3GB Time Capsule MD033 lowers data rate down to 5Mb/s during large files transfers

I just happen to be working on a thread with identical problem.

weirdly irregular throughput problem with two TCs

It starts fast and then slows to a crawl.

I did not solve it on the other thread.. so let me ask how much patience you have and ability to diagnose network issues?

I don't have a magic bullet solution.. even of silver.

(Actually a single shot from a 3030 would fix it permanently.. !!)

If you are prepared to pursue it I am happy to try..

NOTHING similar happens when I copy a large file from a PS located in the same network. 50-70Mb/s ever.

This indicates the issue is not the TC.. it is the Mac that has problems. (PS = PlayStation??)

In the Mac

You must have wireless OFF.

It is best to use ethernet setup with link-local only for IPv6 unless you require full IPv6.. tell me if that is the case.

I am doing the same thing you are. But I cannot reproduce your slowdown.

Open the activity monitor and track your CPU usage. One process is hitting 30% usage during the Read.. so it is fairly high.

Also check the network. Read (green) and Write (red) do make perfect scope scans.

I am wondering if the TCP window is set wrongly by default in your computers.. I am using a Mac Mini i7 with SSD so it is fast albeit 2011.

It is hard while I cannot reproduce the problem.

Ran out of time.

Performance tuning the network stack in Mac OS.

https://rolande.wordpress.com/2014/05/17/performance-tuning-the-network-stack-on -mac-os-x-part-2/

Might be a bit complicated.. but worth a google to see if others have options to improve the settings.

It can also be slow disk performance in the Mac.. are you running SSD??

I cannot see the ethernet port in the thunderbolt hub so I must track down that issue.

PS What is a TCP windows and how it could be checked?

TCP window is a part of the Network stack that has to do with buffering.

Mostly these things should be self adjusting but something has apparently gone wrong.

But the article I referenced has the info you require.. did you read it?

To help break it down just do the first part.. we need to see if memory buffers were starved.

In a terminal type the command, netstat -m

Here are my results.

netstat -m

329/2056 mbufs in use:

300 mbufs allocated to data

29 mbufs allocated to socket names and addresses

1727 mbufs allocated to caches

257/1934 mbuf 2KB clusters in use

0/517 mbuf 4KB clusters in use

0/72 mbuf 16KB clusters in use

7829 KB allocated to network (7.8% in use)

0 requests for memory denied

0 requests for memory delayed

0 calls to drain routines

As your system slows down please do that to see if memory is delayed or denied .. post the result as a copy and paste as I have done directly from terminal.

PS How do you take this screenshots - advice me please.

This is simple key combo.

Command + Shift +4 .. it gives you ability to select just a part of the screen. When you release the mouse key after the drag it will snap the shot and land it on the desktop.

BTW the activity monitor has much less useful info than what I posted which is from Mountain Lion.. they make all the utilities less useful as time goes on at Apple.

I have now swapped to a computer running El Capitan

And yes.. I can reproduce your problem.

This is to the Gen4 TC..

but it doesn't happen to a Gen 5 TC. Just to make life really confusing.

The write speed of this particular unit has always been poor but the read speed was always very good.

I can see no issues in Activity monitor..

Netstat shows running on the Gen 5 is using far less resources.

netstat -m

71/1940 mbufs in use:

53 mbufs allocated to data

18 mbufs allocated to socket names and addresses

1869 mbufs allocated to caches

18/1886 mbuf 2KB clusters in use

0/538 mbuf 4KB clusters in use

0/72 mbuf 16KB clusters in use

7774 KB allocated to network (0.8% in use)

0 KB returned to the system

0 requests for memory denied

0 requests for memory delayed

0 calls to drain routines

This is on Gen4..

netstat -m

306/1967 mbufs in use:

277 mbufs allocated to data

29 mbufs allocated to socket names and addresses

1661 mbufs allocated to caches

206/1886 mbuf 2KB clusters in use

0/534 mbuf 4KB clusters in use

0/72 mbuf 16KB clusters in use

7774 KB allocated to network (6.4% in use)

0 KB returned to the system

0 requests for memory denied

0 requests for memory delayed

0 calls to drain routines

After 30min still on the Gen4.

netstat -m

480/1987 mbufs in use:

466 mbufs allocated to data

14 mbufs allocated to socket names and addresses

1507 mbufs allocated to caches

379/1886 mbuf 2KB clusters in use

0/538 mbuf 4KB clusters in use

0/72 mbuf 16KB clusters in use

7774 KB allocated to network (11.4% in use)

0 KB returned to the system

0 requests for memory denied

0 requests for memory delayed

0 calls to drain routines

And after a few more minutes.

netstat -m

1267/2061 mbufs in use:

1267 mbufs allocated to data

794 mbufs allocated to caches

1092/1886 mbuf 2KB clusters in use

0/538 mbuf 4KB clusters in use

0/72 mbuf 16KB clusters in use

7774 KB allocated to network (32.3% in use)

0 KB returned to the system

0 requests for memory denied

0 requests for memory delayed

0 calls to drain routines

Note the buffer demand is getting worse.

So.. The info I have so far is that the issue is El Capitan on older TC. (Gen1-Gen4)

It is using a lot more network resources to maintain the dataflow.

That being the case we might be able to force things to AFP and see if that helps.

No.. no change using AFP instead. After the first few reads the data speed drops dramatically.

Life is not that simple however.. it seems to oscillate between slow and fast.

Here is the speed as a BTW to a single disk Synology.

You can see the tail end of the test to the Gen4 TC..

And the network buffer resources drops dramatically.

netstat -m

106/1949 mbufs in use:

80 mbufs allocated to data

1 mbufs allocated to ancillary data

25 mbufs allocated to socket names and addresses

1843 mbufs allocated to caches

25/1886 mbuf 2KB clusters in use

0/538 mbuf 4KB clusters in use

20/72 mbuf 16KB clusters in use

7774 KB allocated to network (5.2% in use)

0 KB returned to the system

0 requests for memory denied

0 requests for memory delayed

0 calls to drain routines

Just to confirm my results I got another Mac .. my laptop MBP 2011.. I can boot to Mavericks, Yosemite or El Capo..

The same problem showed up on el capo..

No problem at all with Mavericks.

It has been running now for 15min with no slow down.. and in fact the speed is better overall.

Both read and write manage about 40MB/s

Test on Yosemite shows the issue.

It did take quite a few cycles to start.. but there it is.

No.. even in reproducing it I cannot get consist results.. it happens but then can run perfectly for an hour..

I moved onto using a synology NAS which has no issues.. I'm afraid the old TC will simply not work due to their age.. the Gen5 TC I do have works fine.

http://www.apple.com/feedback/