My darn NAS...

you said gigabit Ethernet and now you are saying Wifi?

Wifi is to put it bluntly, utter shit designed for morons. Especially on laptops with no proper antennae.

I have NEVER gotten more than 10Mbps *actual transfer rate* out of a basic 2.4Ghz wifi link, even feet away from the router. Even when it said it was connected at 65Mbps or 72Mbps.

Remember wifi is half duplex., Every time you send an ack back, it stops the forward channel.

And if any other device is on the wlan, you are sharing the link speed with that, too.

It is worse than old coaxial Ethernet was at 10Mbps. It is to put it bluntly consumer crap for morons. Like StupidPhones?.

Using iwconfig I have watched connection rates and attenuation vary by

paul lee wrote: [snip speeds etc.]

FWIW here are some figures I just got copy a large file across my network to my Pi 'NAS':-

chris@esprimo$ scp 2020-08-20-raspios-buster-armhf-lite.zip backup:/bak/esprimo/cur/home/chris

As you can see I was using scp, the first copy is to an external USB3 hard drive, the second is to the Pi's SD card. As you can see the speed is identical which suggests to me that it's limited almost entirely by the network rather than the Pi's internals.

It's all Gigabit (I checked), out of 'esprimo' which is a desktop machine, via a switch near my desktop, along buried UTP to another switch in the garage and thence to the Pi.

Is that megabits or megaBytes per second?

I'm using a 4B with a USB 3.1 HD. The HD does about 100MB/s (megaBytes) read/write locally, and using Cyrstal Diskmark over Samba its showing maximum transfer rate of 72MB/s read and 58MB/s write. When backing up from my other Pis over NFS I'm seeing rates of 40-50MB/s.

Check you are getting 1000Mb/s Full duplex using the command:- ethtool eth0

Also try another Ethernet cable (at least Cat 5e), as it wasn't until I started using that Pi as a NAS did I find its upload speed was very poor. Up to then it was only used for web browsing and its download speed was fine. I think the cable had been kinked, replacing it restored the upload speed.

---druck

TN> you said gigabit Ethernet and now you are saying Wifi? TN> TN> Wifi is to put it bluntly, utter shit designed for morons. Especially on TN> laptops with no proper antennae. TN> TN> I have NEVER gotten more than 10Mbps *actual transfer rate* out of a TN> basic 2.4Ghz wifi link, even feet away from the router. Even when it TN> said it was connected at 65Mbps or 72Mbps. TN> TN> Remember wifi is half duplex., Every time you send an ack back, it stops TN> the forward channel. TN> TN> And if any other device is on the wlan, you are sharing the link speed TN> with that, too. TN> TN> It is worse than old coaxial Ethernet was at 10Mbps. It is to put it TN> bluntly consumer crap for morons. Like StupidPhones?. TN> TN> TN> Using iwconfig I have watched connection rates and attenuation vary by TN> 3:1 for no apparent reason whatsoever. Or simply stop working altogether TN> until reconnected. Yes, I have foil in all my walls and that makes for a TN> tricky wifi environment, but even so. TN>

Ok... I do appreciate this reply. :P Thanks.

So, yes... my daily driver machine is an older T430S Thinkpad laptop, probably with a less than current WiFi chip/card... however, I was running my NAS on a Pi 3; and just upgraded (actually still running both) to a Pi 4. My LAN/ethernet network is all gigabit+ hardware.

So, what I think I should do is simply plug my Thinkpad into the ethernet port and retest both the Pi 3 and the Pi 4 NAS systems and see what I get then. I, being a fairly versed an knowledgable NEWBIE, didn't release I should be 'happy' with 12-14mb over my laptops WiFi (which again, is probably less than current since the laptops I run are from 2012).

I'll connect over ethernet to both the Pi NAS systems and post again with the results, but... is this a fair and valid test that I should persue?

I thought I would get better speeds OVER that WiFi connection I'm speaking of, but... understand what you've stated here. :P I am, as you can tell, still learning..

THANK YOU.

|07p|15AULIE|1142|07o |08.........

