very slow file transfer between my 2 macs
 

hi,

i have a airport network that connects a g4 mdd dual 1 ghz and a ti powerbook 400 mhz. the abs is connected to 1 cable modem and both computers share the modem. the base station is extreme and both makes have older airport cards.

i find that when copying any size file between computers the transfer speeds a very, very slow. a 300 mb file can take over 30 mins. i have file sharing turned on both macs.

i know it was this slow before i even had a airport, so maybe i am doing something wrong.

any suggestions are greatly appreciated.


thank you,

-c

Please correct me if my math is bad..

AirPort transfers at 11 megabits / second
That is 11,000,000 bps, divided by 8 (1 byte = 8 bits)
= 1,375,000 Bps
1,375,000 / 1024 (1024 bytes in a kilobyte) = 1342.773 KBps
1342.773 / 1024 (1024 kilobytes in a megabyte) = 1.311 MBps

300 MB file / 1.311 MBps = 228.83 seconds to transfer the file
228.83 seconds = 3.814 minutes.

Defintely something wrong here. Are you using AFPoverTCP or Samba or Appletalk?

Or is my math wrong?

i just use afp - command-k through the finder. i have the mdd g4 set to 10.0.1.201, abs- 10.0.1.1 and the powerbook is 10.0.1.3. i use my g4 as a ftp server so i can can transfer large files from work to home.

before i had the airport network, i had a netgear 10-100 router and it was just as slow. we have a network here at work set up the same way between 10 macs and it's super fast.

it's so painfully slow...

Try using the FTP server to see if the transfer is faster. Also, when you use cmd-k, are you just selecting the other machine from a list or are you entering it's IP?

Don't know if this will go anywhere, but how about an 'ifconfig en1' (I think it is en1 for airport) for both machines. As well, do you get better transfer rates from the internet?

ok, here is what i just came up with. bothe tests use the same 11.3 mb jpeg file.

afp:

tibook to g4- 3 mins
g4 to tibook- 2 mins

pureftpd:

pb-g4- 5 mins

wtf?????

it's slow.

yellow, when i command-k i just select the name of the computer i am connecting. i connect as the admin account.

when i transfered the file from the g4 to tibook itunes 4.0.1 started stutering and quit.


please help me, this is awful!!!!

Try entering the remote computer's IP address rather than selecting from the list. Then we're using AFPoverTCP rather than straight AppleTalk. This should be faster. However, since your FTP client upload was so slow, there's something else at work here.

If you're regularly transferring large files, you might use Target Disk Mode (where one of the computers acts as a FireWire drive). Hold down T during startup on the computer with the file you want to transfer.

You'll need a FireWire cable to string between the two computers.

I don't bother coping large files over Ethernet if I can at all avoid it.

it's still just as slow, if not worse. i find that if i copy a 28 mb file from my g4 to the tibook it will take 2 mins. however if i copy the same file back to the g4 it will take 5 mins.

what can cause such a thing? my tibook is further away from the abs and i get a weaker signal than the g4.

firewire target disk mode is way to inconvienent...
i would rather burn a cd or use my ipod to transfer the file. thanks anyway.

we have a similar setup at work without a airport and a imac as a server. there are 10 users that copy huge files and a 28 mb file would take 30 secs.

Are there a lot of apps running at the same time on both machines? Possibly there's not enough CPU cycles to run all your apps and copy large files? I'm struggling here. You might want to check out the man page on tcpdump and try to see exactly what's going on on your little network. Maybe a neighbor has hijacked your signal/bandwidth?

i'm going to add my $0.02 here and say that i, too, have this problem. i have a snow airport basestation that distributes my DSL connection to two laptops. one is a 15" TiBook (roommate) and the other is a brand-spankin'-new 12" AlBook (mine ;) )

i have tried two different methods of connecting the two computers via airport. (1) using cmd+k in the finder and selecting his computer from the list or sometimes i type in the IP directly. and (2) using the airport.menu to create an ad-hoc computer-to-computer network with just the two computers joining.

