Technology Is Supposed to Be Simple, Right?

In his comment to the announcement of my NTP article, Joe said:

This is part of the problem with NTP. It's way more complicated then it needs to be. You shouldn't have to understand so much of it to use it on your routers. Take a look at openntpd. It's free and runs on bsd or linux.

I have to disagree with him on several counts:

  • NTP is supposed to solve a pretty hard problem of synchronizing multiple independent time sources over communication paths with unpredictable delay and jitter. Considering the limitations it's faced with, it does an amazingly good job.
  • NTP configuration on IOS is no more complex than the openntpd configuration if the only thing you want is to do is to configure an upstream NTP server. The only commands you need are ntp server and ntp master.

However, the most important point, in my opinion, is the difference between "aiming for a short recipe" and "understanding the technology". If the only task you ever need to perform is to configure upstream NTP servers, don't even bother to read the IOS documentation or my article, you don't need more than a single configuration command … but then, when things really break, you'll be in trouble.

Likewise, the only thing some people want to know about OSPF are the following two commands:

router ospf 1
 network 0.0.0.0 255.255.255.255 area 0

There are others, however, that might need a slightly more in-depth understanding of OSPF principles, operations, design, configuration and troubleshooting… and if you ignore those fundamentals you’ll have fun time when things go south.

14 comments:

  1. Comments about timing being easy make Timing industry people cry.

    Companies like Symmetricom http://www.symmttm.com/ have huge product lines that would meet any possible need for timing.

    I never fully understood the need for accurate timing until I went to a timing conference and learned just how elaborate a timing network can be, and how useful precision is in resolving network incidents (security or failure).
    Replies
    1. Another option is http://www.endruntechnologies.com/
      1U appliance can support 200,000 NTP Clients
      Varying Oscillator choices depending on your needs.
      Can use CDMA or GPS.
      Devices also Support IEEE1588 (PTP)
    2. P.S.
      I have a hard time thinking of why you would NOT want your Network devices to just be NTP Clients and use the 1U Stratum 1 Appliances. It removes any dependency on the Internet and provides significantly better service.
    3. Some people lack CapEx needed to achieve simple accuracy ... or someone figures out synchronized time is important after all the budget has been burnt.
  2. Can't agree more, Ivan.

    Speaking of Tom Thomas, do you know what is he working on now? A new book perhaps? Just curious...
  3. In simple single area OSPF with nothing but LAN links, I've always wondered what are ALL the arguments against doing:

    network 0.0.0.0 0.0.0.0 area 0

    :)
  4. I can't find any. If all you want to do is run OSPF in area 0 on all interfaces, that's the way to configure it.
  5. @William: no idea.
  6. suppose to be like this right.

    0.0.0.0 255.255.255.255 area 0
  7. Wan Tajuddin
    YM: [email protected]

    sorry forgot to leave my detail.
  8. You're right, but recent IOS releases have built-in "artificial intelligence" ... more about that in an upcoming post.
  9. I also recommend Jeff Doyle's excellent OSPF and IS-IS: Choosing an IGP for Large-Scale Networks it not only does a great job in breaking down protocol design and troubleshooting, but tackles the all controversial isis vs. ospf argument.
  10. IOS release 12.4T accepts either wildcard bits or a subnet mask. More details ...
  11. That's cool. Because from my point of view. It really make no sense of using wild card.

    Same goes the way they use access list. Why so tricky.

    Wan T
    ym: [email protected]
Add comment
Sidebar