1 |
On 16/1/21 6:56 am, thelma@×××××××××××.com wrote: |
2 |
> On 1/15/21 1:11 AM, Raffaele BELARDI wrote: |
3 |
>>> -----Original Message----- |
4 |
>>> From: thelma@×××××××××××.com <thelma@×××××××××××.com> |
5 |
>>> Sent: Friday, January 15, 2021 07:57 |
6 |
>>> To: Gentoo mailing list <gentoo-user@l.g.o> |
7 |
>>> Subject: [gentoo-user] network transfer speed |
8 |
>>> |
9 |
>>> |
10 |
>>> On both of my systems the network card speed is showing 1000 |
11 |
>>> cat /sys/class/net/enp4s0/speed 1000 |
12 |
>>> |
13 |
>>> but when I do rsync larage file I only see about: 20 to 22MB/s On my home |
14 |
>>> network I get about 110MB/s between PC's |
15 |
>>> |
16 |
>>> Both PC's have SSD and the swith is Gigabit (I think). |
17 |
>>> How to find a the bottleneck? |
18 |
>> If the PCs attached to the switch show 1000 then the switch _is_ gigabit. |
19 |
>> |
20 |
>> On my 1Gb home network I have an FTP transfer speed between Gentoo PCs A and B of almost 900Mbps, the other way round is almost half of that. One difference between the two systems is the disk, A uses SATA-2 disk while B has SATA-3. |
21 |
>> |
22 |
>> Does the 'B' in 110MB/s stand for byte? If so you have 880Mbps which is not bad, the problem probably lies somewhere else. Otherwise you could check the switch error count (if you have a managed switch) or the network card error count, just to ensure you don't have a cabling/connector problem. |
23 |
>> |
24 |
>> Have you tried other transfer methods just for comparison? I think FTP is still the fastest way to transfer files, though insecure or inconvenient as it might be. I have no experience with rsync. |
25 |
>> |
26 |
>> raffaele |
27 |
> On a remote network I run ethtool on both cards and I got both 1000Mb/s speed |
28 |
> |
29 |
> 1.) |
30 |
> ethtool net0 |
31 |
> Settings for net0: |
32 |
> Supported ports: [ TP MII ] |
33 |
> Supported link modes: 10baseT/Half 10baseT/Full |
34 |
> 100baseT/Half 100baseT/Full |
35 |
> 1000baseT/Half 1000baseT/Full |
36 |
> Supported pause frame use: No |
37 |
> Supports auto-negotiation: Yes |
38 |
> Advertised link modes: 10baseT/Half 10baseT/Full |
39 |
> 100baseT/Half 100baseT/Full |
40 |
> 1000baseT/Half 1000baseT/Full |
41 |
> Advertised pause frame use: Symmetric Receive-only |
42 |
> Advertised auto-negotiation: Yes |
43 |
> Link partner advertised link modes: 10baseT/Half 10baseT/Full |
44 |
> 100baseT/Half 100baseT/Full |
45 |
> 1000baseT/Full |
46 |
> Link partner advertised pause frame use: Symmetric |
47 |
> Link partner advertised auto-negotiation: Yes |
48 |
> Speed: 1000Mb/s |
49 |
> Duplex: Full |
50 |
> Port: MII |
51 |
> PHYAD: 0 |
52 |
> Transceiver: internal |
53 |
> Auto-negotiation: on |
54 |
> MDI-X: on (auto) |
55 |
> |
56 |
> 2.) |
57 |
> ethtool enp4s0 |
58 |
> Settings for enp4s0: |
59 |
> Supported ports: [ TP ] |
60 |
> Supported link modes: 10baseT/Half 10baseT/Full |
61 |
> 100baseT/Half 100baseT/Full |
62 |
> 1000baseT/Full |
63 |
> Supported pause frame use: Symmetric |
64 |
> Supports auto-negotiation: Yes |
65 |
> Advertised link modes: 10baseT/Half 10baseT/Full |
66 |
> 100baseT/Half 100baseT/Full |
67 |
> 1000baseT/Full |
68 |
> Advertised pause frame use: Symmetric |
69 |
> Advertised auto-negotiation: Yes |
70 |
> Speed: 1000Mb/s |
71 |
> Duplex: Full |
72 |
> Port: Twisted Pair |
73 |
> PHYAD: 1 |
74 |
> Transceiver: internal |
75 |
> Auto-negotiation: on |
76 |
> MDI-X: on (auto) |
77 |
> |
78 |
What brand/type of interface (Realtek, int/ext USB ?) There is an |
79 |
obscure bug that drops throughput to ~1/3 in Realtek RTL8153 chip on |
80 |
USB2 and the r8152 kernel driver. Some interfaces (e.g., most raspberry |
81 |
pi) connect via internal USB with its own throughput problems. Probably |
82 |
not the problem in this case but worth mentioning. |
83 |
|
84 |
Rsync is not a good tool for measuring throughput as its optimisations |
85 |
are problematic - and this can show up when comparing local to remote |
86 |
tests. If on a file system it will always transfer the whole file (and |
87 |
it considers a network filesystem local) while on a network it will use |
88 |
the transfer algorithm which can vary depending on commandline switches |
89 |
used and where the remote file lives, and this can change between tests |
90 |
depending on conditions. Whole files have little overhead, whereas the |
91 |
algorithm overhead can actually be much slower (hence why it tries to do |
92 |
it this way) Try and use a largish file created with /dev/random and dd |
93 |
as well as a simple netcat type transfer mechanism to measure. Note |
94 |
that encryption can severely limit transfers due to cpu limiting if on a |
95 |
low powered system and rsync over ssh uses encrytion. And if a secure |
96 |
VPN is in the mix its orders of magnitude worse. Also note that |
97 |
something like an inline IDS on a link can drop throughput by 50% or |
98 |
more (measured on an older enterprise Cisco device) which further |
99 |
complicates local/remote via internet comparisons. |
100 |
|
101 |
|
102 |
Good luck in finding the problem - its really not simple and easy. |
103 |
|
104 |
BillK |