CG> FWIW here are some figures I just got copy a large file across my CG> network to my Pi 'NAS':- CG> CG> chris@esprimo$ scp 2020-08-20-raspios-buster-armhf-lite.zip CG> backup:/bak/esprimo/cur/home/chris CG> 2020-08-20-raspios-buster-armhf-lite.zip 100% CG> 434MB 11.2MB/s 00:38 CG> chris@esprimo$ scp 2020-08-20-raspios-buster-armhf-lite.zip backup: CG> 2020-08-20-raspios-buster-armhf-lite.zip 100% CG> 434MB 11.2MB/s 00:38 CG> CG> As you can see I was using scp, the first copy is to an external USB3 CG> hard drive, CG> the second is to the Pi's SD card. As you can see the speed is CG> identical whichsuggests to me that it's limited almost entirely by the CG> network rather than thePi's internals. CG> CG> It's all Gigabit (I checked), out of 'esprimo' which is a desktop CG> machine, via a CG> switch near my desktop, along buried UTP to another switch in the garage CG> andthence to the Pi. CG> CG> -- CG> Chris Green

Thanks for your reply... I was told by another poster that I, since the laptop I use connects via wifi, should consider my 12mb speeds normal -

So I am going to connect said laptop to the ethernet connection and test both Pi's again. (pi 3 and pi 4...)

Thanks for your info; that is about what I'm getting from BOTH the Pi 3 and the Pi 4; which are both connected to the ethernet connection, but pulling TO my laptop Thinkpad T430s which is on its older WiFi chip/card.

Thanks again...

|07p|15AULIE|1142|07o |08.........

dr> I'm using a 4B with a USB 3.1 HD. The HD does about 100MB/s (megaBytes) dr> read/write locally, and using Cyrstal Diskmark over Samba its showing dr> maximum transfer rate of 72MB/s read and 58MB/s write. When backing up dr> from my other Pis over NFS I'm seeing rates of 40-50MB/s.

Are *all* of your computers connected via ethernet? I am connecting FROM a laptop on wifi, and was told that its not as fast/reliable... so; THATS probably my bottleneck. :P

dr> Check you are getting 1000Mb/s Full duplex using the command:- dr> ethtool eth0

On the Pi 3, which is connected via ethernet - I'm NOT. On the Pi 4, which is connected via ethernet - I AM.

dr> Also try another Ethernet cable (at least Cat 5e), as it wasn't until I dr> started using that Pi as a NAS did I find its upload speed was very dr> poor. Up to then it was only used for web browsing and its download dr> speed was fine. I think the cable had been kinked, replacing it restored dr> the upload speed. dr> dr> ---druck

I think I was just misinformed, since I was trying to get the higher speeds TO a Thinkpad T430s laptop connected via WiFi. I am going to connect said laptop TO the ethernet and then retest on both the Pi 3 and Pi 4 NAS systems.... I think this just might be me... user error.

:P

But, I wouldn't have known and would have been pulling out my hair had I not asked so... Hope this all makes sense to you, too; and I'm learning. I will post the results with the laptop CONNECTED TO ETHERNET shortly.

|07p|15AULIE|1142|07o |08.........

The above speeds are wrong, I think one of my switches was playing up, revised speeds as follows:-

Desktop to backup SD card:- bone-debian-9.4-console-armhf-2018-07-08-1gb.img 100% 850MB 20.6MB/s 00:41 bone-debian-9.4-console-armhf-2018-07-08-1gb.img 100% 850MB 21.3MB/s 00:39

Desktop to backup external USB3 hard drive:- bone-debian-9.4-console-armhf-2018-07-08-1gb.img 100% 850MB 47.3MB/s 00:17 bone-debian-9.4-console-armhf-2018-07-08-1gb.img 100% 850MB 45.3MB/s 00:18

Backup to desktop, from SD card:- bone-debian-9.4-console-armhf-2018-07-08-1gb.img 100% 850MB 36.9MB/s 00:23

Backup to desktop, from external USB3 hard drive:- bone-debian-9.4-console-armhf-2018-07-08-1gb.img 100% 850MB 39.2MB/s 00:21

So, getting on for half the theoretical speed over a Gigabit network in the best case.

Mmm. I get pretty close to my 100Mbps network against a linux server with NFS.

Do you find that the RPi3 acting as a Plex server is powerful enough to transcode recordings from the format in which you recorded them (MPEG TS or H264 TS) into whatever esoteric format Plex clients require?