both result in absolutely horrible data transfer rates. i know that i can rule out the airport cards in both machines because both can indepently download larges files from outside the network at great speeds via the DSL.

here is the result of ifconfig en1 on the TiBook:
and here is the result on my 12" AlBook:
there are usually NO other programs open at the time of transfer, and neither of us run any third party hacks or other such nonsene things. we also run a secure Wi-Fi network, so unless someone is sitting out there packet sniffing like a mad man, i don't think my bandwith has been compromised.

bustthis, i hope some people have some suggestions for the both of us.

i do see another name in my airport menu sometimes called thecrib85, but i have a closed network and have posted here about it.

my g4 has 1.75gb ram and is dual 1ghz. i have illustrator, photoshop and mailsmith open, the tibook has nothing open.

my g4's ifconfig looks like this:
lo0: flags=8049<UP,LOOPBACK,RUNNING,MULTICAST> mtu 16384
inet6 ::1 prefixlen 128
inet6 fe80::1%lo0 prefixlen 64 scopeid 0x1
inet 127.0.0.1 netmask 0xff000000
gif0: flags=8010<POINTOPOINT,MULTICAST> mtu 1280
stf0: flags=0<> mtu 1280
en0: flags=8822<BROADCAST,SMART,SIMPLEX,MULTICAST> mtu 1500
tunnel inet -->
ether 00:03:93:da:ab:30
en1: flags=8863<UP,BROADCAST,SMART,RUNNING,SIMPLEX,MULTICAST> mtu 1500
inet6 fe80::230:65ff:fe0c:76b7%en1 prefixlen 64 scopeid 0x5
inet 10.0.1.201 netmask 0xffffff00 broadcast 10.0.1.255
ether 00:30:65:0c:76:b7
media: autoselect status: active
supported media: autoselect

and this is ifconfig en1:

en1: flags=8863<UP,BROADCAST,SMART,RUNNING,SIMPLEX,MULTICAST> mtu 1500
inet6 fe80::230:65ff:fe0c:76b7%en1 prefixlen 64 scopeid 0x5
inet 10.0.1.201 netmask 0xffffff00 broadcast 10.0.1.255
ether 00:30:65:0c:76:b7
media: autoselect status: active
supported media: autoselect

At the risk of being *really* simple here, what sort of signal strength do you get when transfer files? I ask because I have an airport network and when I try to transfer from my desktop to my TiBook, I find that the transfer times vary markedly on the strength of the signal (for both machines).

My desktop gets good reception at the end of the hall, but the TiBook doesn't do so well in that room. When I transfer files that way, it takes forever. If I walk down and sit closer to the router with my TiBook, transfer times improve dramatically.

Finally, if I enable computer to computer networking and sit in the same room, the transfer rates are great!

Obviously, I can't move my desktop around very easily, and it's signal remains steady at 85-90%. The TiBook signal varies from 60-95%, depending on where I use it. As the signal strength improves, so too do the transfer rates.

Just a thought.

Best,
darin

I have no experience with airports, but does the 1500 mtu cause it grief? I notice jdhorner's roommate has his set to 1492 (as I have on my normal ethernet as some devices choke on 1500).

Might be worth a try doing on both ends and see if thing speed up....

well, both computers and the basestation are almost always in the same room. and for testing purposes, i have ruled that out anyway. bustthis... what about you.

well, my g4 is 4 ft away from aebs and i get full signal and the tibook was on the same desk when i was copying the 28 mb file. it's usually about 15 ft away from the aebs, but it was pretty slow before i even had a airport.

how would i know if someones stealing my connection? last month my super was stealing my cable, he lives in the basement, i live on the 4th floor. from time to time i see another available airport network called thecrib85 in my airport menu. it's weird it will only appear there after i click the drop down menu a couple of times in a row. i posted here a while back about this and someone help me make a closed network. i have a lot ports mapped to my g4. because i need to use it remotely from work for file transfers via ftp.

