Effect of distance on transfer speed (increases over time)

networking

I'm in Europe and I'm running a web site in the Unites States.

The service provider where the site is hosted offers large up/down bandwith (at good prices).

The server now have very low load (on network/cpu/storage)

As the server is far from me and the route the data have to take is long I have a quite high response time (ping gives 150 to 200 ms) but this is acceptable for me.

Data transfer speed is also good due to large upload bandwidth offered by the provider.

However it seem to me that data transfer starts slow and then reach its peak speed after a few seconds.

Questions:

Is this an effect of the long route/distance too?

What causes this "inertial" behaviour of the data transfer speed?

Best Answer

However it seem to me that data transfer starts slow and then reach its peak speed after a few seconds.

This is expected behavior, especially on connections with higher latency. Look up TCP Slow Start and TCP Window Scaling for more info on this.

Is this an effect of the long route/distance too?

No, it's an effect of high latency. Latency and geographic distance are indeed correlated, but aside from from the speed of light in fiber, there is no strong correlation between the two. There are many factors that play into latency (congestion, number of EDFAs in line, number of route hops, etc.), and distance is only a small part of that.

What causes this "inertial" behaviour of the data transfer speed?

When a TCP socket opens, the network stack errs on the side of being conservative about how much data it will send before receiving acknowledgement from the other end of the socket that the packets have been received. As the socket ages, and more information is known about the quality of the connection, the network stack can increase the Window Size, which allows more packets to be "in flight" before receiving acknowledgment.

Related Solutions

Slow Transfers over Distance

Making sure the TCP window is opening up wide enough to cover the Bandwidth Delay Product would have been my first guess too. Assuming that is configured properly (and supported by both ends) I would next examine a packet trace to make sure that the window really is opening up and that one of the hops in the path isn't stripping the window scaling. If that is all good, and you are certain you are not banging into a bandwidth constrained hop in the path, the likely cause to your problems is random packet drops. This hypothesis is supported by the indication of the duplicated ACKs you mentioned. (Duplicated ACKs are generally a direct result of lost data). Also note that with a large bandwidth delay product and therefore a large open sliding window, even low levels of random packet drops can significantly hamper the total throughput of the connection.

Side Note: For bulk data transfers over TCP and over a multi-hop WAN connection, there should be no need or reason to disable Nagle. In fact, that exact scenario is why Nagle exists. Generally, Nagle only needs to be disabled for interactive connections where sub-MTU sized datagrams need to be forced out without any delay. ie: For bulk transfers, you want as much data in each packet as possible.

Need advice about network distance & speed

Based on Distance I would go with Fiber between the sites to make sure that you have a maintained speed. Although you can put switches in between that really isn't the best solution. Below is a list from DSL Reports FAQ on cable lengths. As you can see to hit your 200 Meter requirement you should really be running fiber. This will eliminate the need for another repeater or switch. This can be easily accomplished with switches that support fiber or have an SFP port that a fiber SFP can be put into. An example of such a switch would be the Cisco SG200-26 24 Two of these ports have sfp adapters that you can put an SFP adapter that matches the type of fiber you have run between the buildings.

COPPER (UTP & COAXIAL) LENGTHS:

Gigabit Ethernet (over copper), 1000baseT
Length: 100 Meters
Cable: UTP (Cat5 and up ONLY)

Fast Ethernet, 100BaseTX
Length: 100 Meters
Cable: UTP (Cat5 and up ONLY)

Twisted Pair Ethernet, 10BaseT
Length: 100 Meters
Cable: UTP (Cat 3 and up)

Thin Ethernet, 10Base2
Length: 185 Meters
Cable: RG-58 type coax

Thick Ethernet, 10Base5
Length: 500 Meters
Cable: RG-58 type coax

FIBRE OPTIC LENGTHS:

10BaseFL (Multimode fibre)
Length: 2000m or 2km

100BaseFX Half-duplex (Multimode fibre)
Length: 412m

100BaseFX Full-duplex (Multimode fibre)
Length: 2000m or 2km

100BaseFX Full-Duplex ONLY (Singlemode fibre)
Length: 2-40,000m or 2-40km *(see note)

1000Base-LX
Length: 550m (62.5 or 50 multimode fibre)
3000m (10 singlemode fibre)

1000Base-SX
Length: 250m (62.5 multimode fibre)
550m (50 multimode fibre)

1000Base-CX
Length: 25m (requires new type of shielded cable)`

Best Answer

Related Solutions

Slow Transfers over Distance

Need advice about network distance & speed

Related Topic