Slashdot Mirror
The NTP Pool Needs More Servers — Yours, If Available
Do you have a static IP or two? If so, you might be able to spread some Internet infrastructure well-being with very little effort. An anonymous reader writes "The NTP Pool project is turning 10 soon, and needs more servers to continue serving reasonably accurate time to anyone in the world."
//puts on sunglasses//
Are we talking about about stratum 1 servers here?
Give me Classic Slashdot or give me death!
"The NTP pool is a dynamic collection of networked computers that volunteer to provide highly accurate time via the Network Time Protocol to clients worldwide." "Network Time Protocol (NTP) is a networking protocol for synchronizing the clocks of computer systems over packet-switched, variable-latency data networks. In operation since before 1985, NTP is one of the oldest Internet protocols in use." - wikipedia.
My AC stalker: " I personally agree with your posts most of the time, but that won't keep me from modding you troll"
This seems like something that almost every country and government in the world, could thrown down a couple hundred dollars a year for. 3rd world, and war-torn countries need not apply for obvious reasons....
In the US, is NIST involved in this at all? If not, why not? Just seems like something that they'd be all over.
Anyone considering this should carefully read the NTP pool's page on the matter. In addition to having a static IP, you need to have fairly good availability over a long period of time, and more importantly you need to be able to handle a lot of traffic. Even though the traffic is fairly low most of the time, you could experience spikes that would be difficult to handle for small businesses or amateurs. Also, anyone with metered bandwidth on their server/colo would almost certainly be unable to handle the cost.
The NTP pool is something that you have to consider carefully. You can't help out for 18 months and then decide to quit. You can expect to receive traffic for up to YEARS after you leave the pool.
-d
"Here Lies Philip J. Fry, named for his uncle, to carry on his spirit"
Can Google/Apple/Amazon not just throw some money at this?
It is easy and they do provide documentation. I added my server and it took about 10 minutes. Stop being a lazy shit.
Some quick searching shows one can get a USB GPS receiver for $27 and the comments say it works with linux/gpsd, showing up as /dev/ttyUSB0.
Somebody could make a simple OS image that would narrow the scope of the problem to the availability of ~$60 and an available public IP address.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
I've always wondered about the defaults to have every RH/Debian/Suse/Ubuntu/etc. box talk directly to the pool. I know that for years, the pool has been considered fully sufficient to meet these needs, but it just always struck me as more efficient for an organization to run its own NTP server--one machine talking to the pool--and have other machines in the organization talk to that, rather than having all the machines in the organization talk to the pool.
For home use, I actually use ntpupdate in a once-a-day cron job, rather than having a full ntpd talking to the pool all day long. It was a little more work to set up (which is also something I wish could be addressed), but combined with automatic drift correction, it seems more than adequate for my needs.
Not that I want to discourage people from contributing to the pool! That's a great idea. I just think it might also be beneficial if people learned to be less abusive of the pool, and if distro makers made it easier to not abuse the pool.
needs the Model B, of course.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Perhaps you could also point out a source for a Raspberry Pi.
I once took an excursion to Reddit, and later HN. Unlimited up/down voting sucks when dealing with a hive-mind.
They can use my system if they don't mind pretty crappy latency.
Have gnu, will travel.
These three are the US master clock's stratum-1 servers. They most likely will not run out of bandwidth. The last one isn't (intended) for civilian users, so don't come to me if an aircraft carrier, F/A-18 Hornet, etc. smashes through your front door.
tick.usno.navy.mil
tock.usno.navy.mil
ntp.usno.navy.mil
More information.
Thank you, Edward Snowden.
"Arguments from authority are worthless." —Carl Sagan
Without metrics, this is just "Please sir, may I have some more?"
How about telling us how many servers are there, what their utilization is, client load, etc?
I want to delete my account but Slashdot doesn't allow it.
"no to pizza"
Why would you make up an acronym for a concept that doesn't exist for words that cannot be spoken?
Boot Windows, Linux, and ESX over the network for free.
They can be read up on here:
http://tns.its.psu.edu/networking/timeReference.cfm
It would make more sense for ISPs to be providing NTP service, since the shortest routes have to go through their peering points or other gateways anyway. Has the NTP Pool been trying to bring them in?
