Expected network throughput on Mediator?

[Heiroglyph]

My system is an A4000 060/50/96MB, Mediator, Realtek 8139D and Radeon 128MB with 93MB used for fast RAM.

I'm running Roadshow for TCP. (I love it so far, well worth the price)

When I use AmiFTP to transfer files from my server (local server, gigabit, fast RAID and CPU) I only get about 780,000 cps whether I transfer to RAM: or a hard drive.

I've done a few of the Roadshow tweaks mentioned in another thread, is there anything Mediator specific that can be adjusted?

Is this normal? It seems really low for 100Mb.

[danbeaver]

Well, 100 Mb/s (megabits per second) equals 12.5 MB/s (12.5 megabytes/sec) with no overhead (error checking,mostly), but only about 50% can be used, so 50 Mbits/s data transfer =~ 4.7 to 6.0 MB/s (max).  Now you are getting < 1 MB/s, so there is a bottleneck.  The Buster chip limits the estimated 150 MB/s Zorro 3 speed to about 13.5 MB/s and the limit on Amiga to PCI card in a Mediator seems a max of 12 MB/s.  Jens of icomp.de says,  "Roadshow is a commercial product still available today, but since it does not reach the speeds that AmiTCP/Genesis can reach" but I can't find results of his testing.  He does state elsewhere, "Transfer rates of over 1600kBytes/s [1.6 MB/s] have been measured with the X-Surf-100. Please note that the classic Amiga architecture and the current TCP/IP stacks for the Amiga don't allow saturating a 100MBit Ethernet link. Performance in your system may be slower due to CPU performance and programs running in multitasking."

I think that somewhere in there is your answer.

[Heiroglyph]

I found something of use on eab, a netio benchmark from matthey using an 060/75 and Mediator.

NETIO - Network Throughput Benchmark, Version 1.32
(C) 1997-2012 Kai Uwe Rommel

TCP connection established.
Packet size  1k bytes:  2033.78 KByte/s Tx,  321.98 KByte/s Rx.
Packet size  2k bytes:  405.09 KByte/s Tx,  574.51 KByte/s Rx.
Packet size  4k bytes:  723.33 KByte/s Tx,  766.37 KByte/s Rx.
Packet size  8k bytes:  1320.06 KByte/s Tx,  876.72 KByte/s Rx.
Packet size 16k bytes:  1618.56 KByte/s Tx,  1356.06 KByte/s Rx.
Packet size 32k bytes:  1644.96 KByte/s Tx,  1663.01 KByte/s Rx.
Done.

I'm not sure my setup will actually complete the test, it seems to stop, but not lock up.

With the tweaks given here http://www.amiga.org/forums/showpost.php?p=751014&postcount=48 it just stops after giving Packet size 1k bytes:  923.34 KByte/s Tx and never seems to finish Rx.

Removing all the tweaks gives:
1K bytes: 901.88  KByte/s Tx,  946.49 KByte/s Rx.
2K bytes: 1067.00  KByte/s Tx,

Then it just stops again.

Edit: it seems that after running netio, Roadshow is no longer functional. GetNetStatus looks good, but not even ping will complete, it just sits there with no crash.

[Heiroglyph]

I uninstalled Roadshow and installed Genesis (yuk)

It will now complete the test and has somewhat better performance, but still crap. It also locks the machine when I go offline The machine was incredibly stable prior to this.

1k: 1375 tx, 184 rx
2k 147 tx, 272 rx
4k 275 tx, 276 rx
8k 510 tx, 531 rx
16k 946 tx, 835 rx
32k 1004 tx, 868 rx

And on FTP I get 1,138,000 cps.

I've tried two different realtek cards with no change and two different versions of the FastEthernet.device

[Heiroglyph]

Changing the 10/100 card for an 8029 and still using Genesis gives me:
1k  843 tx, 446 rx
2k 231 tx, 427 rx
4k 400 tx, 654 rx
8k 551 tx, 651 rx
16k 599 tx, 739 rx
32k 703 tx, 714 rx

FTP gives 982,000 cps.

In many cases, that's actually an improvement.

WTF?

[danbeaver]

The X-surf 100 has tweaks for AmiTCP in the SANA2 ENV file and AmiTCP version 4 is available from Amigakit as Easynet Pro.  As Jens posts commonly on EAB, I'd search there for the tweaks.

I WTF'd your findings in the above post too.

