latest max speed with Gigabit switch!

  • Thread starter Thread starter Geoff
  • Start date Start date
G

Geoff

Hello

I have connected the 1 Gigabit TP-Link switch and get odd results!

I have a Windows 7 Home Premium and a Windows XP Pro PC connected to
the switch and the ADSL modem/router also connected to it.

The Internet connections seems OK - download/upload speeds as normal
for both PCs.

But! When I transfer a 1GB file from the Windows 7 PC to the XP Pro I
see a transfer speed of approx 200KB/sec.

From the XP Pro to the Windows 7 I see approx 4MB/sec!

Why the different speeds and why are both speeds less than before
without the switch?!

Any thoughts?

Geoff
 
Hello

I have connected the 1 Gigabit TP-Link switch and get odd results!

I have a Windows 7 Home Premium and a Windows XP Pro PC connected to
the switch and the ADSL modem/router also connected to it.

The Internet connections seems OK - download/upload speeds as normal
for both PCs.

But! When I transfer a 1GB file from the Windows 7 PC to the XP Pro I
see a transfer speed of approx 200KB/sec.

From the XP Pro to the Windows 7 I see approx 4MB/sec!

Why the different speeds and why are both speeds less than before
without the switch?!

Any thoughts?

Geoff

I replaced the 1000 Mb/sec switch with my older10/100 Mb/sec switch
and was able to get approx 10MB/sec in both directions.

Does this point to why the 1000 Mb/sec switch is not providing faster
transfers?

Geoff
 
Hello

I have connected the 1 Gigabit TP-Link switch and get odd results!

I have a Windows 7 Home Premium and a Windows XP Pro PC connected to
the switch and the ADSL modem/router also connected to it.

The Internet connections seems OK - download/upload speeds as normal
for both PCs.

But! When I transfer a 1GB file from the Windows 7 PC to the XP Pro I
see a transfer speed of approx 200KB/sec.

From the XP Pro to the Windows 7 I see approx 4MB/sec!

Why the different speeds and why are both speeds less than before
without the switch?!

Any thoughts?

Geoff


have just been and bought cat 5e cables - but no improvement!?

Geoff
 
have just been and bought cat 5e cables - but no improvement!?

Geoff

There are so many things to try that I'm at a loss to suggest where to
begin. As usual with most problems it is almost certain that others
(sometimes many others) have seen them all before and at least some of the
time they've resolved them.

For myself, I'd be tempted to totally uninstall the NIC drivers on both
machines and then re-install them fresh (and newly-downloaded) on the
chance that they installed originally with parameters suited to the slower
switch. I'd also do some cable/port/nic swapping if for no other reason
than that it is easy and free. Then I'd google the search term
'troubleshoot ethernet speed problems' (without the quotes of course) and
then browse through the various forum postings and blogs and sites to see
what is being said. I saw so many possibilities that it would be impossible
to start repeating them wholesale. But some which showed up and looked
likely were: various forms of auto negotiation, flow control and block
sizes. You might also want to look at
http://support.microsoft.com/kb/325487 for ideas.
 
Geoff said:
have just been and bought cat 5e cables - but no improvement!?

Geoff

One reviewer here (Feedback section), reports 110-120MB/sec file
transfer rate. So the box is capable. Other people report lower
rates, but not as low as you.

http://www.newegg.com/Product/Product.aspx?Item=N82E16833704042

Apparently, if it comes in the black plastic case, it's V4. If in
a white case, it is V3 and a different chipset (9K jumbo frames).

One reviewer on Newegg, claims this is what is used for V4, but this only
mentions 9K frames.

http://www.realtek.com/products/productsView.aspx?Langid=1&PFid=15&Level=5&Conn=4&ProdID=265

This is the other chip, that might have been used in the white one.
Doesn't seem too much different, in terms of the acronym disease.

http://www.realtek.com/products/productsView.aspx?Langid=1&PFid=15&Level=5&Conn=4&ProdID=197

I would fire up a copy of Wireshark, and trace packets, and see
if there are reports of duplicate packets or abnormal behavior.
Test with a "working" setup, then put the TPLink box back
in the circuit and repeat the test, and compare the traces.

http://en.wikipedia.org/wiki/Wireshark

I would also carefully inspect the setup, to make sure nothing
is connected in a loop or the like.

While the box could be defective in some way, I'd expect more
of a "go no-go" behavior, rather than working "somewhat". The
results suggest investigating the packet protocols (excessive
duplicate packets or the like, or some routing protocols
stuck in a loop) may give a hint.

Paul
 
One reviewer here (Feedback section), reports 110-120MB/sec file
transfer rate. So the box is capable. Other people report lower
rates, but not as low as you.

http://www.newegg.com/Product/Product.aspx?Item=N82E16833704042

Apparently, if it comes in the black plastic case, it's V4. If in
a white case, it is V3 and a different chipset (9K jumbo frames).

