NTP Pool & Network Load (Was: zeroconf for ntpd?)

NTP Pool & Network Load (Was: zeroconf for ntpd?)

Post by jaso » Mon, 28 May 2007 05:53:28

I think that any home user that is running *nix machines with NTP would know
better than to have them all sync to pool servers. I'm sure a good majority
have their own little hierarchy. As for windows machines, with most people
just running the default windows SNTP client... I don't want to get into

One thing I think would be nice is if home broadband routers started to
implement NTP (or even just SNTP) server abilities (for the LAN side only by
default). That way local LAN clients could just sync to the router.

Anyhow, I had to pull my server (this is the 2nd time) from the NTP pool
because of load/network issues that only cropped up once I joined the pool,
and since I removed myself and configured NTP not to respond to network
requests it has gone away.

It wasn't the usual traffic that was causing problems, but a handful of
abusive clients. I monitor and log everything, NTP connections were no
exception. Even after I started to block some IPs via my firewall I would
still get pounded with thousands of requests daily from them. Even though my
server is at a hosting facility with a 100Mb connection, with only the
minimum speed set in the NTP Pool (256k?) I was having those problems! I'm
sure I am not the only one that has experienced such issues, and for those
using a DSL/Cable connection with much less bandwidth the problem has to be
more intensified.

I don't know if it was intentional abuse or just some poorly written clients
(perhaps some routers or other appliances, I don't know). Either way it is
not good for the pool project, and as the pool grows (hopefully) the issue
will probably only worsen if people continue to ignore it.


questions mailing list

1. [pool.ntp.org] {0,1,2}.<zone>.pool.ntp.org; new monitoring system

2. [crivera@visual-tools.com: [ntp:questions] ntpd in isolated network]

Thanks for the answer. However I have important restrictions
on the hardware to be used, I can not simply add a new machine.
So what I have tried is to fiddle with the source code a little
bit and I have been able to get startup times of the order of
15 to 20 seconds.
I have changed the following in ntp.h (I am working with version 4.1.1c-rc1):

* NTP protocol parameters. See section 3.2.6 of the specification.
#define NTP_MINDPOLL 1 /* log2 default min poll interval (64 s) *
#define NTP_MINPOLL 1 /* log2 min poll interval (16 s) */
#define NTP_MINCLOCK 1 /* minimum survivors */
#define NTP_MAXCLOCK 1 /* maximum candidates */
#define BURST_INTERVAL1 1 /* first interburst interval (log2) */
#define BURST_INTERVAL2 1 /* succeeding interburst intervals (log2)

I asked to the list in case it was possible to speed up startup time without
changing the code, but now I am certain that that's not possible.


Free the mallocs!

Carlos Rivera Cordero
R&D Department / Departamento de I+D
C/Isla Graciosa, 1.
28034 Madrid - Spain
Telephone: +34 91 72948 44
Fax: +34 91 358 52 36

3. [ntp:questions] ntpd in isolated network

4. Why so many dead NTP servers in pool.ntp.org?

5. Bad NTP servers jeopardizing the pool.ntp.org initiative

6. [ntp:questions] [pool.ntp.org] 87 servers so far/webserver configuration / some other small things

7. NTP does not sync when using pool.ntp.org

8. [ntp:questions] [pool.ntp.org] 87 servers so far /web server configuration / some other small things

9. [Pool] NTP Pool Servers and DynDNS

10. [ntpwg] sntp: rejects packets from ntpd when "disable ntp" is used

11. [ntp:questions] Unexpected ntpd behavior

12. [ntp:questions] using ntpd to initialise time instead of ntpdate sets wrong time.

13. [ntp:questions] Strange ntpd problem

14. Documentation comments (Was: Finding out where ntpd gets its ntp.conf file)

15. Finding out where ntpd gets its ntp.conf file