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Because their load balancing shouldn't have to resolve your IP every time they send someone to your server. Tends to fuck up the accuracy of the, you know, time... Also, your connection is not reliable enough based on your comment. This is not folding at home, SETI at home, etc. They don't want people like you fucking things up.
The NTP protocol doesn't support changing IPs -- there's a long-term relationship among hosts in an NTP group. Servers like yours that hop on and off the network are only useful for single-sync applications and therefore are not suitable for inclusion in an NTP pool.
Since all broadband connections have bufferbloat (to some degree or other), in all technologies (fiber, DSL and cable alike), it isn't a good idea to volunteer to run an NTP server on such a connection, even if it is/has been reliable. Bufferbloat will induce transient bad timing into your time service; even more fun, in often a asymmetric way, pretty much any time you do anything over that link.
- Jim
Many ISPs will give you a static IP for a reasonable monthly charge.
Some do it by default, like mine.
I used to have a computer in the pool, but removed it due to disgust with the NTP abusers out there. When I looked at the logs, I would see that the vast majority of incoming traffic was from a relatively small handful of IP address. For normal well behaved users, you would see them hit you every 64 seconds and over a period of a few hours slowly back off until they do a query only once every 1024 seconds. Reasonable and well behaved. Even a relatively low bandwidth DSL line could handle a lot of users like that.
Unfortunately, not all the users are reasonable and well behaved. There were a few addresses that were hitting me with a query per second. And you can't blacklist these anti-social idiots because if you do, they're still consuming inbound bandwidth. After a period of time where 1% of the users were consuming 99% of my donated resources, I left the pool out of disgust. Was still getting hits from the idiot users a year later.
To make their idiocy even more evident, the SHORTEST interval that NTPD will hit a server is once per 16 seconds. So those once a second idiots were using software that itself was written by idiots.
Would I donate to the pool again? Nope. Not at long as there are invalid NTP clients that hit that often. If I could be assured that the idiots are gone, then I'd donate. Until then, I don't need the headaches.
An USB GPS means no Pulse Per Second (actually 1000ms). The PPS fires an interrupt on the serial port, which should result in an interrupt every 1000ms accurate within 100us.
The lack of PPS will result in a ntpd with lots of jitter, my experience is about +/- 150ms but this depends heavily on actual USB usage and the GPS device itself. This is unsuitable for a low stratum ntpserver IMHO, so don't use it as the only timesource if you want to participate in the pool unless you advertise it as some high stratum source (I would guess 5-10).
So how do I get the GPS receiver to get a time signal in my basement or datacenter?
NTP requires long-term relationships among the hosts in the peer/server group. As implemented that means static IPs, but even if you changed the system to do repeated DNS lookups the NTP pool couldn't use hostnames -- the DNS-based pooling currently in use does not include any mechanism to distribute hostnames, nor do most NTP clients provide any method to easily consume such data even if it were available.
Serial. USB has variable latency.
I use this receiver, which is quite reasonably priced. The wiring diagram at this site makes it quite easy to assemble.
Rather than driving the PPS LED directly from the PPS line, I used an NPN transistor to switch the LED on and off with each pulse. The transistor draws a negligible current from the PPS line.
I got the whole setup wired in less than an hour. Works quite well.
As I understand it, an NTP server closer to you on the Internet will provide more accurate time. Fewer hops away generally means a shorter ping and less jitter. Adding more servers in underserved countries adds more servers closer to users in those countries.
Real Americans pay for the time and don't rely on handouts.
Serial. USB has variable latency.
What's the cause of the variability of the USB latency? Does it apply on a dedicated bus?
This testing makes it look fairly stable.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
You don't run an NTP server in a VM. NTP servers need realtime (or as close as possible on a non-RTOS) access to the clock and network, and no matter much you jack up the priority of your NTP server VM, it's not going to be stable enough for anyone to bother using it. This is why e.g. VMWare ESX run an internal ntp daemon that the VMs can sync to, which itself syncs to the ntp pool.
Yes, if a USB receiver makes it accurate enough for the monitoring system then it's fine. (Though the monitoring system has been tuned to be stricter and stricter over the years).
However: if the USB receiver has more "jitter" than the other internet servers you'd be syncing from as backup, then there's not much point in having it.
--
ask bjoern hansen
A long cable. :-) Depending on the equipment you can have cables several hundred feet long.
An (expensive) way is to use a CDMA receiver. The CDMA protocol needs accurate timing, so it's included in those signals. It's not as accurate as GPS, but it can work in places with no "sky view" access.
