uescjp
asked on
Slow LAN network file copy performance when copied between subnets
I am having an issue as I try to migrate my basically flat network to a more robust subnetted network. As a very basic design I have the following:
pc1(on new network)<-->access switch<-->distribution switch(layer 3)<-->firewall<-->router<- ->layer 2 unmanaged switch<-->pc2(on old flat network)
Everything is Gigabit ethernet. When I try to copy a file from pc1 to pc2, it is relatively fast at about 4MB/sec. When I try to copy the same file back from pc2 to pc1, it is atrociously slow at about 100KB/sec. I am able to test everything in a duplicate test environment and cannot duplicate the issue. I suspect that it is the router because the router that is the only thing I don't have access to (because it is managed by our ISP) in the production environment and can verify in the production environment everything on either side of it works just fine.
I have also tried hard-coding speed and duplex at each location to no avail. Setting 100 M half actually didn't work on the "new" end of things so I didn't continue to the old.
Any help is greatly appreciated. Thanks in advance!
pc1(on new network)<-->access switch<-->distribution switch(layer 3)<-->firewall<-->router<-
Everything is Gigabit ethernet. When I try to copy a file from pc1 to pc2, it is relatively fast at about 4MB/sec. When I try to copy the same file back from pc2 to pc1, it is atrociously slow at about 100KB/sec. I am able to test everything in a duplicate test environment and cannot duplicate the issue. I suspect that it is the router because the router that is the only thing I don't have access to (because it is managed by our ISP) in the production environment and can verify in the production environment everything on either side of it works just fine.
I have also tried hard-coding speed and duplex at each location to no avail. Setting 100 M half actually didn't work on the "new" end of things so I didn't continue to the old.
Any help is greatly appreciated. Thanks in advance!
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Plus: what if you introduce another device (simplest: a notebook) first to new net, then to old net, testing transfer to/from both pcs? (you may get some surprise, e.g. pc2->nb being too slow even if pc2 and nb are in same network)
ASKER
@Surrano Yes, file transfer is the problem not ping connectivity. Traceroutes complete in both directions and in the inverse order. All are equal to or under 1 millisecond.
New devices on either side act the same way as the original devices.
New devices on either side act the same way as the original devices.
ASKER
@WalkaboutTigger The following is what I can guess is the traffic shaping config on my ISP managed router:
And this is the policy-map:
I only see on the multilink interface any mention of "service-policy output etm" which I would think (quite possibly incorrectly) is only going to shape the traffic going across that multilink interface (across our MPLS). The GigE lan ports do not mention any service-policy on the interfaces. Does that help any in ruling out traffic shaping as a potential issue or does it point more toward it as being the problem?
I am currently reading up on traffic shaping to see if I can decipher one way or another as well.
Thanks again to you both for the assistance so far.
class-map match-any ef
match dscp cs5 ef
class-map match-any be
match any
class-map match-any af4x
match dscp cs4 af41 af42 af43
class-map match-any af3x
match dscp cs3 af31 af32 af33 cs6
class-map match-any af2x
match dscp cs2 af21 af22 af23
class-map match-any af1x
match dscp cs1 af11 af12 af13
And this is the policy-map:
policy-map etm
class ef
set ip dscp ef
priority 3012
class af4x
bandwidth remaining percent 40
random-detect
set ip dscp af41
class af3x
bandwidth remaining percent 39
random-detect
set ip dscp af31
class af2x
bandwidth remaining percent 16
random-detect
set ip dscp af21
class af1x
bandwidth remaining percent 1
random-detect
set ip dscp af11
class be
set ip dscp default
bandwidth remaining percent 3
random-detect
I only see on the multilink interface any mention of "service-policy output etm" which I would think (quite possibly incorrectly) is only going to shape the traffic going across that multilink interface (across our MPLS). The GigE lan ports do not mention any service-policy on the interfaces. Does that help any in ruling out traffic shaping as a potential issue or does it point more toward it as being the problem?
I am currently reading up on traffic shaping to see if I can decipher one way or another as well.
Thanks again to you both for the assistance so far.
ASKER
Would a multilink interface being down possibly negate any service policy that would otherwise being effecting other interfaces? This is again copied from the config of the router supplied to me from my ISP.
interface Multilink1
description Possible cause of slowness
ip address 111.111.111.111 255.255.255.252
no peer neighbor-route
ppp chap hostname blah
ppp multilink
ppp multilink links minimum 1
ppp multilink group 1
ppp multilink fragment disable
!
service-policy output etm
!
ASKER
It also is making sense to me that this is happening but I don't know why. Mainly because the "show policy-map etm" on my test router is showing a default bandwidth of 3012kbps which would compare to the 3-4Mb/sec I see in the file transfer in the one direction that it isn't really bad.
ASKER
The issue I was seeing was actually due to a wan optimization device between the router and layer 2 unmanaged switch that was throttling the network speed down to 8mbps. That was removed and speeds picked up to what was expected.
ASKER
The reason I am not connecting the old subnet to the distribution switch is I was trying to avoid any situation that I make part of the change and something doesn't work. If I keep them separated like this, it is very easy to go back even partially to the old network.