You are not logged in.

#1 2020-04-19 06:20:04

xenobro
Member
Registered: 2019-05-25
Posts: 33

Only fast, speedtest-cli, pacman and curl get good speeds

Hi everyone.
I have a "Positivo 2 em 1 Q432A" pc, motherboard CHT12CP, quad core, 4 GB of RAM, running a pretty stripped down and minimalistic instance of Arch on a USB 3.0 pendrive, all programs updated to its latest versions, and I'm connected to the net through a TP-Link TP-UE300 USB-to-Ethernet Gigabit adapter, with 300 Mbps download and 150 Mbps upload as my nominal speeds, using Mitrastar GPT-2741GNAC-N2 as my "fiber modem".

My issue is that while I ALWAYS get 295+ Mbps download on the fast cli program and 300+ Mbps dl and 150+ ul on speedtest-cli, I only get "compatible" download speeds of 35-36+ MB/sec (i.e. 280-288+ Mbps) on pacman or curl - and that is on a specific server, though lately it seems that, after the first times I download a file from them, when I try it again, it'll be full speed for some 10-20 secs, then go down fast to 2 or 3 MB/sec then climb back up to 9 or 10 MB/sec and stay there for the rest of the download. That could be some sort of server protection that I've unwittingly triggered, since I've had to download the largest single file on Arch's repositories (cuda) a huge number of times to test this with pacman, curl, wget, Firefox, Chromium, etc. Is that even possible?

Anyway, I can't get these speeds on any other sites (e.g. downloading Ubuntu's, Arch's or Debian's isos) or programs. Every time I try, one of these will happen:

1. The download speed will climb fast, then "hit a ceiling" that could be anywhere between 24-36 MB/sec, and then immediately fall even faster to 2 or 3 MB/sec, climb back to anything between 6 to 12 MB/sec and stay there for the rest of the transference;
2. Like the previous one, but when it "hits the ceiling", it starts to "stutter" or get "mini-freeze-ups" for a few secs (inside the program only, all numbers freeze for a few secs, but the mouse, keyboard, etc. are not affected) and then fall; in these cases, the final speeds will be slower, betweem 3 to 5 MB/secs;
3. The speeds will stay between 9-12 MB/sec all the way through;

It might be worth mentioning that every first time I download something with Firefox, everything (this time including mouse, keyboard, etc.) is frozen for some 40 secs then continues as if nothing happened - and it's clear from the download progress that the pc was getting the file while everything was frozen at the measured speed. After that I can get the same file or any other and there'll be no "freezings" whatsoever. Also, depending on the speedtest or download I'm doing on Firefox, I hear some "clicks", but it seems to be only on FF.

Finally, there's rtorrent's case: on a popular download, the speeds will climb REALLY fast, until it's measuring 24-27 MB/sec, but from the changing numbers on downloaded file size, it's actually about 32-40 MB/sec, but the "speedometer" hasn't caught up with it yet. Then, everything inside the program will start to freeze for a few secs a number of times, and the speeds will fall below 5 MB/sec, afterwards climbing back up to anything between 9 and 12 MB/sec, but never go beyond this. If it somehow crosses that threshold again, the "fast fall" will happen again and the "ceiling" speeds will be even lower. Also, it'll get extremely slow to respond to commands after that.

The best metaphor I can get for the really fast speed fall is this: imagine the download speeds are a tennis ball; you take it and throw it up with all your strength, then it hits the ceiling (and it may or may not be stuck there for a few secs). Now it somehow comes back down not only with all the strength you threw it up, but with the added force of gravity - that's how fast and how much the speed will fall.

Using TP-UE300 on USB 3.0 or 2.0 doesn't make any difference, nor does the value of net.ipv4.tcp_ecn. Also, on Windows 10, MS Edge tends to get better dl speeds than Firefox, and the Speedtest app also always gets 300+ Mbps dl and 150+ Mbps ul.

Does anyone have a clue what's happening here? Could it be that my pc can't handle these dl speeds? But if that was the case, shouldn't the fast program and speedtest-cli have issues getting those speeds? And why does pacman and curl get max speeds from a specific server but no other program can?