i think /sbin/ifconfig en1 mtu 1492 made it slower, if you can beleive that one!

my itunes craps out when copying the 28 mb file back and forth between both computers. the sound gets muffled until it quits. i have no clue!!!

i just woke up my TiPB/500 and it has a wifi connection to a wifi access point hanging of my router.

i had abysmal xfer speed over AFP.

i switched it to the wired enet port and got 7-8MB/s

switched it back to wifi and then got the expected 7-8 xfer rate.

so, does a sleeping wifi connection get eyeglue and need a wipe?

real-world performance
 

I don't think your math is wrong. But your assumption about the transfer speed is wrong. The 11 Mbps is the nominal rate for an 802.11b network. But the real-world maximum transfer rate is less than half of that.

See http://www.pcworld.com/news/article/...83,pg,2,00.asp
for some measurements that show a maximum transfer rate less than 5 Mbps with WEP off. Enabling WEP will lower the transfer rate significantly.

If you are interested in why the actual rate is lower than the nominal rate, see this article that calculates a theoretical maximum rate of 5.6 Mbps for 802.11b:
http://www.oreillynet.com/pub/a/wire...hroughput.html

So, your calulation is off by a factor of at least 11/5, which brings the minimum time for transfer of that 300 MB file up to 8.4 minutes.

I.e. his 30 minute transfer time is only a factor of 4 slower than the best he could expect. So yes, there likely is something that could be improved in his setup, something that is slowing it down, but the problem isn't quite as big as it might seem.

I also note that you shouldn't expect more than about 50% of the nominal transfer rate for Ethernet networks either.

Well, it was worth a try... As for the iTunes thing, I would think that you system is to busy transferring that file, so iTunes cannot get the throughput it needs to keep playing.

Cool, thanks hayne. It's been a while since I dusted off the calculator. :) I suppose if I'd thought about it I would have realized that 11 Mbps was the optimal speed.. but anyway, I concur that there's something amiss here.

Bustthis, you really should test without any apps running on either machine and see what kinda speed increase we get (if anything noticable at all). At this point I think you've become a lab rat for 802.11b transfer rates. :)

no apps running on either machine after a restart tried to copy a 350 mb folder and it took 49 mins!!!

i'm lost at this point... could it be a bad ethernet cable going from the abs to the cable modem? what can i do!!!

As far as I know (correct me if I'm wrong) the cable modem has nothing to do with this. The traffic between machines should be passed (repeated) only through the base station, so cabling should be an issue.

Hmm, do I understand the situation correctly? You are trying to transfer files between your two machines, through and airport extreme base station, and these machines are communicating (however slowly) through the airport connections, right?

If that's the case, I would assume (perhaps incorrectly) that the ethernet cable connecting the aebs to the cable modem is not a factor.

I would be interested to know what sorts of transfer rates you get when you transfer from machine to machine. I realize that JDHorner gets bad rates even here, so maybe I'm just a curious bird. But hey, that's me.

Best,
darin

[edit]Ya, what Yellow said.[/edit]

your right and that makes sense...

any ideas on what i could do at this point?

I'm just musing here, so take what you will---

Would it be at all useful to try the same transfer with the computers wired to the base station, so that they were transfering through ethernet cables as opposed to through airport connections? Is this possible in your location?

I'm wondering two things: 1. could you isolate this problem to the airport connection and 2. is there perhaps, despite Apple's claims, a less than idea connection made when you combine an older airport card with a newer extreme base station.

Again, just musing, and I don't have a base station, so I don't know too much about them. I hope that somebody will correct my errors.

Best,
darin

my numbers
 

Prompted by this thread and another one about slow network transfers between Mac & PC, I spent some time to measure the performance I get on my home network.

