Major problems with home network, I'm desperate for help

[System Error Message]

When you use wifi did you see the wifi stats such as the link rate and protocol?

NUCs are small, so try plugging both server and client to the router itself and try a transfer over the router's own switch. Than do the same with the switch to find out where the problem is. If everything appears fine than problem is wifi. There could be overlapping signals such as a router's auto being terrible and always using the channels that you use even when you change it.

If your switch is d-link or tp-link it could also be a problem. D-link isnt really well known when it comes to reliability however if you find one of the devices to be at fault than perhaps using a compatible PSU (make sure they have the same output voltage but more output amps) could help.

Wifi is half duplex. If at the same radio one transmits to the other it has to go through the radio. Lets say both clients connect at the same rate of 800Mb/s, the transfer will even out (assuming there is a good radio chip) at 400Mb/s at best, otherwise you will get varying speeds that go up and down, mostly down. If the speeds arent the same rate than expect less than half the shown rate. Remember that wifi is a convenience and not performance orientated. It doesnt matter how many Gb/s your wifi can have it will never exceed the performance of gigabit ethernet. You should expect 60% of link speeds so instead of 866Mb/s you will get around 500Mb/s max so you're looking at about 250Mb/s average or best case per client. Since you get 20MB/s this means your wifi is transmitting at 160Mb/s per client from wifi to wifi transfers, not far off from practical rates.

Use wires where you can and like powerline, practical speeds are far from rated. With powerline you will get speeds closer to what you see but the sync speeds will be less than half of the rated speeds while with wifi being a half duplex medium, speeds will be much lower than expected.

Its actually possible to get 90% of the wifi speeds shown but not without significant packet losses just to fill up the medium.

 

[road hazard]

Hi, if you copy from a wired PC to a wireless device with 866Mbps max data rate then the Max teorethical throughput would be just 866/8 = 108 MByte/s, which mean a real speed of something more than half that data (or about 60MB/sec would be reachable) once we take out all the protocol induced overhead, but take into account that this is with the link speed established at 866... note that most times the link speed is negotiated slower due to interferences, distance, echo, lack of antennas/bands in one part. or just bad reception by one of the ends and then the real throughput gets even worst.

Regarding slowdowns induced by other devices connected but not transmitting, no, there should be no measurable slowdowns.

When you use wifi did you see the wifi stats such as the link rate and protocol?

NUCs are small, so try plugging both server and client to the router itself and try a transfer over the router's own switch. Than do the same with the switch to find out where the problem is. If everything appears fine than problem is wifi. There could be overlapping signals such as a router's auto being terrible and always using the channels that you use even when you change it.

If your switch is d-link or tp-link it could also be a problem. D-link isnt really well known when it comes to reliability however if you find one of the devices to be at fault than perhaps using a compatible PSU (make sure they have the same output voltage but more output amps) could help.

Wifi is half duplex. If at the same radio one transmits to the other it has to go through the radio. Lets say both clients connect at the same rate of 800Mb/s, the transfer will even out (assuming there is a good radio chip) at 400Mb/s at best, otherwise you will get varying speeds that go up and down, mostly down. If the speeds arent the same rate than expect less than half the shown rate. Remember that wifi is a convenience and not performance orientated. It doesnt matter how many Gb/s your wifi can have it will never exceed the performance of gigabit ethernet. You should expect 60% of link speeds so instead of 866Mb/s you will get around 500Mb/s max so you're looking at about 250Mb/s average or best case per client. Since you get 20MB/s this means your wifi is transmitting at 160Mb/s per client from wifi to wifi transfers, not far off from practical rates.

Use wires where you can and like powerline, practical speeds are far from rated. With powerline you will get speeds closer to what you see but the sync speeds will be less than half of the rated speeds while with wifi being a half duplex medium, speeds will be much lower than expected.

Its actually possible to get 90% of the wifi speeds shown but not without significant packet losses just to fill up the medium.