Here's lshw and, if there's anything I should post, please just tell me:

[x@archlinux ~]$ sudo lshw
[sudo] senha para root: 
archlinux                   
    description: Notebook
    product: CHT12CP (3601127)
    vendor: Positivo Tecnologia SA
    version: Tablet
    serial: 4A675V90H
    width: 4294967295 bits
    capabilities: smbios-3.1 dmi-3.1 smp vsyscall32
    configuration: boot=normal chassis=notebook family=CHT12CP sku=3601127 uuid=0C021ECE-9052-0244-8EEA-9A5C1ECE9052
  *-core
       description: Motherboard
       product: CHT12CP
       vendor: Positivo Tecnologia SA
       physical id: 0
       version: 11130211
       serial: 4A675V90H
       slot: family
     *-firmware
          description: BIOS
          vendor: Positivo Tecnologia SA
          physical id: 0
          version: V00.06.X
          date: 09/21/2017
          size: 64KiB
          capacity: 3008KiB
          capabilities: pci upgrade shadowing cdboot bootselect edd int9keyboard int14serial int17printer int10video acpi usb zipboot biosbootspecification netboot uefi
     *-cpu
          description: CPU
          product: Intel(R) Atom(TM) x5-Z8350  CPU @ 1.44GHz
          vendor: Intel Corp.
          physical id: 4
          bus info: cpu@0
          version: Intel(R) Atom(TM) x5-Z8350  CPU @ 1.44GHz
          serial: To Be Filled By O.E.M.
          slot: CHV
          size: 941MHz
          capacity: 1920MHz
          width: 64 bits
          clock: 83MHz
          capabilities: x86-64 fpu fpu_exception wp vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp constant_tsc arch_perfmon pebs bts rep_good nopl xtopology tsc_reliable nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 movbe popcnt tsc_deadline_timer aes rdrand lahf_lm 3dnowprefetch epb pti ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid tsc_adjust smep erms dtherm ida arat md_clear cpufreq
          configuration: cores=4 enabledcores=4 threads=4
        *-cache:0
             description: L1 cache
             physical id: 6
             slot: L1 Cache
             size: 32KiB
             capacity: 32KiB
             capabilities: synchronous internal write-back instruction
             configuration: level=1
        *-cache:1
             description: L2 cache
             physical id: 7
             slot: L2 Cache
             size: 1MiB
             capacity: 1MiB
             capabilities: synchronous internal write-back unified
             configuration: level=2
     *-cache
          description: L1 cache
          physical id: 5
          slot: L1 Cache
          size: 24KiB
          capacity: 24KiB
          capabilities: synchronous internal write-back data
          configuration: level=1
     *-memory
          description: System Memory
          physical id: 17
          slot: System board or motherboard
          size: 4GiB
        *-bank
             description: DIMM DDR3 Synchronous 1066 MHz (0,9 ns)
             vendor: 0000
             physical id: 0
             serial: 00000000
             slot: ChannelA-DIMM0
             size: 4GiB
             width: 16 bits
             clock: 1066MHz (0.9ns)
     *-pci
          description: Host bridge
          product: Atom/Celeron/Pentium Processor x5-E8000/J3xxx/N3xxx Series SoC Transaction Register
          vendor: Intel Corporation
          physical id: 100
          bus info: pci@0000:00:00.0
          version: 36
          width: 32 bits
          clock: 33MHz
          configuration: driver=iosf_mbi_pci
          resources: irq:0
        *-display
             description: VGA compatible controller
             product: Atom/Celeron/Pentium Processor x5-E8000/J3xxx/N3xxx Integrated Graphics Controller
             vendor: Intel Corporation
             physical id: 2
             bus info: pci@0000:00:02.0
             version: 36
             width: 64 bits
             clock: 33MHz
             capabilities: pm msi vga_controller bus_master cap_list rom
             configuration: driver=i915 latency=0
             resources: irq:117 memory:90000000-90ffffff memory:80000000-8fffffff ioport:1000(size=64) memory:c0000-dffff
        *-multimedia
             description: Multimedia controller
             product: Atom/Celeron/Pentium Processor x5-E8000/J3xxx/N3xxx Series Imaging Unit
             vendor: Intel Corporation
             physical id: 3
             bus info: pci@0000:00:03.0
             version: 36
             width: 32 bits
             clock: 33MHz
             capabilities: pm msi bus_master cap_list
             configuration: driver=intel_atomisp2_pm latency=0
             resources: irq:255 memory:91000000-913fffff
        *-generic:0
             description: Signal processing controller
             product: Atom/Celeron/Pentium Processor x5-E8000/J3xxx/N3xxx Series Power Management Controller
             vendor: Intel Corporation
             physical id: b
             bus info: pci@0000:00:0b.0
             version: 36
             width: 64 bits
             clock: 33MHz
             capabilities: msi pm bus_master cap_list
             configuration: driver=proc_thermal latency=0
             resources: irq:167 memory:91818000-91818fff
        *-usb