[matthey]

Hi Heiroglyph,

I'm the Matt Hey with the Mediator+Genesis results you found on EAB. I do have a very fast 68060 and memory so my results are probably not obtainable by most 68060 owners. There seems to be some bottleneck on my system also as my results vary between fast and slow. I have no connection or stability problems though.

I don't know what is causing your problems so I will give you my working setup.

3000T, CSMK3 68060@75MHz, AmigaOS 3.9 with fixes

My ENVARC:Mediator variables:

FastEthernet 0
MMU YES
VoodooInt YES (may be different depending on gfx board)
VoodooMem 30 (I have a Voodoo 4 so this will likely be different for you)

Library and device versions:

Libs:68060.library 43.4 Mu/MMU.lib from ThoR
Devs:Networks/FastEthernet.device 1.25
Libs:pci.library 9.7
Libs:Picasso96/Voodoo.card 4.34

Note that some of my versions are older because of new bugs.

I uninstalled Roadshow and installed Genesis (yuk)

Genesis isn't that bad is it? It has a GUI with a choice of ClassAct or MUI and it is fast.

My Genesis prefs:

(Interface tab)
Name: FastEthernet
Comment:
IP Address: 192.168.0.100 (static)
Destination:  (dynamic)
Gateway: 192.168.0.1 (static)
NetMask: 255.255.255.0 (specify)
MTU: 1500
Params:

[Heiroglyph]

Thanks for chiming in.

Hi Heiroglyph,

My ENVARC:Mediator variables:

FastEthernet 0
MMU YES
VoodooInt YES (may be different depending on gfx board)
VoodooMem 30 (I have a Voodoo 4 so this will likely be different for you)

Ethernet = 0

I have no MMU variable and MedConfig isn't asking me about one. I'll manually set it and see what happens.

Edit: Setting MMU didn't seem to change anything.

The Radeon Memory is set to 32 and RadeonMemOS = YES.

Library and device versions:

Libs:68060.library 43.4 Mu/MMU.lib from ThoR
Devs:Networks/FastEthernet.device 1.25
Libs:pci.library 9.7
Libs:Picasso96/Voodoo.card 4.34

Note that some of my versions are older because of new bugs.

I'm using the same 060 library and fast ethernet device.

My PCI library is 9.10 though.

Genesis isn't that bad is it? It has a GUI with a choice of ClassAct or MUI and it is fast.

My Genesis prefs:

(Interface tab)
Name: FastEthernet
Comment:
IP Address: 192.168.0.100 (static)
Destination:  (dynamic)
Gateway: 192.168.0.1 (static)
NetMask: 255.255.255.0 (specify)
MTU: 1500
Params:

Mine is set the same, plus your suggested changes from the eab thread, increasing the send and recv buffer sizes.

Thanks,
John

[matthey]

I have no MMU variable and MedConfig isn't asking me about one. I'll manually set it and see what happens.

Edit: Setting MMU didn't seem to change anything.

The Radeon Memory is set to 32 and RadeonMemOS = YES.

Yes, the directions have changed over the years but my old config still works. It's probably not the best for a new Radeon Mediator setup though.

I'm using the same 060 library and fast ethernet device.

My PCI library is 9.10 though.

I had problems with the new pci.library 9.8+. The screen will lose sync while using some Warp3D programs (Alan Thellier's Cow3D for example). I don't know if this would be a problem with 2D apps or the Radeon driver but there was at least one bug introduced after pci.library 9.7. The new pci.library is supposed to allow faster ethernet speed with fast ethernet but I wonder if it is broken or the timings are too tight for some setups.

You could try to find an old Voodoo 3 for testing. This would allow you to play with Warp3D programs also.

[Heiroglyph]

Yes, the directions have changed over the years but my old config still works. It's probably not the best for a new Radeon Mediator setup though.



I had problems with the new pci.library 9.8+. The screen will lose sync while using some Warp3D programs (Alan Thellier's Cow3D for example). I don't know if this would be a problem with 2D apps or the Radeon driver but there was at least one bug introduced after pci.library 9.7. The new pci.library is supposed to allow faster ethernet speed with fast ethernet but I wonder if it is broken or the timings are too tight for some setups.

You could try to find an old Voodoo 3 for testing. This would allow you to play with Warp3D programs also.

I've got a Voodoo, but I can't use resolutions that match my monitor with it. I did go back to pci.library 9.7, but I'm not sure if it was needed, I'm just staying there for safety sake right now.

For the time being, I've also gone back to Roadshow. I have better stability with it other than this one benchmark and it now seems to give similar performance to Genesis.

Who knows, it could be a bug there or I may find that I hit the bug during real-world use eventually.