One reviewer on Newegg, claims this is what is used for V4, but this only
mentions 9K frames.

http://www.realtek.com/products/productsView.aspx?Langid=1&PFid=15&Level=5&Conn=4&ProdID=265

This is the other chip, that might have been used in the white one.
Doesn't seem too much different, in terms of the acronym disease.

http://www.realtek.com/products/productsView.aspx?Langid=1&PFid=15&Level=5&Conn=4&ProdID=197

I would fire up a copy of Wireshark, and trace packets, and see
if there are reports of duplicate packets or abnormal behavior.
Test with a "working" setup, then put the TPLink box back
in the circuit and repeat the test, and compare the traces.

http://en.wikipedia.org/wiki/Wireshark

I would also carefully inspect the setup, to make sure nothing
is connected in a loop or the like.

While the box could be defective in some way, I'd expect more
of a "go no-go" behavior, rather than working "somewhat". The
results suggest investigating the packet protocols (excessive
duplicate packets or the like, or some routing protocols
stuck in a loop) may give a hint.

Paul

Thanks Paul - will have a go with Wireshark ..

The TP-Link switch box is black ..

I have just downloaded the latest drivers for Windows 7 and XP from
the realtek site and with auto negociation (default) I see

Windows 7 PC to XP Pro PC 200KB/sec

XP Pro to Windows 7 8MB/sec

?!

Geoff
 
I would fire up a copy of Wireshark, and trace packets, and see
if there are reports of duplicate packets or abnormal behavior.
Test with a "working" setup, then put the TPLink box back
in the circuit and repeat the test, and compare the traces.

Paul

I have tried Wireshark when moving a file from Windows 7 PC to XP Pro
PC - the slower connection I see a lot of TCP Dup ACK message which
apparently indicate dropped/missing packets ..

'afraid I am at a loss really getting much more from Wireshark ..

Geoff

No. Time Source Destination Protocol Length
Info
487 28.754339 192.168.0.2 192.168.0.3 TCP 66
[TCP Dup ACK 485#2] netbios-ssn > 49270 [ACK] Seq=21375 Ack=128395
Win=65535 Len=0 SLE=129855 SRE=132775

Frame 487: 66 bytes on wire (528 bits), 66 bytes captured (528 bits)
Ethernet II, Src: 00:a1:b0:69:6b:11 (00:a1:b0:69:6b:11), Dst:
00:a1:b0:69:6b:5e (00:a1:b0:69:6b:5e)
Internet Protocol Version 4, Src: 192.168.0.2 (192.168.0.2), Dst:
192.168.0.3 (192.168.0.3)
Transmission Control Protocol, Src Port: netbios-ssn (139), Dst Port:
49270 (49270), Seq: 21375, Ack: 128395, Len: 0
 
Geoff said:
Hello

I have connected the 1 Gigabit TP-Link switch and get odd results!

I have a Windows 7 Home Premium and a Windows XP Pro PC connected to
the switch and the ADSL modem/router also connected to it.

The Internet connections seems OK - download/upload speeds as normal
for both PCs.

But! When I transfer a 1GB file from the Windows 7 PC to the XP Pro I
see a transfer speed of approx 200KB/sec.

From the XP Pro to the Windows 7 I see approx 4MB/sec!

Why the different speeds and why are both speeds less than before
without the switch?!

Any thoughts?

Geoff

Since you put the "1" (a value) before "Gigabit", the assumption is that
you meant "1 Gigabit" as the bandwidth rating for the switch, not that
it was a Gigabit brand of switch. "TP-Link" is a manufacturer name. I
saw over 20 gigabit-rated switches at the tp-link.com site. I also saw
some only rated up to 100 Mbps. You didn't mention the model that you
are using.

Hard to know which speeds where you used the wrong values. You don't
have a 1 GB switch. You have a 1 Gbps switch. That's bits, not bytes.

You have a 1 GB file. That's 8 gigaBITS.

A transfer rate of 200 KB/sec is 1.6 gigabits/sec which exceeds the
rating for the gigabit switch. A transfer rate of 4MB/sec is 32
megabits/sec. You're mixing MBps with Mbps and GBps with Gbps. Can't
be sure of your numbers because you mixed byte with bit.

You never identified what network card, onboard controller, USB-attached
NIC, or what you have in each host. You never identified what software
you installed that interrogates your network traffic (anti-virus,
firewall, web shields or site content checking, HIPs, etc) and if you
disabled or uninstalled it all during bandwidth testing. You never
mention your actual network setup, as in which devices are connected via
wired or wireless, and if the devices go straight to the switch or, if
wireless, go through APs. If wired, how long are the network cables and
what type are they (CAT3, CAT5, CAT5E, CAT6)? You didn't mention if all
other communication was quiescent (no Internet traffic, not intranet
host traffic other than the file transfer). Did you use SysInternals'
TCPview or Wireshark on each intranet host to ensure they (and the
router) weren't generating traffic (i.e., all hosts quiescent) before
and after the file transfer test? Since the source and target intranet
hosts are only involved in the file transfer, what happens when you
disconnect all other hosts from the switch, like the router? Is QoS
(quality of service) settings, if available, enabled or disabled in the
switch?

