FTP get scenario
I'm conducting some tests on TCP extensions and Cisco's RBSCP for poor, long delay bandwidth WAN links (768 kbps at 1200 ms with BERs from 10-5 to 10-8). Typically, FTP transfers range from 2 to 60%. The better performance being at lower BERs (10-8). What is really strange is that FTP gets are quicker than puts in all cases. What gives? Also, any ideas on improving the FTP flows? I've tried adjusting the TCP stacks in Windows, but with limited success. Also, I'm finding that Cisco's RBSCP only gives me a small performance increase (10-15%).

    Requires Free Membership to View

    Login

    By submitting your registration information to SearchNetworking.com you agree to receive email communications from TechTarget and TechTarget partners. We encourage you to read our Privacy Policy which contains important disclosures about how we collect and use your registration and other information. If you reside outside of the United States, by submitting this registration information you consent to having your personal data transferred to and processed in the United States. Your use of SearchNetworking.com is governed by our Terms of Use. You may contact us at webmaster@TechTarget.com.

In an FTP get scenario, there is less that happens from a security standpoint. In a put scenario, all security must be evaluated to determine rights and to make sure that the file can be reassembled with proper file check sum packets. The speed of the machine doing these operations will have an effect on how quickly they occur. Also in some network protocol configurations both on the network card, operating system software, etc., there may be a difference in packet receive buffers and packet send buffers. If a system is optimized for one over the other, there will be a difference.

On circuits with a BER of 10–5, this simply tells the network the acceptable level of errors. These errors may cause retransmissions, etc. It is hard to determine the best balance in a live environment as other factors may come into play during the testing of sends and receives.

Also, some circuits are asymmetrical, meaning that the circuit itself is optimized for either inbound or outbound traffic. I would suggest working with your circuit carrier to be sure that both sides are equal, and you will want to test with exactly the same configuration on both sides of the link. This would include routers, machines, operating system revisions, virus scan software and firewall software. Some machines (for instance on your PUT side) may be scanning each portion of the file as it arrives, while the get machine may only scan the complete file.

As you can see, there are several factors that will affect performance on both ends. The best bet is to set up a testbed where you can control other traffic and better monitor each end. RMON and other performance software will also help. If you send the same file both ways, is there a large discrepancy in the number of packets or octets in the transmission? This would also be a good thing to look at.

As for the Cisco software, it may need some configuration. Vary few packages will work for all companies straight out of the box. I would suggest talking to Cisco TAC to determine if there are some parameters that you can tune to optimize the effectiveness of the product.

This was first published in September 2005