At 02:26 PM 5/13/2004 +0800, zamri wrote:
Ray, Thank for your reply,
> The only ready way I know of to test the speed of a connection is to try to > do a large transfer over it ... usually, I find an ftp transfer to be the > handiest test. This can be difficult when you do not control hosts at both > ends of the link you want to test, but you still want to find some > approximation of this approach.
may you advice me how 'large' the file should used for my 1Mbps or 2Mbps connections speed? it is just okay with 10MB file transfer or it's still not _large_ enough? i could setup the ftp server at main office and branches office so i could test it from main office to branches office and vice versa. Is there any standard way for file size transfer to accurate the speed test?
I don't know of a standard. You want it to be large enough to give you time to observe the transfer rate, and to average over any transient effects that might be misleading. I tend myself to use something in the 30 MB range, like a Linux source kernel ... or something even bigger for testing a faster connection (E-to-E, say).
> you want to be as "close" as you can to your LAN (don't to a download > that involves multiple satellite hops from the opposite hemisphere); and > you want to be sure that the upload source serves fast enough that it will > not limit connection speed.
I thinks this is the hardies part, even if i setup the ftp server at branches office, it still get about 7-9 hops to reach main office LAN and of course i was using a diffrent ISP for the branches and main office. At main office the connections was 2Mbps Up and Down lease line, while on the branches office, i use 1.5 Up and Down SDSL line. And both running Bering Uclibc 2.1 as the router.
Right. Obviously (I hope this is obvious) you cannot use a source with a 1.5 Mbps upload limit to test a 2 Mbps download ... unless, I suppose, the 2 Mbps connection is running *way* below rated speed. You might do better to find a public ftp repository (a Linux-distro mirror, for example) near you and do some downloads from it.
> Beyond that ... when you say "i'm experiance a slow connections", what are > you actually observing? Anything other than the Web-page problem you > describe? Although you say "this 2Mbps are slower than 1Mbps that i > subcsibe before", that doesn't tell me what tests you are using now.
i know this wasn't accurate and reliable, but i thinks it will gave me some point to consider, for the temporary test, i using a mcafee speedometer test at http://us.mcafee.com/root/speedometer.asp . Before i upgrade the line, i get an average between 500 to 600Kbps but after i upgrade the line, i get an average between 100 to 400Kbps. How i'm doing the test, i test the connections about 50 times, plus all the results and divide by 50. Maybe i should do more in the next time and still, i dont thinks this was the best way to test the speed, that why i try to _asking_ an expert here for _advice_ on how i could do the speed test more reliablely.
I just ran this test here (only once) and it looks to me like a good test. Of course, I'm in the USA so probably "close" to the source site, unlike you. But to be honest, if I got here a result like the one you got, I'd stop wondering if my line was slow and start wondering where the problem was (in the line or in the router, that is).
[...]
> As to your Web-page problem ...[...]
I just move from dnscache to dnsmasq but the upstream dns still reffer to the same one. I will try to change back to dnscache and see the results.
Even simpler ... on the LAN host you are testing from, change its DNS from the router to use the "upstream" DNS resolver directly. If this improves performance, than suspect dnsmasq as the cause of the problem. If it doesn't, then use an app like "host" to test the response speed of the upstream DNS (which I assume is your ISP's DNS forwarder).
[...]> Is the problem with images a general one or is it limited to specific Web > pages? Details may matter here ... for example, if the images come from a > very different URL than the main page, you may be seeing a DNS timeout (the > images eventually load because the DNS request eventually resolves). But > this is just a wild guess, offered more to indicate the value of your > attending to the details of what does not load proprely.
The problem was a general one, even if i browse my own web site which sit at the dmz network from my local LAN still have this problem. For other site, it took an age to finnished loading the whole page espeacialy if the page have a lot of images :( i feel like i'm using an old 22.8Kbps dial-up modem.
Do you mean that images actually *stored* on your DMZ Web server (not just offsite ones referenced by it) exhibit this problem? If that's the case, then problems with connections from the DMZ to the LAN (do you mean the "staff ntwork" or the "student network"? or both? you seem to have 2 "local" LANs) are almost surely related to the router swap.
If this is not a DNS problem of some sort, I don't have any ideas to suggest based on this information. Perhaps someone else will be able to suggest something.
The only other thing I would direct your attention to is the extraordinarily high collision rate on eth3. As you report it ...
6: eth3: <BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast qlen 1000
link/ether 00:04:ac:6e:52:b8 brd ff:ff:ff:ff:ff:ff
RX: bytes packets errors dropped overrun mcast
3229811836 726754647 0 0 0 0
TX: bytes packets errors dropped carrier collsns
310837687 6346146 0 0 0 2149416
... 2 million collisions against 6 million packets is simply awful. But since I don't know if the "local LAN" exhibiting the problem is this one or eth0 (or both), I do not know if this is relevant to your problem.
------------------------------------------------------- This SF.Net email is sponsored by: SourceForge.net Broadband Sign-up now for SourceForge Broadband and get the fastest 6.0/768 connection for only $19.95/mo for the first 3 months! http://ads.osdn.com/?ad_id=2562&alloc_id=6184&op=click ------------------------------------------------------------------------ leaf-user mailing list: [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/leaf-user SR FAQ: http://leaf-project.org/pub/doc/docmanager/docid_1891.html