Have you used Device Manager to look at the properties for the NICs in
your intranet hosts? The switch should support autonegotiation but do
your NICs? Is the NIC optimized for CPU or throughput? Is it set to
autonegotiation or a particular speed? If already set for
autonegotiation, can you set it to use 1000 full duplex? Is it using
half- or full-duplex mode?
 
WoW! Let me try to give some further info then.
Since you put the "1" (a value) before "Gigabit", the assumption is that
you meant "1 Gigabit" as the bandwidth rating for the switch, not that
it was a Gigabit brand of switch. "TP-Link" is a manufacturer name. I
saw over 20 gigabit-rated switches at the tp-link.com site. I also saw
some only rated up to 100 Mbps. You didn't mention the model that you
are using.

Hard to know which speeds where you used the wrong values. You don't
have a 1 GB switch. You have a 1 Gbps switch. That's bits, not bytes.

You have a 1 GB file. That's 8 gigaBITS.

The switch is TP-Link model TL-SG1005D. On the cardboard box it also
states 1000Mbps and describes it as a 5-Port Gigabit Desktop Switch.
A transfer rate of 200 KB/sec is 1.6 gigabits/sec which exceeds the
rating for the gigabit switch. A transfer rate of 4MB/sec is 32
megabits/sec. You're mixing MBps with Mbps and GBps with Gbps. Can't
be sure of your numbers because you mixed byte with bit.

I see what you mean but I had copied what I see when I use Windows
Explorer to copy and then paste a file from the Windows 7 PC to the XP
Pro PC and expand the details part of the window. There I see

200KB/sec and 8MB/sec for the reverse copying !
You never identified what network card, onboard controller, USB-attached
NIC, or what you have in each host.

The 2 NICs are ADDON Fast Ethernet PCI Adapters, NIC1000Rv2.

When installed under Windows 7 Home Premium they are in Device Manager
asRealtek PCI GBE Family Controller.

Under XP Pro Device Manager - Realtek RTL8169/8110 Family Gigabit
Ethernet NIC.
You never identified what software
you installed that interrogates your network traffic (anti-virus,
firewall, web shields or site content checking, HIPs, etc) and if you
disabled or uninstalled it all during bandwidth testing.

I have AVG Internet Security 2011 installed. I have tried turning the
firewall off and that does not change the transfer speeds mentioned
above.
You never
mention your actual network setup, as in which devices are connected via
wired or wireless, and if the devices go straight to the switch or, if
wireless, go through APs.
If wired, how long are the network cables and
what type are they (CAT3, CAT5, CAT5E, CAT6)?

The 2 PCs are directly connected to the switch, the XP Pro with a 10m
cat 5e cable and the Windows 7 PC with a 3m cat 5e cable.

The Netgear DG834G v5 modem/router is connected to the switch by a
0.5m cat 5e cable.
You didn't mention if all
other communication was quiescent (no Internet traffic, not intranet
host traffic other than the file transfer). Did you use SysInternals'
TCPview or Wireshark on each intranet host to ensure they (and the
router) weren't generating traffic (i.e., all hosts quiescent) before
and after the file transfer test? Since the source and target intranet
hosts are only involved in the file transfer, what happens when you
disconnect all other hosts from the switch, like the router? Is QoS
(quality of service) settings, if available, enabled or disabled in the
switch?

I have disconnected the modem/router from the switch but the transfer
speeds are as above.
Have you used Device Manager to look at the properties for the NICs in
your intranet hosts? The switch should support autonegotiation but do
your NICs?
Is the NIC optimized for CPU or throughput? Is it set to
autonegotiation or a particular speed? If already set for
autonegotiation, can you set it to use 1000 full duplex? Is it using
half- or full-duplex mode?

Both NICs can be set for autonegociation and were for the transfer
figures above.

I have tried 10 Mbps/full duplex which gives a transfer speed of 1.2
MB/sec (according to Windows ...) in both directions and 100 Mbps/full
duplex which gives a transer speed of 12 MB/sec in both directions.

When I use 1 Gbps/full duplex the transfer speeds fall to 144 KB/sec.

Now accepting that these units are wrong how to I get a mesaure in the
correct units?

Cheers

Geoff
 
You never identified what network card, onboard controller, USB-attached
NIC, or what you have in each host. You never identified what software
you installed that interrogates your network traffic (anti-virus,
firewall, web shields or site content checking, HIPs, etc) and if you
disabled or uninstalled it all during bandwidth testing. You never
mention your actual network setup, as in which devices are connected via
wired or wireless, and if the devices go straight to the switch or, if
wireless, go through APs. If wired, how long are the network cables and

I forgot to add that there is no wireless connection being used at the
moment.