I still can't fathom why (when the NUCs are both hardwired into the router) my PC can push to one of them at roughly 100MB/s and the other at 50MB/s but when they're both hardwired into the router, they can push a 10 gig MKV back and forth at 90-100MB/s. This fact alone makes me want to put my head into the wall as it makes no sense at all. Can anyone offer a hypothesis on that?

I'm bringing a cross-over cable and a new 1 gig network card home tonight and will do some more testing.

 

[FTC]

Hi, not sure about this last one, but I would make some tests with the router port where the NUCs are. There could be some specific configuration for a port or priority that is affecting, and/or check /swap the ethernet cables used to connect the NUCs maybe one is not good and ends up producing errors and/or negotiating to half duplex or so.

 

[road hazard]

When you use wifi did you see the wifi stats such as the link rate and protocol?

NUCs are small, so try plugging both server and client to the router itself and try a transfer over the router's own switch. Than do the same with the switch to find out where the problem is. If everything appears fine than problem is wifi. There could be overlapping signals such as a router's auto being terrible and always using the channels that you use even when you change it.

If your switch is d-link or tp-link it could also be a problem. D-link isnt really well known when it comes to reliability however if you find one of the devices to be at fault than perhaps using a compatible PSU (make sure they have the same output voltage but more output amps) could help.

Wifi is half duplex. If at the same radio one transmits to the other it has to go through the radio. Lets say both clients connect at the same rate of 800Mb/s, the transfer will even out (assuming there is a good radio chip) at 400Mb/s at best, otherwise you will get varying speeds that go up and down, mostly down. If the speeds arent the same rate than expect less than half the shown rate. Remember that wifi is a convenience and not performance orientated. It doesnt matter how many Gb/s your wifi can have it will never exceed the performance of gigabit ethernet. You should expect 60% of link speeds so instead of 866Mb/s you will get around 500Mb/s max so you're looking at about 250Mb/s average or best case per client. Since you get 20MB/s this means your wifi is transmitting at 160Mb/s per client from wifi to wifi transfers, not far off from practical rates.

Use wires where you can and like powerline, practical speeds are far from rated. With powerline you will get speeds closer to what you see but the sync speeds will be less than half of the rated speeds while with wifi being a half duplex medium, speeds will be much lower than expected.

Its actually possible to get 90% of the wifi speeds shown but not without significant packet losses just to fill up the medium.

So I got some tonight and used a cross-over cable between the server and my PC down here in the basement and the result was roughly 50MB/s. I installed USB 3.0 to Ethernet adapters in both and re-did the test with the cross over cable connected to the adapters, same result, crap speed. I tested the adapters before I left work and hooked up to our network at my job, I was copying a 20+ gig file and maintained a speed of roughly 105MB/s so I know the adapters are good.

Tried booting the server and my machine in safe mode with networking and re-did the test, same crap speed.

Heck, now, the only time >MY< PC can sustain a fast transfer is when it's talking to NUC # 1 (hardwired) and then I see the usual 120MB/s when copying large files. (Speaking of which, will have to re-try that test because who knows, maybe that's crap speed now as well.) This just gets more and more bizarre by the day. I can't think of anything else to try. Is every computer in my house suffering some sort of motherboard damage or severe electronic interference of some kind?!?!?!?

I'm on the verge of giving up and living with 1/2 my old speed.

 

[System Error Message]

So I got some tonight and used a cross-over cable between the server and my PC down here in the basement and the result was roughly 50MB/s. I installed USB 3.0 to Ethernet adapters in both and re-did the test with the cross over cable connected to the adapters, same result, crap speed. I tested the adapters before I left work and hooked up to our network at my job, I was copying a 20+ gig file and maintained a speed of roughly 105MB/s so I know the adapters are good.

Tried booting the server and my machine in safe mode with networking and re-did the test, same crap speed.

Heck, now, the only time >MY< PC can sustain a fast transfer is when it's talking to NUC # 1 (hardwired) and then I see the usual 120MB/s when copying large files. (Speaking of which, will have to re-try that test because who knows, maybe that's crap speed now as well.) This just gets more and more bizarre by the day. I can't think of anything else to try. Is every computer in my house suffering some sort of motherboard damage or severe electronic interference of some kind?!?!?!?