The network:
1) MacSense router connected to DSL "modem". The switching side of this box does "fast" Ethernet: 100 Mbs
2) Airport 802.11b basestation, configured to act as a "bridge" to the router. 128-bit WEP is enabled.
3) Cat-5 cable between all nodes

The computers:
A) iBook 600 Mhz, running 10.2.6
(connects via Airport and/or Ethernet to the router)
B) HP Pavilion PC, 800 Mhz, running Windows ME
(connects via Ethernet to the router)

In the following experiments, I mounted a shared folder from the PC on the iBook via SMB ("Connect to server" in the Finder) and copied a 33.5 megabyte QuickTime movie file back and forth, deleting the copied file after each test.

When I disconnect the Ethernet cable from the iBook (so that it is connected only by Airport), copying a 33.5 MB file from the iBook to the PC takes about 2.5 minutes - i.e. 150 seconds.
That translates into a transfer rate of 1.9 Mbps
(to be compared to the nominal 11 Mbps of 802.11b).

When I turn off Airport on the iBook and reconnect the Ethernet cable, it takes only 21 seconds to copy that same file. That translates into a transfer rate of 13.4 Mbps (to be compared to the nominal 100 Mbps of "fast" Ethernet).

One thing to note: I have turned down the MTU on my iBook's Airport card to 1384 since anything higher seems to give trouble when sending larger email messages via my DSL connection. Presumably I would get a bit better Airport transfer rate if my MTU was at the default 1500.

I might be trying to improve my network speeds if my configuration were stable. But, except for my iBook, most of my machines (there are more than the 2 I mentioned) are turned off or hibernating most of the time - to cut down on the noise. So my typical transfer time is dominated by the time to boot the machines!

dhayton you bring up a good point here. or, at least, a good idea. i was excited to think that perhaps it was the fact that there was at least one airport extreme device in the mix that's the culprit. (both my basestation and the 15" are NOT extreme. only the new 12" AlBook is).

but then i realized that the 12" iBook that this new computer just replaced was also non-extreme. so even back when everything was only 802.11b i still had the same problems.

i'm going to bed now, but will hopefully be able to test some of the suggestions here some time tomorrow. my first test will be to disable WEP completely.

as a side note, and i never even really noticed it before, does anyone know what the option to "use interface robustness" in the airport.menu item actually DO?

keep in mind that file transfers are greatly affected by the disk subsystem itself. It tends to be a major bottleneck. If you want to eliminate issues with the network, you need to test the connection by using a program such as "ttcp" and do a memory-to-memory transfer of bytes, omitting disk involvement. you should do this in one direction and then in the other. something like ttcp lets you set the buffer size of the socket. experimenting with this will tell you where the bottleneck is, if any, on transfer size. Note that programs that doen't allow u to set transmit/receive buffer sizes make the problem harder.

laptop disks tend to be slower and less performing than SCSI,fast IDE etc. Also the program used to transfer a file has inherent limitations due to over head of the protocol it uses.

The ability of a system to buffer data in front of the disk subsystem is also a factor. that's why some disks have large caches to help, but small writes can hurt this.

In benchmarking systems, I tend to use something like ttcp to help "tune" tcp/ip and when I think I have that as best I can, I then look at the disk syubsystem. Systems that have raided disk will obviously write more efficiently than trying to pound data onto a single disk, especially a laptop.

your mileage will vary.

diagnostic tools
 

There is a newer tool 'iperf' which is claimed to be better than 'ttcp'. Version 1.6.5 is available via fink but a newer version is available from the iperf web site:
http://dast.nlanr.net/Projects/Iperf/
I haven't tried it yet.

That web site led me to a page with a very interesting analysis of OS buffer sizes and how they prevent us from attaining the maximum bandwidth available on high-speed networks:
http://www.psc.edu/networking/perf_tune.html

And that web site linked to a useful diagnostic tool (Java applet + specialized back-end server) that can tell you if your Internet connection is suffering from misconfiguration:
http://miranda.ctd.anl.gov:7123/
Even though you have been talking about LAN performance, this tool might give some clues since it of course tests part of your LAN in getting to the Internet.