Geoff
 
VanguardLH said:
Have you used Device Manager to look at the properties for the NICs in
your intranet hosts? The switch should support autonegotiation but do
your NICs? Is the NIC optimized for CPU or throughput? Is it set to
autonegotiation or a particular speed? If already set for
autonegotiation, can you set it to use 1000 full duplex? Is it using
half- or full-duplex mode?

I think this is a good direction to look.

There are more properties in the NIC properties entry in Device Manager
than just the speed. It could be some kind of flow control option which
is incorrectly set.

My NIC has:

"Flow Control" (with a possible option of "Disabled")
"Jumbo Packet" (currently 1514, for compatibility)
"Receive Buffers" 256
"Transmit Buffers" 256

I also have 802.1p which is currently enabled, but could be turned off.

http://en.wikipedia.org/wiki/802.1p

Experimenting with some settings there might help.

*******

With regard to collecting information, you can use Wireshark on
both computers, and collect traces on them at the same time.

First, go to the Date and Time panel, and request NTP time synchonization.
Use the same NTP server for both computers. The intention of doing that,
is so the time stamps (displayed in hours:minute:seconds) match. Then,
when you attempt a transfer between machines, you can compare the
traces on the two machines, and see if a packet sent by one machine,
is showing up at the other machine (within the time stamp uncertainty).

I'm having trouble understanding how "duplicate ACKs" could be
showing up, when we're talking about a network connection that
only goes through a switch. That seems pretty weird.

*******

And just for fun, there is another way to analyse file copying,
thanks to Mark Russinovich.

"Inside Vista SP1 File Copy Improvements"

http://blogs.technet.com/b/markrussinovich/archive/2008/02/04/2826167.aspx

Paul
 
Geoff said:
WoW! Let me try to give some further info then.


The switch is TP-Link model TL-SG1005D. On the cardboard box it also
states 1000Mbps and describes it as a 5-Port Gigabit Desktop Switch.

10/100/1000 Mbps rated, so 1 Gbps is supported.

Auto MDI/MDIX supported, so you don't need a cross-over cable when
chaining together multiple switches or routers to this device. Good
because otherwise there could've been a problem with connecting your
router to this switch.

No wireless connects so nothing needs to be discussed about that (e.g.,
max bandwidth of, say, 54 Mbps for wireless versus 1 Gbps for wired).

No configuration, so forget the QoS question.

Automatic power reduction but no mention of idle detect, sleep, and
wakeup delays. Does mention it'll up the power output for longer cables
but that won't help on the other end to make the hosts up their power
for longer cables.

There's mention of supporting 9KB (15KB in their manual) jumbo frames
(instead of the old 1500KB frame) but I'm not sure Windows is going to
fluctuate its frame size based on some auto-negotiation with an upstream
device, plus it won't match what your ISP expects. If the frame size is
larger than can be handled at either end, you end up with layer 2
fragmentation or reassembly that causes delay. I haven't investigated
Windows 7 to see if it auto-negotiates to larger frame sizes but I know
Windows XP won't (you'll have to use a network tweaker but then the size
is fixed and won't match with your ISP). Whether or not the "store and
forward" mechanism of this switch forces the use of jumbo frames is not
clear from its manual or description, so maybe something like Wireshark
could tell you the frame size.

Alas, I didn't see mention that its port LEDs show negotiated speed by
using different colors. It just shows status regarding link/active.
The 2 NICs are ADDON Fast Ethernet PCI Adapters, NIC1000Rv2.

"Fast Ethernet" means 100 Mbps (100BASE-TX) with backward compatibility
(usually auto-negotiated) with 10 Mbps (10BASE-TX). A NIC listed as
"Fast Ethernet" means it supports 10/100 Mbps, not 1000 Mbps. See
http://en.wikipedia.org/wiki/Fast_ethernet. For 1000 Mbps (1 Gbps),
you'd want something that identifies itself as Gigabit Ethernet; see
http://en.wikipedia.org/wiki/Gigabit_Ethernet.

I've seen a NIC detect a problem and auto-negotiate down to 10 Mbps
(from an initial connect of 100 Mbps) but which will not auto-negotiate
back up. That is, once the rate gets dirty and has to drop then it
stays there. I forget the typical culprits of having to lower the rate
but having to resend over some percentage of lost (unacknowledged)
packets is probably one cause.

Did you actually install a network card? If so, you should be able to
see a sticker with an actual model number instead of getting just some
specs detected within Windows. If not, is it an onboard NIC (i.e., a
backpanel RJ-45 connector from a controller on the motherboard)? In
that case, knowing the motherboard maker and model could help in
identifying just what specs its NIC controller can support.