I'm on the verge of giving up and living with 1/2 my old speed.

what CPUs and storage do your NICs use? CPU has to be fast enough for NAS and storage has to be fast enough for that throughput. Some flash storage will either be slow around 10Mb/s or around 40-50MB/s or 100MB/s or more. Hard drive speeds vary but recent ones will have no issues saturating half a 1 Gb/s link ( gigabit ethernet is full duplex so is capable of 2Gb/s).

 

[L&LD]

If your 'normal' was up around 120MB/s, that suggests you are (or were) using jumbo frames. If not everything on your network supports the JF value(s) you are using (and it seems on only 'some' equipment), then you'll have issues like this.

http://rickardnobel.se/actual-throughput-on-gigabit-ethernet/

Using USB 3.0 Ethernet adaptors is also throwing another wrench into the mix. They have their own unique overhead, over and above what Ethernet introduces into the link path.

What NUC's are you using? What is the ram and cpu in them? What ssd are you running and are you using any ssd utilities to enhance that speed from them?

Have you changed any of the default NIC parameters? Or any of the default os networking parameters? What os are you running? Was this an upgrade or a clean install?


In addition to all the above, the actual files you're transferring matter too. If your tests haven't been using the same identical (set) of files, then your comparisons are not really trustworthy.

 

[road hazard]

Update for everyone........ and interesting turn of events.

My server has a SSD boot drive (C and 2, 4TB drives in RAID 1 (D and 2, 1.5TB drives in RAID 1 (E. In the vast majority of my testing, I was reading .MKV movie files from my server's D: drive to my PC in the basement and the NUCs upstairs. The average was around 60MB/s a lot of the time, never getting back to the normal 120MB/s. Every now and then, I would copy from my PC to NUC 1 and 2 and NUC 1 was fast, NUC 2, slow. No matter what, I was never able to pull a big .MKV file from my server's D:\Movies folder at more than 60MB/s to any PC in my house. (Even tried a cross over cable, direct to my PC.) A slow movie copy from the server was a common constant.)

So today, via a friends suggestion, I copied the big .MKV file to my SSD boot drive on my server and from THAT drive, I did the same copy. Low and behold, I was able to copy to NUC # 1 at around 110MB/s the entire time. This is the FIRST time that the server has hit that speed since all this started. But, a slight fly in the ointment. Doing the same test to NUC # 2 (all hard wired)...same crappy 60MB/s speed. I'm not too worried about the speed to NUC # 2, will dig in to that later.....maybe its' SSD is acting up or I need to make sure Windows caching is active for that drive (but a SSD should be able to keep up with the writes, shouldn't it? But I digress, this issue can wait.)

When I get home, I'm going to do the same copy from the server's SSD to my PC in the basement. If that is fast, which I'm hoping/assuming it will be, it's definitely the RAID at the heart of my probem. As of right now, the RAID is looking very guilty.

Something else I discovered...... this is another weird issue;

Copying from SSD to D:=fast (120MB/s, maxing out mechanical drive
Copying from D: to SSD=slow (60MB/s)

Copying from SSD to E:=slow (60MB/s)
Copying from E: to SSD=fast (120MB/s)

I really think this is no longer a network issue but a RAID one. I'm using my motherboard ICH10R for the mirrors. According to the Intel Storage utility, the arrays are both fine. SMART tests come back AOK. There any RAID experts here?

 

[L&LD]

Try defragging the drive array completely first, before you mess with the RAID array setup.

 

[john9527]

I'm not too worried about the speed to NUC # 2, will dig in to that later.....maybe its' SSD is acting up or I need to make sure Windows caching is active for that drive (but a SSD should be able to keep up with the writes, shouldn't it? But I digress, this issue can wait.)

Make sure that you have trim support set up correctly.