I found that even a RPi4 gets very hot and runs at very high CPU % if it has to do any transcoding. Using the Plex client on a Roku box, it seems that it will play SD recordings (MPEG TS) natively but has to transcode HD recordings (H264 TS). I suppose this is an improvement over Windows, where it seems to transcode everything.

Why did Plex devise a client-server architecture where the client cannot play the files in their native form, but must instead get the server to transcode them? Is there a format that recordings could be converted to (offline) which allows them to be played without transcoding? What a faff having to do that for every file of my several TB of recordings...

At least on RPi4, VLC will run fast enough to play either SD or HD TS files, either from a local disk or over Ethernet/SMB. All I need to do is to get the sound to work - either via the analogue output or via the HDMI lead to my TV. I may end up binning PLex server and playing to the TV over VLC on the Pi - at least VLC can play whatever file format is available.

From desktop to/from laptop I too get something over 100MBps on my Gigabit network. The Pi4 (as can be seen above) is somewhat slower, but not hugely. ... and the Pi doesn't have a particularly fast disk, unlike my desktop and laptop which both have NVME SSDs.

CG> I had a T430 that died (very unusual for Lenovo T series), I now have CG> a T470 I bought used off eBay for rather less than I expected, lovely!

I love the ThinkPad hardware; and while the T430/T440 series have SOME of the cool stuff from the old days, they are just beginning to be a little long in the tooth for me. I'm not very hardware intensive, but... I will be looking at some other ThinkPad models in the future. I might just bite the bullet and go CURRENT T-series, but I haven't decided just yet.

CG> My WiFi connection reports that it is 300Mb/s but the real speed is CG> never anything like that. Here are my results sending from T470 CG> laptop (WiFi connection, reports as 300Mb/s) to desktop:- CG> CG> bone-debian-9.4-console-armhf-2018-07-08-1gb.img 100% CG> 850MB 16.2MB/s 00:52 CG> CG> So I get about half the 'expected' speed, also quite similar to your CG> speeds I think.

Understood... however, you are getting a LITTLE better speeds than me; I wonder what type of WiFi chip/card is in the T470 vs what is CURRENT??? Maybe it would be worth it, for me, to upgrade the WiFi chip/card in my T430s to the best it will take, OR whatever is current in 2021...

Even if I won't get the 100mb/s that I had THOUGHT, I'd probably see a better transfer rate anyway. Hmmmmm...

I get around 80mb down on average, with my Thinkpad...

|07p|15AULIE|1142|07o |08.........

Absolutely

I had a T430 that died (very unusual for Lenovo T series), I now have a T470 I bought used off eBay for rather less than I expected, lovely!

My WiFi connection reports that it is 300Mb/s but the real speed is never anything like that. Here are my results sending from T470 laptop (WiFi connection, reports as 300Mb/s) to desktop:-

bone-debian-9.4-console-armhf-2018-07-08-1gb.img 100% 850MB 16.2MB/s 00:52

So I get about half the 'expected' speed, also quite similar to your speeds I think.

On Tue, 12 Jan 2021 10:57:04 +0000, Chris Green declaimed the following:

Are you taking into account the IP header size, the TCP (or UDP) header size, and MTU size? The latter will tend to determine how many packets need to be sent (and for TCP, ACKed). Also, does your transfer method apply any sort of CRC or ECC logic, which will also consume some space in those packets?

On Tue, 12 Jan 2021 00:31:38 +1300, snipped-for-privacy@f420.n05.z1.binkp.net (paul lee) declaimed the following:

On Tue, 12 Jan 2021 00:35:31 +1300, snipped-for-privacy@f420.n05.z1.binkp.net (paul lee) declaimed the following:

As I recall, the R-Pi 3 Ethernet is internally a USB dongle, so throughput will be comparable to USB-2... On the Pi 4, which is connected via ethernet - I AM.

Real Ethernet on R-Pi 4

When I copy from laptop to desktop (both quite fast machines with fast disks) I get something quite a bit over 100MB/s on wired Gigabit connections. So the overhead isn't that great given that the theoretical maximum would be 1000/8 which is 125MB/s. So on a 300Mb/s wireless link between the same two machines one would, sort of, expect something a bit more than 30MB/s whereas in reality one gets about half of that.

Which starts to make more sense when you realise the link is essentially half-duplex and subject to noise induced retries.

Testing with ipferf, I get about 996MB/s, which is close to the maximum.

---druck