Below is an example of an AddOn NIC that declares itself as a "Fast
Ethernet" device. Notice it never mentions 1000 Mbps (1 Gbps) because
it isn't a gigabit NIC.

http://www.addon-tech.com/new/product_info.php?id=76

Looks like you added the switch trying to get around the limitation of
the switch in the router. That is, you added a 10/100/1000 Mbps switch
so you weren't limited to the 10/100 Mbps switch in the router for
throughput between your intranet hosts. However, you need to perform
the additional step of getting better NICs in your intranet hosts. They
only have FastEther NICs which limits them to 10/100 Mbps maximum.
When installed under Windows 7 Home Premium they are in Device Manager
asRealtek PCI GBE Family Controller.

Under XP Pro Device Manager - Realtek RTL8169/8110 Family Gigabit
Ethernet NIC.

Can't really tell if that is for a daughtercard (a NIC card) or for a
controller on the motherboard. The controller on the motherboard is the
same one often used on a NIC card.
I have AVG Internet Security 2011 installed. I have tried turning the
firewall off and that does not change the transfer speeds mentioned
above.

However, the anti-virus program is still going to interrogate all the
bytes for a newly created or modified file. That interrogation takes
time. Although you could disable the AV program, that doesn't always
work. Some AV programs will continue routing the network traffic
through their proxy or handler. Disabling just means the content won't
be interrogated but if there is a problem with the proxy or handler than
throughput remains affected. I had that problem way back when I used to
have Norton 2003 installed: disabling interrogation didn't eliminate the
problem of blocked traffic when their proxy went unresponsive so I had
to reboot to get it responsive again. As I recall with AVG, you have to
uninstall it to really get it out of the way.

Since you are testing throughput between intranet hosts, you could
disconnect the router (or power it off) during testing and uninstall AVG
(and other security software) on the source and target hosts. Reinstall
the security software after testing and before reconnecting the Internet
setup.
The 2 PCs are directly connected to the switch, the XP Pro with a 10m
cat 5e cable and the Windows 7 PC with a 3m cat 5e cable.

The Netgear DG834G v5 modem/router is connected to the switch by a
0.5m cat 5e cable.

Well within the 100 m maximum for Ethernet, and capable cables, too.
I have disconnected the modem/router from the switch but the transfer
speeds are as above.

Did you make sure the source and target hosts (and any other hosts
connected to the switch) were all quiescent? That is, with the router
disconnected or powered off, do you see any traffic over your intranet
going through the common switch? I don't know if your switch's LEDs
will blink when there is traffic or if you'll have to use a packet
sniffer, like Wireshark, to be sure.
Both NICs can be set for autonegociation and were for the transfer
figures above.

I have tried 10 Mbps/full duplex which gives a transfer speed of 1.2
MB/sec (according to Windows ...) in both directions and 100 Mbps/full
duplex which gives a transer speed of 12 MB/sec in both directions.

When I use 1 Gbps/full duplex the transfer speeds fall to 144 KB/sec.

Which makes me wonder if that new "feature" of jumbo frames is getting
in the way and causing fragmentation and reassembly (to generate 1500 B
frames expected by the host NICs). Alas, the description for this
switch says there is no user configurable settings to, say, disable the
jumbo frame feature or limit the frame size (so it matches your hosts).
Now accepting that these units are wrong how to I get a mesaure in the
correct units?

http://www.bitplumber.net/2009/03/how-to-configure-jumbo-frames/

Notice the 3rd bullet that says, "You are sure all your host operating
systems and NIC drivers will support Jumbo Frames". I don't know that
just getting drivers that have jumbo frame support will fix your program
unless the hardware mentions it also supports larger frames. Probably
but maybe not.

Back in my Windows XP box, with the onboard NIC on the mobo, there is no
selection in the NIC properties to support jumbo frames. Some switches
let you enable/disable jumbo frame support but this switch is
non-configurable so I don't know what it might be doing with frame size.
Since it is a "store and forward" switch, it's possible it accumulates
the 1518 B packets to store them and then assembles them into jumbo-
sized frames which your other intranet host cannot handle directly.

Did you check if the NIC properties in Device Manager on your source and
target intranet hosts has a jumbo setting and that it's enabled?
 
10/100/1000 Mbps rated, so 1 Gbps is supported.

Auto MDI/MDIX supported, so you don't need a cross-over cable when
chaining together multiple switches or routers to this device. Good
because otherwise there could've been a problem with connecting your
router to this switch.

No wireless connects so nothing needs to be discussed about that (e.g.,
max bandwidth of, say, 54 Mbps for wireless versus 1 Gbps for wired).

No configuration, so forget the QoS question.

Automatic power reduction but no mention of idle detect, sleep, and
wakeup delays. Does mention it'll up the power output for longer cables
but that won't help on the other end to make the hosts up their power
for longer cables.

