Re: OT: time and computer networks
Thanks to everyone who responded, I appreciate the help! Thx especially to Alex, I feel I understand wha I needed to a little better. m -- .''`. Matt Price : :' : Debian User `. `'` & hemi-geek `- -- if you're an evil spambot, these addresses are for you: [EMAIL PROTECTED], [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: OT: time and computer networks
On Thu, Dec 02, 2004 at 12:59:57AM -0500, Matt Price wrote: > hi folks, > > ok... this is way OT. but I thought I'd put this question to the > most knowledgable group of people I know... > > I have to give a lecture on the history of timekeeping technologies. > I want to end with late c.20/ early c21 technologies of synchronized > timekeeping. GPS is one obvious example, NTP is another. But > puttingthe lecture together I realized I don'trelaly understand why > it's important for computer networks to have fine-grain > synchronization. So I thought I'd ask some geeks (as my sig says, I'm > only a hemi-geek): why does a network need careful clock > synchronization? Are packets like railroad cars -- in the sense that > it's VERY important to know which got sent first, and which is ocming > next -- and if you screw up the timeable, you get a catastrophe? Or > is there more flexibility in the system? > In San Francisco, in a museum of local history, there is a large clock which is described as the first official clock of the SF business community. Public time is important to business, as it is the way people (including business people, not just geeks) sequence events, decide cause and effect, discover defects in the data that is available, etc. This easily pre-dates the internet, and indicates why people are interested in knowing the 'correct' time. Real technologists, i.e. people like Edison and Westinghouse, know this and design systems that have built in sequencing and synchronization. For example, Y2K was a real big deal for the billing department of the local electric company, but had nothing to do with keeping the various AC generators on the grid in sync. (Please, no flames about hidden chips inside essential boxes. Those were not designed by real technologists.) The best internet time is many orders of magnitude inferior to the best NIST laboratory clock time. The time available from GPS is in between in quality, i.e. far better than internet time, but far inferior to the best NIST lab. time. GPS time is 'good enough' for it job, namely determining position from signal time of arrival measurements to an accuracy of a few meters. Also, GPS time can never be as good as NIST lab. time, because the motion of the GPS satelites makes synchronizing two clocks to NIST lab. accuracy a meaningless exercise. -- Paul E Condon [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: OT: time and computer networks
On Thu, 2004-12-02 at 00:59 -0500, Matt Price wrote: > puttingthe lecture together I realized I don'trelaly understand why > it's important for computer networks to have fine-grain > synchronization. So I thought I'd ask some geeks (as my sig says, I'm > only a hemi-geek): why does a network need careful clock > synchronization? Kerberos and Windows 2000+ domains (which are based on kerberos) require synchronized clocks so that tickets' timestamps can actually be honored. If your system's time is too far off from the kerberos ticket server, you cannot authenticate. -Mark -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: OT: time and computer networks
On Thu, 02 Dec 2004 08:56:44 -0600, Alex Malinovich <[EMAIL PROTECTED]> wrote: > Modern timekeeping systems have primarily been put in place for the > benefit of the humans using the computers, not so much the computers > themselves. The reason that things like NTP are so accurate is not > because the majority of people using them NEED them to be so accurate, > but because: There's another reason to keep your computer well-synched, and that's if you're performing micro-benchmarks. If you're really interested in how many microseconds something takes (say because you're going to scale it by a factor of several million), then you'd like to know at what rate your clock tends to drift, and by how much that varies. If you're just interested in having a reasonably accurate clock (within a minute), then you certainly need less precision. -- Michael A. Marsh http://www.umiacs.umd.edu/~mmarsh -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: OT: time and computer networks
On Thu, 2004-12-02 at 00:59 -0500, Matt Price wrote: > hi folks, > > ok... this is way OT. but I thought I'd put this question to the > most knowledgable group of people I know... > > I have to give a lecture on the history of timekeeping technologies. > I want to end with late c.20/ early c21 technologies of synchronized > timekeeping. GPS is one obvious example, NTP is another. But > puttingthe lecture together I realized I don'trelaly understand why > it's important for computer networks to have fine-grain > synchronization. So I thought I'd ask some geeks (as my sig says, I'm > only a hemi-geek): why does a network need careful clock > synchronization? Are packets like railroad cars -- in the sense that > it's VERY important to know which got sent first, and which is ocming > next -- and if you screw up the timeable, you get a catastrophe? Or > is there more flexibility in the