             description: USB controller
             product: Atom/Celeron/Pentium Processor x5-E8000/J3xxx/N3xxx Series USB xHCI Controller
             vendor: Intel Corporation
             physical id: 14
             bus info: pci@0000:00:14.0
             version: 36
             width: 64 bits
             clock: 33MHz
             capabilities: pm msi xhci bus_master cap_list
             configuration: driver=xhci_hcd latency=0
             resources: irq:119 memory:91800000-9180ffff
           *-usbhost:0
                product: xHCI Host Controller
                vendor: Linux 5.6.4-arch1-1 xhci-hcd
                physical id: 0
                bus info: usb@1
                logical name: usb1
                version: 5.06
                capabilities: usb-2.00
                configuration: driver=hub slots=7 speed=480Mbit/s
              *-usb:0
                   description: Generic USB device
                   product: USB 10/100/1000 LAN
                   vendor: TP-LINK
                   physical id: 2
                   bus info: usb@1:2
                   version: 30.00
                   serial: 000001000000
                   capabilities: usb-2.10
                   configuration: driver=r8152 maxpower=180mA speed=480Mbit/s
              *-usb:1
                   description: Video
                   product: USB 2.0 Web Camera
                   vendor: Alcor Micro, Corp.
                   physical id: 3
                   bus info: usb@1:3
                   version: 0.08
                   capabilities: usb-2.00
                   configuration: driver=uvcvideo maxpower=200mA speed=480Mbit/s
              *-usb:2
                   description: Keyboard
                   product: USB KEYBOARD
                   vendor: HAILUCK CO.,LTD
                   physical id: 4
                   bus info: usb@1:4
                   version: 1.00
                   capabilities: usb-1.10
                   configuration: driver=usbhid maxpower=100mA speed=1Mbit/s
           *-usbhost:1
                product: xHCI Host Controller
                vendor: Linux 5.6.4-arch1-1 xhci-hcd
                physical id: 1
                bus info: usb@2
                logical name: usb2
                version: 5.06
                capabilities: usb-3.00
                configuration: driver=hub slots=6 speed=5000Mbit/s
              *-usb
                   description: Mass storage device
                   product: Ultra Fit
                   vendor: SanDisk
                   physical id: 1
                   bus info: usb@2:1
                   version: 1.00
                   serial: 4C530001190318108265
                   capabilities: usb-3.00 scsi
                   configuration: driver=usb-storage maxpower=896mA speed=5000Mbit/s
        *-generic:1
             description: Encryption controller
             product: Atom/Celeron/Pentium Processor x5-E8000/J3xxx/N3xxx Series Trusted Execution Engine
             vendor: Intel Corporation
             physical id: 1a
             bus info: pci@0000:00:1a.0
             version: 36
             width: 32 bits
             clock: 33MHz
             capabilities: pm msi bus_master cap_list
             configuration: driver=mei_txe latency=0
             resources: irq:166 memory:91700000-917fffff memory:91600000-916fffff
        *-isa
             description: ISA bridge
             product: Atom/Celeron/Pentium Processor x5-E8000/J3xxx/N3xxx Series PCU
             vendor: Intel Corporation
             physical id: 1f
             bus info: pci@0000:00:1f.0
             version: 36
             width: 32 bits
             clock: 33MHz
             capabilities: isa bus_master cap_list
             configuration: driver=lpc_ich latency=0
             resources: irq:0
  *-battery
       description: Lithium Ion Battery
       product: SR Real Battery
       vendor: Intel SR 1
       physical id: 1
       version: Date
       serial: 123456789
       slot: I2C2
       configuration: voltage=3,8V
  *-power UNCLAIMED
       description: OEM Define 1
       product: OEM Define 5
       vendor: OEM Define 2
       physical id: 2
       version: OEM Define 6
       serial: OEM Define 3
       capacity: 42mWh
  *-scsi
       physical id: 3
       bus info: scsi@0
       logical name: scsi0
       capabilities: scsi-host
       configuration: driver=usb-storage
  *-network:0
       description: Ethernet interface
       physical id: 4
       logical name: enp0s20u2
       serial: d0:37:45:6d:9a:b0
       size: 1Gbit/s
       capacity: 1Gbit/s
       capabilities: ethernet physical tp mii 10bt 10bt-fd 100bt 100bt-fd 1000bt 1000bt-fd autonegotiation
       configuration: autonegotiation=on broadcast=yes driver=r8152 driverversion=v1.11.11 duplex=full firmware=rtl8153a-3 v2 02/07/20 ip=192.168.0.86 link=yes multicast=yes port=MII speed=1Gbit/s
  *-network:1 DISABLED
       description: Wireless interface
       physical id: 5
       logical name: wlan0
       serial: 58:b3:fc:cb:1e:12
       capabilities: ethernet physical wireless
       configuration: broadcast=yes driver=rtl8723bs driverversion=5.6.4-arch1-1 multicast=yes wireless=unassociated