There's mention of supporting 9KB (15KB in their manual) jumbo frames
(instead of the old 1500KB frame) but I'm not sure Windows is going to
fluctuate its frame size based on some auto-negotiation with an upstream
device, plus it won't match what your ISP expects. If the frame size is
larger than can be handled at either end, you end up with layer 2
fragmentation or reassembly that causes delay. I haven't investigated
Windows 7 to see if it auto-negotiates to larger frame sizes but I know
Windows XP won't (you'll have to use a network tweaker but then the size
is fixed and won't match with your ISP). Whether or not the "store and
forward" mechanism of this switch forces the use of jumbo frames is not
clear from its manual or description, so maybe something like Wireshark
could tell you the frame size.

Alas, I didn't see mention that its port LEDs show negotiated speed by
using different colors. It just shows status regarding link/active.


"Fast Ethernet" means 100 Mbps (100BASE-TX) with backward compatibility
(usually auto-negotiated) with 10 Mbps (10BASE-TX). A NIC listed as
"Fast Ethernet" means it supports 10/100 Mbps, not 1000 Mbps. See
http://en.wikipedia.org/wiki/Fast_ethernet. For 1000 Mbps (1 Gbps),
you'd want something that identifies itself as Gigabit Ethernet; see
http://en.wikipedia.org/wiki/Gigabit_Ethernet.

'just a quick comment re the NICs - on the box for this NIC1000Rv2 it
does say 10/100/1000 Ethernet Adapter - so that does mean 1 Gbps
possible presumably On th back of the box I read Port One
10M/100M/1000M STP/UTP Port and 1000Base-T:Cat. 5 or above Cat.5 UTP.

Does that mean I didn't need to spend £24 on cat. 5e cables?!

I am reading the rest of your reply and will get back asap.

Incidentally I set the NICs to 100Mbps/full duplex and timed the
transfer of 1 GB file - it took 90 secs so if

1GB = 1000MBytes
transfer speed 1000MB/90secs is approx 11MB/sec - which is what the
details in the transfer window shows.

Cheers

Geoff
 
Did you actually install a network card? If so, you should be able to
see a sticker with an actual model number instead of getting just some
specs detected within Windows. If not, is it an onboard NIC (i.e., a
backpanel RJ-45 connector from a controller on the motherboard)? In
that case, knowing the motherboard maker and model could help in
identifying just what specs its NIC controller can support.

Taking a NIC out I see a chip with Pulse H5007 and 1003-c on it. The
serial number on the NIC is 2010H040213 and the same model description
as on the box, NIC1000Rv2.

Two LEDs ACT and Link.

Cheers

Geoff
 
Geoff said:
Taking a NIC out I see a chip with Pulse H5007 and 1003-c on it. The
serial number on the NIC is 2010H040213 and the same model description
as on the box, NIC1000Rv2.

Two LEDs ACT and Link.

Cheers

Geoff

The part with the Pulse brand name on it, is the transformers chip.
Ethernet is AC coupled, and a transformer is used for galvanic
isolation. There is no DC path between computers, due to
transformers being on either end. "Pulse" is a popular brand
name for those.

A GbE NIC transformer package, has twice as many coils inside it,
as the one shown in this diagram. Two of the coils are chokes,
while the other two function as transformers. The turns ratio
on the transformer, can be used to boost the voltage level if
necessary.

http://www.coolcomponents.co.uk/catalog/resources/CS8900A/cs8900a-sch.gif

*******

I tried your file sharing test here, just for fun.

I installed the Cenatek RAMDisk program on two computers. Each
computer has a GbE NIC on it. I connected them to my Netgear
gigabit switch (which very likely uses the same silicon chip
as yours does, because everyone likes to use the cheapest chip
possible). I probably paid twice as much as you did for your
TPLink, and all I got was a nice plastic casing :-)

With a 1GB sized RAM disk running on each computer, I could store
the source file on one end, and using file sharing, copy the
file into a RAMDisk at the other end. The RAMDisk removes the
disk as a limiting factor in terms of performance.

When I ran the transfer test, I got an average of 55MB/sec. The
CPU on the laptop was flat out at 100%, while I think my desktop
wasn't quite as heavily loaded. It might have gone faster, if
the CPU clock rate was higher.

Paul
 
I think this is a good direction to look.

There are more properties in the NIC properties entry in Device Manager
than just the speed. It could be some kind of flow control option which
is incorrectly set.

My NIC has:

"Flow Control" (with a possible option of "Disabled")
"Jumbo Packet" (currently 1514, for compatibility)
"Receive Buffers" 256
"Transmit Buffers" 256

I also have 802.1p which is currently enabled, but could be turned off.

http://en.wikipedia.org/wiki/802.1p

Experimenting with some settings there might help.

*******

With regard to collecting information, you can use Wireshark on
both computers, and collect traces on them at the same time.