 

[road hazard]

BANGING HEAD INTO WALL

Just when I was breaking out the champaign to celebrate.......... got home tonight and copied that same test from the my server's SSD to my PC and........... 60MB/sec. Copying it to NUC # 1, 100+ MB/s the entire time. I can dual boot between Windows 7 on my PC down here by the server and Linux Mint and yeah....... the server can push to NUC # 1 at high speed (from its' SSD drive) but to my PC, 60-65MB/s. What in the world is going on?!?!?!?!??!

So MY PC can copy to NUC # 1 at high speed, the server can copy to NUC # 1 at high speed but my PC and the server...60-65MB/s. All testing was done from the server's SSD. Again, none of this makes sense and I'm at a loss.

The RAID arrays on the server are definitely a problem but this weird speed issue between different systems is illogical and driving me crazy. Other the replacing every single PC, router, switch and network cable in my house..... any other ideas?

 

[L&LD]

Curious, how full are your ssd's on all computers?

 

[john9527]

Curious, how full are your ssd's on all computers?

I think that is a VERY good question

 

[road hazard]

Curious, how full are your ssd's on all computers?

Server is a 120 gig SSD with about 40 gigs free.
My PC beside it is a 480 gig SSD with 200 gigs free.
NUC 1 and NUC 2 are both 256 gig SSD with about 120 gigs free on both.

More clues for everyone.... on my server and NUC # 1, I noticed that in Device Manager/SSD hard drive/Policies.... "Enable Write Caching" was enabled. Below it, the "Turn off Windows write-cache buffer flushing on the device" was not checked. I mimicked those same setting on my PC and NUC # 2. The result.... I can routinely max out my connection when copying a file from my server (from the server's SSD....the slow RAID reads are a battle for another day) to my PC and from my PC to NUC # 1. This is progress. Unfortunately, I think I'm just masking some other problem. I remember turning that feature off a long time ago and have no idea how NUC # 1 got it enabled. Do you guys have that turned off or on? Can you do speed tests with a big file to see what the results are?

I had a friend who had that option active, he turned it off on the target machine and copied a 4 gig ISO file and was able to maintain over 100MB/s. For me, using it brings me back to the speeds I'm use to...........except on NUC # 2. Once I turned it on for that one, instead of the 60MB/s I'm use to seeing, the transfer rate started nice but quickly fell to a steady 80MB/s. Better but still not a rock solid 110+/-MB/s.

This picture is from Taskmanager on my PC while copying a file to NUC # 2 (Write cache stuff enabled on the NUC)

......as you can see, network utilization bounces around the 80% mark. Xfer rate was AROUND 70-80MB/s the entire time. When I turned off the write caching, the xfer rate was around 60MB/s, if that. So that caching option helped but maybe is masking something?! My SSD on the NUC failing?

And this picture is from Taskmanager on the SSD as I first copy a file to NUC 2 (caching enabled) then the same file to my PC right after it:

....with write caching enabled for my PC, it maxed out the connection. Turn off that write caching SSD option, back to 50-60MB/s.

The mystery continues down the weird-o path.

 

[sfx2000]

as you can see, network utilization bounces around the 80% mark

Not seeing a problem here - that's what is supposed to happen...

 

[john9527]

Just to satisfy my curiosity....what's the response to

fsutil behavior query disabledeletenotify

on each of the systems

 

[road hazard]

Not seeing a problem here - that's what is supposed to happen...

When the server is copying a file from it's SSD to my PC (with SSD caching enabled), the network utilization is 99% the entire time. Can you do a multi-gig file transfer in between systems at your house and post a screen shot like that? Just curious what it looks like on another system.

 

[road hazard]

Just to satisfy my curiosity....what's the response to

fsutil behavior query disabledeletenotify

on each of the systems

Result is zero on all systems..........trim enabled.

 

[L&LD]

Sorry if you've already posted it, but what version of Windows are you running?

Are you running anything like Intel Storage Manager (or the like)?

Is everything up to date?

 

[L&LD]

I also think that you are slowly killing your ssd's (at least their performance) by repeatedly testing them like this (for no real benefit).

That 120GB ssd in the server is suspect (much too small to give good performance, consistently).