Thank you very much in advance,
xenobro.

Offline

#2 2020-04-19 06:55:28

seth
Member
Registered: 2012-09-03
Posts: 16,250

Re: Only fast, speedtest-cli, pacman and curl get good speeds

Disk I/O?

Did you try comparing the wget/curl download of a largefile to /dev/null and ~/test.img?

Offline

#3 2020-04-20 08:13:52

xenobro
Member
Registered: 2019-05-25
Posts: 33

Re: Only fast, speedtest-cli, pacman and curl get good speeds

Hi, seth, thank you for your reply. I tried what you suggested and here are the results:

curl -o /dev/null http://archlinux.c3sl.ufpr.br/community/os/x86_64/cuda-10.2.89-5-x86_64.pkg.tar.zst

Doing this 10 times in a row and at other moments, the lowest speed was 33.3 MB/sec, but mostly between 35.5 and 36.3 MB/sec, so that recent issue of the speed dropping some 10 or 20 seconds in after the first download seems to be related. So I checked the disk i/o with the command below:

iozone -I -s2g -r1024k -i0 -i1

Write/rewrite speeds can go from 9 to 61 MB/sec, but usually between 10 to 20 MB/sec, so this could be the bottleneck of this specific issue. Any idea why could there be so much variation here? Read/reread speeds are always 115-123 MB/sec. Also, changing 2GB to 1GB makes no difference.

Anyway, the main issue remains - downloading to /dev/null or RAM seemed to have no effect on the other cases:

wget http://archlinux.c3sl.ufpr.br/community/os/x86_64/cuda-10.2.89-5-x86_64.pkg.tar.zst -O /dev/null

Here, as usual, the speeds were all over the place: from 0 to 32.8 MB/sec, down to 5, back up to 11, climbing up to 15, coming down to 8 and so on, but mostly 10-11 MB/sec with some random climbings to 15 followed by drops to 6 or 7, before coming back up to 10 or 11. One thing I noticed is that the "higher" the max (i.e. 27-32 MB/sec), the lower the min (i.e. 0-2 MB/sec), before stabilizing at ~11 MB/sec. For instance, if it only went up to 15 MB/sec, the drop would be to 7 or 8 MB/sec before coming back to 11 MB/sec.

And Firefox saving to my home folder (on RAM) gets exactly the same results as wget saving to /dev/null.

So, what could be causing these variations and low speeds?

As an aside, I also decided to test other servers from my mirrorlist and confirmed that I can only get 35+ MB/sec speeds from this one specific server. Does anyone get these speeds from any other server that they could suggest to me? If it was from South America, all the better, but from anywhere would be nice.

Thank you very much again,
xenobro.