First, go to the Date and Time panel, and request NTP time synchonization.
Use the same NTP server for both computers. The intention of doing that,
is so the time stamps (displayed in hours:minute:seconds) match. Then,
when you attempt a transfer between machines, you can compare the
traces on the two machines, and see if a packet sent by one machine,
is showing up at the other machine (within the time stamp uncertainty).

I'm having trouble understanding how "duplicate ACKs" could be
showing up, when we're talking about a network connection that
only goes through a switch. That seems pretty weird.

*******

And just for fun, there is another way to analyse file copying,
thanks to Mark Russinovich.

"Inside Vista SP1 File Copy Improvements"

http://blogs.technet.com/b/markrussinovich/archive/2008/02/04/2826167.aspx

Paul

Paul

I have just emailed the Tech Support at ADDON - will get back on
Monday after speaking to them ...

Cheers

Geoff
 
10/100/1000 Mbps rated, so 1 Gbps is supported.

Auto MDI/MDIX supported, so you don't need a cross-over cable when
chaining together multiple switches or routers to this device. Good
because otherwise there could've been a problem with connecting your
router to this switch.

No wireless connects so nothing needs to be discussed about that (e.g.,
max bandwidth of, say, 54 Mbps for wireless versus 1 Gbps for wired).

No configuration, so forget the QoS question.

Automatic power reduction but no mention of idle detect, sleep, and
wakeup delays. Does mention it'll up the power output for longer cables
but that won't help on the other end to make the hosts up their power
for longer cables.

There's mention of supporting 9KB (15KB in their manual) jumbo frames
(instead of the old 1500KB frame) but I'm not sure Windows is going to
fluctuate its frame size based on some auto-negotiation with an upstream
device, plus it won't match what your ISP expects. If the frame size is
larger than can be handled at either end, you end up with layer 2
fragmentation or reassembly that causes delay. I haven't investigated
Windows 7 to see if it auto-negotiates to larger frame sizes but I know
Windows XP won't (you'll have to use a network tweaker but then the size
is fixed and won't match with your ISP). Whether or not the "store and
forward" mechanism of this switch forces the use of jumbo frames is not
clear from its manual or description, so maybe something like Wireshark
could tell you the frame size.

Alas, I didn't see mention that its port LEDs show negotiated speed by
using different colors. It just shows status regarding link/active.


"Fast Ethernet" means 100 Mbps (100BASE-TX) with backward compatibility
(usually auto-negotiated) with 10 Mbps (10BASE-TX). A NIC listed as
"Fast Ethernet" means it supports 10/100 Mbps, not 1000 Mbps. See
http://en.wikipedia.org/wiki/Fast_ethernet. For 1000 Mbps (1 Gbps),
you'd want something that identifies itself as Gigabit Ethernet; see
http://en.wikipedia.org/wiki/Gigabit_Ethernet.

I've seen a NIC detect a problem and auto-negotiate down to 10 Mbps
(from an initial connect of 100 Mbps) but which will not auto-negotiate
back up. That is, once the rate gets dirty and has to drop then it
stays there. I forget the typical culprits of having to lower the rate
but having to resend over some percentage of lost (unacknowledged)
packets is probably one cause.

Did you actually install a network card? If so, you should be able to
see a sticker with an actual model number instead of getting just some
specs detected within Windows. If not, is it an onboard NIC (i.e., a
backpanel RJ-45 connector from a controller on the motherboard)? In
that case, knowing the motherboard maker and model could help in
identifying just what specs its NIC controller can support.

Below is an example of an AddOn NIC that declares itself as a "Fast
Ethernet" device. Notice it never mentions 1000 Mbps (1 Gbps) because
it isn't a gigabit NIC.

http://www.addon-tech.com/new/product_info.php?id=76

Looks like you added the switch trying to get around the limitation of
the switch in the router. That is, you added a 10/100/1000 Mbps switch
so you weren't limited to the 10/100 Mbps switch in the router for
throughput between your intranet hosts. However, you need to perform
the additional step of getting better NICs in your intranet hosts. They
only have FastEther NICs which limits them to 10/100 Mbps maximum.


Can't really tell if that is for a daughtercard (a NIC card) or for a
controller on the motherboard. The controller on the motherboard is the
same one often used on a NIC card.


However, the anti-virus program is still going to interrogate all the
bytes for a newly created or modified file. That interrogation takes
time. Although you could disable the AV program, that doesn't always
work. Some AV programs will continue routing the network traffic
through their proxy or handler. Disabling just means the content won't
be interrogated but if there is a problem with the proxy or handler than
throughput remains affected. I had that problem way back when I used to
have Norton 2003 installed: disabling interrogation didn't eliminate the
problem of blocked traffic when their proxy went unresponsive so I had
to reboot to get it responsive again. As I recall with AVG, you have to
uninstall it to really get it out of the way.