--
ask bjoern hansen
Maybe I'll do that in the future if we setup a separate DNS name for SNTP clients. For "ntpd" it's not currently practical. The ntpd developers are working on some new features that might make occasionally changing IPs work better with it, but it'll be a long time before they're widely deployed.
--
ask bjoern hansen
Something like this, with a proper run down to the receiver. With a RF Amplifier if needed.
For large sets, this will be our guide even unto death, for the LORD will work for each type of data it is applied to...
I actually have made that (running FreeBSD/NanoBSD), but it still costs $300-$400 or so -- seems like too much for a hobby when just running ntpd on some linux box you already have is almost as good. Maybe for people who have a static IP but no server running 24/7? Seems like a small group...
A small computer with the appropriate serial port (Soekris box, for example - $200 - $250 with power supply and small compact flash) plus Garmin 18lvc with the appropriate cabling (~$100). Then you still also need an (extra?) static IP address and space near a window (as you said). Doesn't seem like a big market!
--
ask bjoern hansen
The only down-side to USB GPS devices is that they don't have accurate 1PPS signals. A serial GPS can send the 1PPS signal to the DTR line where the computer can detect it for sub microsecond accuracy. Unfortunately serial ports are getting more and more uncommon, and use annoying +/-12V signalling.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
USB has the controller poll devices. Even on a dedicated bus there's a degree of uncertainty from the polling. Also, relying on NMEA data adds even more uncertainty, as there's no assurance that sentences are delivered in the right order or at timing more precise than one second.
My GPS triggers a serial interrupt when the PPS line goes high. The PPS line is within 1uS of UTC. After an hour or two to settle, NTP holds the time within +/-15uS.
Sure, one second precision is probably "good enough" for normal uses, but one can get more consistent time from most public servers. Providing one second precision time as a public time server is a bad idea, as NTP expects more consistent ticks and this will confuse other clients.
Running a serial GPS+NTP clock is pretty easy and provides much more stable time. Why bother with a USB GPS receiver when a more suitable serial+PPS capable one is available for only slightly more?
An USB GPS means no Pulse Per Second
Hrmmm .... good point - looks like it is available in a few devices.
esr says he can get 1ms on USB with the Macx-1 device. What accuracy is required for each stratum? The bufferbloat people are using that device for their latency measuring project.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
If "low power" wasn't a requirement you could do it for almost free. Old PCs are tossed every day that would run NTPd just fine. Problem is the damn thing would sit there humming away eating power.
my server doesn't need to know where its going, its stationary.
"The hands that help are better far than lips that pray." - Robert Ingersoll (1833-1899)
As I mentioned previously, after a few hours to settle NTP can match time to my GPS clock (which provides PPS output and NMEA sentences over serial) with a jitter of 15 microseconds on Linux (Ubuntu Server 10.04).
Using only the NMEA sentences over serial without PPS, jitter increases to ~250 milliseconds, roughly 16,000x more. Sure, it's "only" a quarter second, but still. It might be good enough for internal use but I wouldn't provide a public time server with a USB GPS clock.
Since USB receivers don't provide PPS output and the USB controller polls devices at intervals (rather than responding to serial interrupts), there's no way for a USB receiver to come close to the precision of a PPS-based serial connection.
I will have to check that out. I have an older machine that sits powered off but I would have no problem re-purposing it for this. I have a static IP on a business class line at home that sits unused most of the time. I didn't know there were cheaper devices than the big Symmetricom devices. Also how weather proof is that puck as the location of the computer I would attach it to would allow it to be put up on a south facing roof for better signal reception.
Time to offend someone
I don't think they WANT people with home connections. They're more interested in people who run their own servers in a remote datacenter or even VPS servers. They need always-on, not "on whenever the user isn't on holiday".
And on that basis, I don't know of a VPS provider that *doesn't* just provide 100's or even 1000's of GB's of traffic to each user.
My current host has a limit of 1TB of data per month, for example, and costs less than £20 a month. And that's not even a particularly cheap example.
You should set up a local router for your local machines to use as an NTP server and tell your DHCP to tell your hosts which NTP server to use. Just watch out when the router reboots since it may have no idea what time it is.
5 years ago I wrote a script that does a traceroute and then finds out of the hosts support NTP.
Its the bottom of my text on NTP Info page