Last edited by xenobro (2020-04-20 08:27:14)

Offline

#4 2020-04-20 13:33:06

seth
Member
Registered: 2012-09-03
Posts: 16,250

Re: Only fast, speedtest-cli, pacman and curl get good speeds

https://gist.github.com/raulmoyareyes/3 … c93be64746

The 100MB linode files should have "1GB" variants
Random arch mirrors are generally not very good test servers because you don't know what you get and the server performance could be impacted by the current overall load.

However the difference seems to be between wget and curl?
(You suggest to have downloaded the same file from the same server into /dev/null but curl outperforming wget 3:1, correct?)

Offline

#5 2020-04-21 23:00:07

xenobro
Member
Registered: 2019-05-25
Posts: 33

Re: Only fast, speedtest-cli, pacman and curl get good speeds

Hi, seth, and thank you once again for your help.
As you mentioned, the main problem here is curl/pacman outperforming wget/firefox 3:1 (and sometimes even 6:1), but also the fact that I can only get these speeds from one specific server on the entire Internet - trying to get ubuntu, debian or even arch isos, for instance, or even Win10 from the official MS site never gets these dl speeds, and I'm pretty sure that, in these cases, they should. I mean, can you test and see if you get more than 30 MB/sec dl'ing any of those?
As for the Arch servers, I used that one to measure it because it's pretty much the only place where I can get the full dl speed, but I'm also looking for new ones with good bandwidths, since most servers in the brazilian mirrorlist are public universities, and those have been having their funds severely cut by the federal gov't, so they might not even exist as servers for much longer, or will have to cut down expenses to a minimum, which might lead to dl speed decreases to make less machines serve the same amount of people, since they're having to cut down even on the electricity bills.
But anyway, I went to the site you suggested and did every test there was there. As you will see below, strangely only the Cachefly main site got me the full dl speeds, and for a change, that happened on curl, wget and even Firefox.
Most other tests had about the same results with wget and curl, but with lower speeds to both. Also, the variation in many of those was huge: the same test would get 0.3 MB/sec throughout its duration, and then would go on to get 20 MB/sec for the next 3 times, then go back to 0.4 MB/sec.
Some things I noticed on all dl tests except Cachefly: the speed up was generally slow; the first time I did a test was almost always the worst - by a factor of 10x to 100x, like 0.2 MB/sec in the 1st test to 20 MB/sec in the second); wget seemed to fail a lot with ipv6, which would never happen to curl.
Another thing is that my ul speeds seem to be, in general, much, much slower than they should be: speedtest-cli, fast and speedtest on Firefox always get me 150+ Mbps, so I should be getting some 16+ MB/sec, but even on rtorrent I've never gotten more than 2 MB/sec, and from the tests at the end of this post, you can see that, many times it's even lower than that - only the last test gave me a reasonable but still very slow result.
Finally, each of those tests was performed 3-10 times in a row; the first result is the lowest final result speed gotten from all of those, and the second, the highest final result. "Failed" means the site has 404'd unless stated otherwise.

Once again, thank you very much in advance,
xenobro.

BENCHMARKING
Here, results were almost the same for both wget and curl: among the many servers, 4.87-13.01 MB/sec dl, 0.37-7.26 MB/sec ul, only North American ramnode servers and one from vortexservers responded, all others giving me 0 MB/sec dl and ul, CPU reported as 4 x Intel(R) Atom(TM) x5-Z8350 CPU @ 1.44GHz, time to generate PI to 5000 decimal places with a single thread ~0m51s and IO test ~611 MB/sec (home folder on RAM).

DOWNLOAD
Cachefly:

wget -O /dev/null http://cachefly.cachefly.net/100mb.test

35.3 MB/sec

curl -o /dev/null http://cachefly.cachefly.net/100mb.test

34.4 MB/sec
Here, even Firefox got max speeds, since I couldn't even measure it, because the 100 MB dl was was over in less than 1 sec.

wget -O /dev/null http://speedtest.dal01.softlayer.com/downloads/test100.zip
curl -o /dev/null http://speedtest.dal01.softlayer.com/downloads/test100.zip

Failed

wget -O /dev/null http://speedtest.sea01.softlayer.com/downloads/test100.zip
curl -o /dev/null http://speedtest.sea01.softlayer.com/downloads/test100.zip