system? Modern computer timekeeping has little to do with actual transmission of data over a network, at least as far as Ethernet-based networks are concerned. Now, if you're talking about a T1, for example, timing is still critical. Due to the way T1's transmit data, they require a timing subcarrier to keep the sender and receiver in tune with each other. Ethernet, on the other hand, is based on sequencing. So a packet sent now vs a packet sent two minutes from now don't necessarily need to arrive in the correct order, as the receiving side will re-order them once they're all in. (There are limitations to this of course, but they're outside of the scope of what I'm going to go into here.) Modern timekeeping systems have primarily been put in place for the benefit of the humans using the computers, not so much the computers themselves. The reason that things like NTP are so accurate is not because the majority of people using them NEED them to be so accurate, but because: 1) The people who originally designed them were uber-geeks who knew that there was only one 'right' way to do things, and getting clocks within 100 ms of each other just wasn't good enough because it wasn't the 'right' way. 2) The people who originally NEEDED them were people who play with nukes and alien specimens (well, maybe not the latter, but who knows) and they absolutely HAVE to have the EXACT time. Today, the big benefit of having closely synchronized networks is in business, and especially financial applications. Say that I'm a huge corporation, with a $1,000,000,000 bank account, and I want to buy another corporation for $1,200,000,000. I just sold a whole bunch of my product and earned $210,000,000 on the deal. I know that the money will be in my account at 11:00 am. At 11:00:00:01 my computers begin the purchase of the new company. If the bank and I aren't on the same time, the transaction is going to fail and I might very well get slapped with a multi-million dollar bank charge in the process. Granted, this is a rather extreme and unrealistic example, but that's one of the reasons why computer networks nowadays need to be synchronized so closely. Hope that helps. And when you're done giving the lecture, be sure to urge the listeners to use Debian and give up the evils that are Windows. :) -- Alex Malinovich Support Free Software, delete your Windows partition TODAY! Encrypted mail preferred. You can get my public key from any of the pgp.net keyservers. Key ID: A6D24837 signature.asc Description: This is a digitally signed message part
RE: OT: time and computer networks
> From: Ron Johnson [mailto:[EMAIL PROTECTED] > Sent: Thursday, December 02, 2004 1:14 AM > > On Thu, 2004-12-02 at 00:59 -0500, Matt Price wrote: > > hi folks, > > > > only a hemi-geek): why does a network need careful clock > > synchronization? > > It's applications & humans that need/want clock sync. A quick, > simple example is forensic analysis of log files. Another popular example would be remote filesysems (i.e. NFS, Samba, etc.). For instance, my soure files are stored on an NFS mount, but I compile on a local box. Make depends on timestamps to know what to compile. If the systems don't synch up very well, there is a chance a changed file will not be remade. Also imagine if your boss looked at your TPS reports on the Samba server and found that they were dated in the future (according to the clock on his PC). Or worse 3 years in the past. (Ron, thanks for the "Reply" head's up) -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: OT: time and computer networks
On Thu, 2004-12-02 at 00:59 -0500, Matt Price wrote: > hi folks, > > ok... this is way OT. but I thought I'd put this question to the > most knowledgable group of people I know... > > I have to give a lecture on the history of timekeeping technologies. > I want to end with late c.20/ early c21 technologies of synchronized > timekeeping. GPS is one obvious example, NTP is another. But > puttingthe lecture together I realized I don'trelaly understand why > it's important for computer networks to have fine-grain > synchronization. So I thought I'd ask some geeks (as my sig says, I'm > only a hemi-geek): why does a network need careful clock > synchronization? Are packets like railroad cars -- in the sense that > it's VERY important to know which got sent first, and which is ocming > next -- and if you screw up the timeable, you get a catastrophe? Or > is there more flexibility in the system? > > anyway, it's just a question. I'd love to hear some answers. "The network" doesn't need careful clock synchronization. As evidence, note that *most* IP nodes do not run any sort of time sync app. TCP packets are ordered at the end point using sequence numbers. It's applications & humans that need/want clock sync. A quick, simple example is forensic analysis of log files. -- - Ron Johnson, Jr. Jefferson, LA USA PGP Key ID 8834C06B I prefer encrypted mail. No matter how hard women work, train, excersize (and some of the woment in the Olympics were *muscular*), lose body fat, or get Title IX money, (the population of athletic) women will *never* be faster, stronger or better at sports(*) than (the population of athletic) men. (*Excluding "small sports" like darts.) It's just how men and women have evolved. Deal with it. signature.asc Description: This is a digitally signed message part
