Very slow transfer speed using sftp or ftp

[Anthony Nova]

We are having an issue of having very slow transfer speed which is effecting our backups. Using sftp I see kb/s as high as 500kb/s when it first kicks off then it slowly decreases to a crawl to 5kb/s then finally stalling. It then fluctuates from 2kb/s to 10kb/s the rest of the transfer.

Syslog and lanadmin are not showing any errors. We have checked that the interface is running at 100/full w/auto neg off. The switch also is at 100/full. Not sure what has happened but ever since this past weekend this has started to happen. Any ideas as to what it can be? Thxs!

PPA Number = 0
Description = lan0 HP PCI 1000Base-T Release B.11.11.22
Type (value) = ethernet-csmacd(6)
MTU Size = 1500
Speed = 100000000
Station Address = 0x303ef816e4
Administration Status (value) = up(1)
Operation Status (value) = up(1)
Last Change = 205
Inbound Octets = 242021882
Inbound Unicast Packets = 199340
Inbound Non-Unicast Packets = 1
Inbound Discards = 0
Inbound Errors = 0
Inbound Unknown Protocols = 0
Outbound Octets = 41027866
Outbound Unicast Packets = 138162
Outbound Non-Unicast Packets = 1
Outbound Discards = 0
Outbound Errors = 0
Outbound Queue Length = 0
Specific = 655367

Ethernet-like Statistics Group

Index = 1
Alignment Errors = 0
FCS Errors = 0
Single Collision Frames = 0
Multiple Collision Frames = 0
Deferred Transmissions = 0
Late Collisions = 0
Excessive Collisions = 0
Internal MAC Transmit Errors = 0
Carrier Sense Errors = 0
Frames Too Long = 0
Internal MAC Receive Errors = 0


Speed = 100 Full-Duplex.
Autonegotiation = Off.

[Ivan Ferreira]

From this side seems to be correct, autonegotiation is disabled and input/output errors are not showed.

Anyway, ensure that the switch speed configuration match the server configuration and autonegotiation is disabled. Also, check the other host for problems. Do the same test from host1 to host2, host1 to host3, host2 to host3 to try to identify the host that presents the problem.

[Ben Dehner]

The above information seems to indicate the problem is not on the host NIC. Few broadcast packets, no errors, discards, or collisions. As Ivan mentioned, make sure all hosts/switch ports have consistent speed settings. Make sure everything is fixed to 100/Full Duplex if you don't like autonegotiate. Every port doesn't have to be the same, but every port has to be consistent with whatever device is connected to it.

You didn't mention what type of switch you were connected to. If it is any type of managed switch, see if there is any sign of errors or dropped packets on the switch. It may be throwing away packets before they get to the host, forcing retransmits. Depending on the switch, check for other problems such as CPU or memory as well.

Also, depending on the environment, check the MTU size. If the hosts are on the same LAN segment -- no routing -- then MTU is not a factor. A pure layer 2 switch usually doesn't care about MTU, but some of the new layer 3 switches do. Also, if this is a routed network, that brings in another layer of device(s) that could be causing problems.

If you can, maybe run a network sniffer and look at the packets going between the hosts. Specifically, look for lots of TCP retransmits, as this indicates someone is loosing packets somewhere.

Hope this gives you some places to look.

[Phil.Howell]

You could use ttcp or ttcpw to get stats on send/receive from each end of your connections.
This usually reveals any problems with switch configuration.
Phil