13.8-16.0 MB/sec

wget -O /dev/null http://speedtest.ams01.softlayer.com/downloads/test500.zip

Up to 13 MB/sec, down to 18 KB/sec, sometimes maxing until the end, sometimes at the minimum all the way through.

curl -o /dev/null http://speedtest.ams01.softlayer.com/downloads/test500.zip

Slowly up to 12-13 MB/sec, sometimes dropping to 9.9 MB/sec.

wget -O /dev/null http://50.23.64.58/downloads/test100.zip

14.2-16.4 MB/sec

curl -o /dev/null http://50.23.64.58/downloads/test100.zip

14.7-16.0 MB/sec

wget -O /dev/null http://speedtest.wdc01.softlayer.com/downloads/test500.zip

Fails on ipv6, 22.8-24.7 MB/sec on ipv4

curl -o /dev/null http://speedtest.wdc01.softlayer.com/downloads/test500.zip

21.1-22.7 MB/sec

wget http://speedtest.wdc01.softlayer.com/downloads/test10.zip > /dev/null

Fails on ipv6, 2.73-2.75 MB/sec on ipv4 - seemingly not enough time to reach higher speeds

curl -O http://speedtest.wdc01.softlayer.com/downloads/test10.zip > /dev/null

2399-2540 KB/sec on ipv4 - seemingly not enough time to reach higher speeds


wget --output-document=/dev/null http://speedtest.wdc01.softlayer.com/downloads/test500.zip

21.5-24.6 MB/sec

curl --output /dev/null http://speedtest.wdc01.softlayer.com/downloads/test500.zip

21.5-21.9 MB/sec

echo `curl --progress-bar -w "%{speed_download}" http://speedtest.wdc01.softlayer.com/downloads/test10.zip -o test.zip|sed 's/....$//'`/
131072|bc|xargs -I {} echo {} MB/sec

14 MB/sec

Linode:

wget -O /dev/null http://speedtest.tokyo.linode.com/100MB-tokyo.bin
curl -o /dev/null http://speedtest.tokyo.linode.com/100MB-tokyo.bin

Failed

wget -O /dev/null http://speedtest.london.linode.com/100MB-london.bin

10.8-11.6 MB/sec

curl -o /dev/null http://speedtest.tokyo.linode.com/100MB-tokyo.bin

10.9-11.8 MB/sec

wget -O /dev/null http://speedtest.newark.linode.com/100MB-newark.bin

12.2-17.8 MB/sec

curl -o /dev/null http://speedtest.newark.linode.com/100MB-newark.bin

17.3-17.6 MB/sec

wget -O /dev/null http://speedtest.atlanta.linode.com/100MB-atlanta.bin

11.0-18.8 MB/sec, mostly 18+

curl -o /dev/null http://speedtest.atlanta.linode.com/100MB-atlanta.bin

11.6-18.7 MB/sec, mostly 18+

wget -O /dev/null http://speedtest.dallas.linode.com/100MB-dallas.bin

2.15-16.9 MB/sec, mostly 16+

curl -o /dev/null http://speedtest.dallas.linode.com/100MB-dallas.bin

1.4-16.5 MB/sec

wget -O /dev/null http://speedtest.fremont.linode.com/100MB-fremont.bin

5.8-13.0 MB/sec

curl -o /dev/null http://speedtest.fremont.linode.com/100MB-fremont.bin

11.3-11.7 MB/sec

wget -O /dev/null http://mirror.nl.leaseweb.net/speedtest/1000mb.bin

11.8-12.9 MB/sec

curl -o /dev/null http://mirror.nl.leaseweb.net/speedtest/1000mb.bin

12.6-12.8 MB/sec

wget -O /dev/null http://mirror.us.leaseweb.net/speedtest/1000mb.bin

12.0-17.7 MB/sec

curl -o /dev/null http://mirror.us.leaseweb.net/speedtest/1000mb.bin

11.2-15.0 MB/sec

wget -O /dev/null http://lg.denver.fdcservers.net/100MBtest.zip
curl -o /dev/null http://lg.denver.fdcservers.net/100MBtest.zip

