Communications on Degraded Networks
Certain applications, especially those involving space communications links, require special network protocols for reliable and efficient operation. The NetAcquire server contains advanced networking technology to dramatically improve network performance in a variety of specialized network environments. Plain TCP, in particular, performs poorly over space communications links because of the link’s long propagation delay and significant bit errors.
Limitations of TCP
A key limitation with TCP in high bit error networks is the lack of error correction capabilities. Since TCP cannot correct bit errors, if even a single bit within a packet is corrupted in transit, the receiver will discard the entire packet. This turns bit errors into packet loss.
In addition, TCP can recover from the loss of only one packet per round trip. If the networks round-trip time is 500 milliseconds, then TCP can tolerate only one packet loss per 500 milliseconds. To illustrate the implication of this limitation, consider what happens if this network has a bit error rate of 10-5: TCP can send data at a maximum rate of 200 kbps, no matter how fast the physical network is!
Another limitation of TCP is the strength of its data corruption protection. TCP uses a relatively weak checksum scheme to detect bit errors in each packet. The approach fails to detect bit errors relatively often at high bit error rates, allowing corrupted data to be delivered to the application undetected!
NetAcquire TCP Enhancements and SCPS
NetAcquire Corporation has enhanced TCP by implementing three network protocol extensions: increased TCP buffer sizes, Forward Error Correction (FEC), and Space Communications Protocol Specification-Transport Protocol (SCPS-TP). Large TCP memory buffers ensure that maximum throughput is achieved even when network delay is long (up to 1000 milliseconds). FEC improves TCP by including a Reed-Solomon Forward Error Correction code in each packet. Most packets that become corrupted in the network reach the destination host, where any errors are corrected and the data delivered to the application. This is a substantial improvement over TCPs policy of simply discarding corrupted packets. Furthermore, the powerful Reed-Solomon scheme ensures that corrupted data will not be delivered to the application.
SCPS-TP is complementary to FECit and recovers quickly when packets get lost in transit. SCPS-TP can recover from the loss of any number of packets in each round-trip time (subject only to bandwidth limitations). This is a substantial improvement over the loss of only one packet per round-trip time that TCP can tolerate.
Using these two extensions together, the NetAcquire server can maintain excellent throughput in even the poorest network conditions.