare events.  I have not worked with systems that require extremely accurate time for other purposes but have set up "time servers" for customers who desired them either for the "cool factor" or just to avoid having to trust an outside source for time.  They preferred to have their own source for reliability.

Best of luck! -Ed

Thank you very much for all your replies, i really appreciate it...

what is ntp poll interval? I'm having a hard time finding what its about

Ntp poll interval is how often we go and resynchronise or 'update' our time from the ntp source.

An NTP network usually gets its time from an authoritative time source, such as a radio clock or an atomic clock attached to a time server. NTP then distributes this time across the network. An NTP client makes a transaction with its server over its polling interval (from 64 to 1024 seconds) which dynamically changes over time depending on the network conditions between the NTP server and the client. The other situation occurs when the router communicates to a bad NTP server (for example, NTP server with large dispersion); the router also increases the poll interval. No more than one NTP transaction per minute is needed to synchronize two machines. It is not possible to adjust the NTP poll interval on a router.

http://www.cisco.com/en/US/tech/tk869/tk769/technologies_white_paper09186a0080117070.shtml

Hope this helps

Sent from Cisco Technical Support iPhone App

Please rate useful posts & remember to mark any solved questions as answered. Thank you.

thanks nawaz

What is the ntp packet size??

Each time a client tries to synchronize the time, how much bandwidth would be required?

Thanks for your help

Review Cisco Networking products for a $25 gift card