Speed issue when using 2 seperate networks

[Harrie]

I've encountered an issue with my networking where my 1Gbps LAN has it's
speed capped at 54Mbps.

I'm using a laptop with a 1Gbps NIC and 54Mbps WIFI adapter. The WIFI is
connected to an wireless access point (network 1). The 1Gbps NIC is connected
to another network. Both networks are private. The networks are not connected.

While transferring between my laptop and a Windows XP box that is only on
the 1Gbps network I've noticed the speed stays at a constant 54Mbps during
the entire transfer. I'm wondering why the speed of an 1Gbps network is being
capped at 54Mbps.

 

[Bob Willard]

I've encountered an issue with my networking where my 1Gbps LAN has it's
speed capped at 54Mbps.

I'm using a laptop with a 1Gbps NIC and 54Mbps WIFI adapter. The WIFI is
connected to an wireless access point (network 1). The 1Gbps NIC is connected
to another network. Both networks are private. The networks are not connected.

While transferring between my laptop and a Windows XP box that is only on
the 1Gbps network I've noticed the speed stays at a constant 54Mbps during
the entire transfer. I'm wondering why the speed of an 1Gbps network is being
capped at 54Mbps.

Because the data must flow through a fat (1Gb/s) pipe *and* through a skinny
(54 MB/s) pipe. The skinny pipe is the bottleneck.

 

[Harrie]

Because the data must flow through a fat (1Gb/s) pipe *and* through a skinny
(54 MB/s) pipe. The skinny pipe is the bottleneck.

Negative, the data stays on the wired network. Let me try to make a little
text diagram:

---------------- -----------------
---------
| WIFI Network | | Wired Network |-----------| Host B |
---------------- -----------------
---------
| |
| --------- |
|-----------| Host A |---------|
---------

Host A isn't supposed to bridge the networks.
In this situation traffic between hosts A and B is only on the wired network.

Also I have noticed that disabling the WIFI on Host A removes the limit,
immediatly increasing the speed.

 

[Harrie]

So much for the diagram. Back to plain describing.

Host A connects to 2 networks. Network A is a WIFI network (802.11g).
Network B is a wired network (1000BASE-T).

Host B is only connected to Network B.

The only path between Host A and Host B is over the wired network.

 

[Bob Willard]

So much for the diagram. Back to plain describing.

Host A connects to 2 networks. Network A is a WIFI network (802.11g).
Network B is a wired network (1000BASE-T).

Host B is only connected to Network B.

The only path between Host A and Host B is over the wired network.

Interesting. How are you measuring speeds? To be safe (against
software that mis-reports data rates), I generally use file-to-file
transfers using network-mapped drives; for which there are four
different STRs (Sustained Transfer Rates): A pushing data to B's HD,
A pulling data from B's HD, B pushing data to A's HD, and B pulling
data from A's HD. In some cases, I've seen very different STRs from
the four cases, and this might offer some insight into your case.
Also, make sure you measure STRs with a single large file; not a folder
full of small files.

If you see differences from the four STR measurements, see if A and B
are running the same OS (i.e., XP PRO SP2), and make sure you measure
STRs using the same credentials (userass) on both A and B.

 

[Harrie]

Just to be concrete:
Host A: Windows Vista Premium. 1GB RAM, AMD Turion 64x2 1.8Ghz
Host B: Windows XP Professional, 512 MB RAM, Intel Pentium M 1.73Ghz.

To measure the speed on Host A I used the resource monitor.
To measure the speed on Host B I used the task manager Networking tab.
The measurements of both seemed to match.

B pushing to A:
Speed ~53-54 Mbps

A pushing to B:
Speed max ~40 Mbps, falling back to ~15 Mbps every ~15 seconds

B pulling from A:
Good speed initially (300+ Mbps), but once Windows Vista starts swapping
memory to hard drive the system takes a huge perfomance hit. Network speed
reduced to ~6 Mbps. Sometimes the system frees up a bit, speeds rise to about
150Mbps only to fall back into sluggish mode after a few seconds.
cpu usage on Vista 40%-50%
used pagefile on Vista 1.6 GB
cpu usage on XP 0%-5%
Used pagefile on XP 600 MB

A Pulling from B: 58Mbps, dropping to 0 for an instant every 10 seconds. No
noticable system lag on Vista.

The picture I'm getting is that I'd be better off ditching Vista. If the
networking has issues like this, what is the rest going to be like?

Interesting. How are you measuring speeds?

I used

 

[Bob Willard]

Just to be concrete:
Host A: Windows Vista Premium. 1GB RAM, AMD Turion 64x2 1.8Ghz
Host B: Windows XP Professional, 512 MB RAM, Intel Pentium M 1.73Ghz.

To measure the speed on Host A I used the resource monitor.
To measure the speed on Host B I used the task manager Networking tab.
The measurements of both seemed to match.

B pushing to A:
Speed ~53-54 Mbps

A pushing to B:
Speed max ~40 Mbps, falling back to ~15 Mbps every ~15 seconds

B pulling from A:
Good speed initially (300+ Mbps), but once Windows Vista starts swapping
memory to hard drive the system takes a huge perfomance hit. Network speed
reduced to ~6 Mbps. Sometimes the system frees up a bit, speeds rise to about
150Mbps only to fall back into sluggish mode after a few seconds.
cpu usage on Vista 40%-50%
used pagefile on Vista 1.6 GB
cpu usage on XP 0%-5%
Used pagefile on XP 600 MB

A Pulling from B: 58Mbps, dropping to 0 for an instant every 10 seconds. No
noticable system lag on Vista.

The picture I'm getting is that I'd be better off ditching Vista. If the
networking has issues like this, what is the rest going to be like?




I used

Egads, those STRs are horrible. If you haven't already, you may want to
poke around on the NIC vendor's web site (particularly on the Vista box) to
see if there is a better driver.

Your observation about swapping is interesting; but I don't see why there
should be any swapping on your (pretty strong) PCs if nothing is going on
other than HD-HD copies across a LAN. No reason for data being copied,
on either inbound or outbound transfers, to flow through the pagefile.

In an earlier era, I found that STRs between XP boxes and W9x boxes, were
significantly lower (2-6 MB/s on a 100 Mb/s LAN) than XP-XP STRs (9-10 MB/s
on the same 100 Mb/s LAN). And, when I used different credentials on the
XP and W9x boxes, the STRs were *extremely* low (0.06-0.76 MB/s on the same
100 Mb/s LAN); I never found an explanation for the strong effect on STRs
due to different credentials (or for the decreased STRs due to cross-OSs).