Since you are testing throughput between intranet hosts, you could
disconnect the router (or power it off) during testing and uninstall AVG
(and other security software) on the source and target hosts. Reinstall
the security software after testing and before reconnecting the Internet
setup.


Well within the 100 m maximum for Ethernet, and capable cables, too.


Did you make sure the source and target hosts (and any other hosts
connected to the switch) were all quiescent? That is, with the router
disconnected or powered off, do you see any traffic over your intranet
going through the common switch? I don't know if your switch's LEDs
will blink when there is traffic or if you'll have to use a packet
sniffer, like Wireshark, to be sure.


Which makes me wonder if that new "feature" of jumbo frames is getting
in the way and causing fragmentation and reassembly (to generate 1500 B
frames expected by the host NICs). Alas, the description for this
switch says there is no user configurable settings to, say, disable the
jumbo frame feature or limit the frame size (so it matches your hosts).


http://www.bitplumber.net/2009/03/how-to-configure-jumbo-frames/

Notice the 3rd bullet that says, "You are sure all your host operating
systems and NIC drivers will support Jumbo Frames". I don't know that
just getting drivers that have jumbo frame support will fix your program
unless the hardware mentions it also supports larger frames. Probably
but maybe not.

Back in my Windows XP box, with the onboard NIC on the mobo, there is no
selection in the NIC properties to support jumbo frames. Some switches
let you enable/disable jumbo frame support but this switch is
non-configurable so I don't know what it might be doing with frame size.
Since it is a "store and forward" switch, it's possible it accumulates
the 1518 B packets to store them and then assembles them into jumbo-
sized frames which your other intranet host cannot handle directly.

Did you check if the NIC properties in Device Manager on your source and
target intranet hosts has a jumbo setting and that it's enabled?

I have emailed ADDON Tech Support and will speak to them on Monday -
will let you know what happens!

Cheers

Geoff
 
1 GB = 1000 MB (correct)

Transfer rate is 1000 megaBITS/second. So working the other way around
without any knowledge of how long an actual file transfer takes:

1 GB = 8 Gb

8 Gb / 1 Gb/s = 8 seconds

You're measuring about 10 times that duration so I'm wondering if the
transfer is dropping down (falling back) to the 100 Mbps rate. At 100
Mbps, you would get 12.5 MB/s (close to the 11 MB/s you reported). For
quick calculation, there are some online calculators available; e.g.,
http://www.t1shopper.com/tools/calculate/downloadcalculator.php and
http://www.esubnet.net/file-transfer-calc.html.

I don't remember all the reasons a NIC would fall back to negotiate and
stick with a slower speed. Misconfiguration is one cause; see
http://www.itsyourip.com/cisco/speed-and-duplex-issues-when-connecting-to-cisco-switches/.
The folks over in the *.networking newsgroups might know more. Poor or
flaky cabling, connectors, poor signal quality, low power, and other
physical factors can affect fall back.

You said you got brand new Cat5E cables. Did you insert/remove them
several times to make sure you wiped the contacts inside the connectors?
With the cables inserted in the connectors (at the host and at the
switch), have you flexed the end of the cable near the connector to make
sure they have solid connection to the wires and the Link LED on each
end (host and switch) doesn't fluctuate between off and on?

will try the above

Cheers

Geoff
 
Geoff said:
Incidentally I set the NICs to 100Mbps/full duplex and timed the
transfer of 1 GB file - it took 90 secs so if

1GB = 1000MBytes
transfer speed 1000MB/90secs is approx 11MB/sec - which is what the
details in the transfer window shows.

1 GB = 1000 MB (correct)

Transfer rate is 1000 megaBITS/second. So working the other way around
without any knowledge of how long an actual file transfer takes:

1 GB = 8 Gb

8 Gb / 1 Gb/s = 8 seconds

You're measuring about 10 times that duration so I'm wondering if the
transfer is dropping down (falling back) to the 100 Mbps rate. At 100
Mbps, you would get 12.5 MB/s (close to the 11 MB/s you reported). For
quick calculation, there are some online calculators available; e.g.,
http://www.t1shopper.com/tools/calculate/downloadcalculator.php and
http://www.esubnet.net/file-transfer-calc.html.

I don't remember all the reasons a NIC would fall back to negotiate and
stick with a slower speed. Misconfiguration is one cause; see
http://www.itsyourip.com/cisco/speed-and-duplex-issues-when-connecting-to-cisco-switches/.
The folks over in the *.networking newsgroups might know more. Poor or
flaky cabling, connectors, poor signal quality, low power, and other
physical factors can affect fall back.

You said you got brand new Cat5E cables. Did you insert/remove them
several times to make sure you wiped the contacts inside the connectors?
With the cables inserted in the connectors (at the host and at the
switch), have you flexed the end of the cable near the connector to make
sure they have solid connection to the wires and the Link LED on each
end (host and switch) doesn't fluctuate between off and on?
 
Back
Top