I also found out that I could use netshutdown followed by addnetinterface to safely recover from the netio hang.

I'm able to get a bit over 1MBps now with the settings I'm using, both send and receive in netio (until it hangs and I have to ctrl-C) and using ftp from an in-house server. Those huge dips seem to be gone from what I can tell.

I set tcp.mssdflt to 536, the standard default size, and it actually seemed to help. 1500 is slower here, 1400 was better than 1500 and 536 seems completely safe and fast enough.

I raised the tcp.recvspace to 65536, but not the send. Raising both seemed to cause issues, but raising just this one seemed to pull up the receive to match the already good send speed. I think this was one of the biggest improvements.

In the interface, I set iprequests to 77 (a magic number from another thread that I'm now afraid to touch) and writerequests to 64. These seemed to also make a noticeable difference.

It still seems odd to get such slow speeds, but with the weak CPU and PCI implementation we have, that's probably about the best I can hope for.

[kolla]

Always thought it was funny how I got better networking and filesystem performance with Linux than with AmigaOS on same hardware, and on Linux I got a much more modern and feature rich IP stack too.

[danbeaver]

The Mediator throws a few things in the calculations, but most of the settings (such as MMU) have nothing to do with an 060 board.  The graphics memory sets aside 32 MB (above) for uses as graphics, with 1 MB required for the pseudo-DMA and the rest under the later PCI.library useful as system RAM.

Note that the 1 MB pseudo DMA eats read/write cycles for every 1 MB of data transferred on the PCI bus.

[olsen]

I found something of use on eab, a netio benchmark from matthey using an 060/75 and Mediator.



I'm not sure my setup will actually complete the test, it seems to stop, but not lock up.

With the tweaks given here http://www.amiga.org/forums/showpost.php?p=751014&postcount=48 it just stops after giving Packet size 1k bytes:  923.34 KByte/s Tx and never seems to finish Rx.

Removing all the tweaks gives:
1K bytes: 901.88  KByte/s Tx,  946.49 KByte/s Rx.
2K bytes: 1067.00  KByte/s Tx,

Then it just stops again.

Edit: it seems that after running netio, Roadshow is no longer functional. GetNetStatus looks good, but not even ping will complete, it just sits there with no crash.

It can be dangerous to change the memory usage settings for the TCP/IP stack. Given enough traffic at a sustained high rate, the TCP/IP stack may end up consuming most available and usable memory (with respect to fragmentation, etc.).

I first saw this with AmiTCP V4, when I tested it with the X-Surf 100, and given that the underlying TCP/IP stack architecture is the same, right down to how the memory management works, Roadshow must be susceptible to the same issues.

Did you read the entire thread on which you found the tweaking information (there's some information on this subject in the Roadshow documentation, too)? There is more information to be found there, including how fast the X-Surf 100 can be expected to be, and under which circumstances.

In my own tests I found that 2.2 MBytes/s should be possible if you don't actually use the data transmitted. In reality with a web browser or an ftp client using the data transmitted you might get more than half of the 2.2 MBytes/s, and that's still a lot more than the Zorro II Ethernet cards can deliver. The X-Surf 100 is the fastest Amiga Ethernet card there is.

[Oldsmobile_Mike]

In my own tests I found that 2.2 MBytes/s should be possible if you don't actually use the data transmitted. In reality with a web browser or an ftp client using the data transmitted you might get more than half of the 2.2 MBytes/s, and that's still a lot more than the Zorro II Ethernet cards can deliver. The X-Surf 100 is the fastest Amiga Ethernet card there is.

X-Surf 100 has a jumper to switch between ZII and ZIII mode.  Pretty sure I've seen numbers somewhere of folks getting as much as 6.6MB in ZIII.  Might be mis-remembering, however.

[Oldsmobile_Mike]

In my own tests I found that 2.2 MBytes/s should be possible if you don't actually use the data transmitted. In reality with a web browser or an ftp client using the data transmitted you might get more than half of the 2.2 MBytes/s, and that's still a lot more than the Zorro II Ethernet cards can deliver. The X-Surf 100 is the fastest Amiga Ethernet card there is.

Probably a dumb question, but how is it that a RapidRoad connected to an X-Surf 100 in ZIII can get up to 6-7MB/sec (source:  http://www.fitzstevesamigaworld.co.uk/?p=253 ), but Ethernet only 2.2MB?  Is it all just protocol overhead?  Seems like 2.2MB/sec isn't a bus limitation if RapidRoad can do so much faster through basically the same pipeline.