Failed

wget -O /dev/null http://proof.ovh.net/files/100Mb.dat

3.57-4.90 MB/sec - seemingly not enough time to reach higher speeds

curl -o /dev/null http://proof.ovh.net/files/100Mb.dat

2916-4078 KB/sec - seemingly not enough time to reach higher speeds

UPLOAD

iperf3 -c iperf.scottlinux.com -u
iperf3 -c iperf.volia.net -R -P 4

Failed

iperf -c iperf.volia.net -r -P 4

267 Kbps

iperf -c s-network1.amcs.tld -P 1 -i 5 -p 5999 -f B -t 60 -T 1
iperf -c iperf.acd.net -P 10 -t 240

Failed

iperf -c ping.online.net -i 2 -t 20 -r

Failed - connection refused

bwctl -T iperf3 -f m -t 10 -i 1 -c ps.ncar.xsede.org

Couldn't find bwctl

iperf -c ping.online.net -i 5 -u -r

1.05 Mbps

iperf -c ping-90ms.online.net -i 5 -u -r

988 Kbps

iperf -c debit.k-net.fr -i 10 -T 100

Failed

iperf -c speedtest.serverius.net

67.5 Kbps

iperf -c iperf.he.net

57.2 Mbps

Offline

#6 2020-04-22 05:54:46

seth
Member
Registered: 2012-09-03
Posts: 16,250

Re: Only fast, speedtest-cli, pacman and curl get good speeds

Downloads some live distro (grml, knoppix, …) and test the behavior there. This is probably a routing issue and the results are rather random.

I cannot test their max performance from here, because, for the time being, the fastest (and only) connection I can get is 30MBit/s …

Offline

#7 2020-04-24 07:07:28

xenobro
Member
Registered: 2019-05-25
Posts: 33

Re: Only fast, speedtest-cli, pacman and curl get good speeds

Hi, seth, and thank you once again.

I'll do what you say and dl grml, knoppix, ubuntu (if there's a live mode) and debian (and maybe alpine) to check how wget/firefox fare against curl, but this will take a few days: I'll update as I go. What tests should I run though? The same from the page you posted, trying to dl random things or isos with wget/firefox and curl or both?
Also, while the results from the latest tests were pratically the same with wget and curl, if it was a routing issue, why should there be such a big difference when dl'ing anything from the high speed arch server? And also, the problem only seems visible when curl's getting higher speeds, like 30+ MB/sec, while on lower speeds they seem to get about the same...
Finally, what routing issues could be happening? If that's the case, is there anything I can do to correct this?
Anyway, just out of curiosity, how much MB/sec dl are you getting on your 30 Mbps, just so I know how much I should expect?

Thank you once again and I'll update as I run the tests,
xenobro.

Last edited by xenobro (2020-04-24 07:12:31)

Offline

#8 2020-07-17 20:52:52

xenobro
Member
Registered: 2019-05-25
Posts: 33

Re: Only fast, speedtest-cli, pacman and curl get good speeds

Hi, seth and everyone.
Sorry for taking so long to reply: there's been a lot going on here with the pandemic, so I couldn't really take time to do the necessary tests until recently.

Anyway, I reran all the tests 3-5 times each, with both wget and curl, on grml and Alpine (Knoppix didn't boot for some reason), and actually got almost exactly the same results, even though it's been 3 months since the original tests. The few ones that came different, when I ran them again in Arch, got the same results as on the 2 others now.
So while these may seem random when comparing speeds gotten between sites, they're actually extremely consistent when you compare the many results you get from each site even throughout 3 months.
Also, as I predicted, one of the few sites where I got 35+ MB/sec speeds, archlinux.c3sl.ufpr.br, now can't go over 12-15 MB/sec. In its place however, br.mirror.archlinux-br.org now gets 35+ most of the time. Cachefly and a few others throughout the net still get it constantly without any problems, while most sites don't.

Any ideas on what should or could I do?
Thank you very much, xenobro.

Last edited by xenobro (2020-07-17 20:54:13)

Offline

#9 2020-07-21 03:57:16

xenobro
Member
Registered: 2019-05-25
Posts: 33

Re: Only fast, speedtest-cli, pacman and curl get good speeds

Oh, also some random file download sites throughout the net can give me those speeds more often than "normal" sites, if that makes any sense.

Any ideas? Can somebody help me figure out why the slow speeds on many others that should be getting way better performance?
Thank you very much, xenobro.

Offline

Board footer

Powered by FluxBB