Many papers have been written on the effect that limited bandwidth and high latency have on application performance across the Wide Area Network (WAN). The purpose of this document is to address a less commonly understood WAN challenge that can also affect the performance of critical business applications–packet loss and reordering.

While packet loss and out of order packets are a nuisance for a network that supports typical data applications1 like file transfer and email, it is a very serious problem when performing data replication and backup across the WAN. The former involves thousands of short-lived sessions made up of a small number of packets typically sent over low bandwidth connections; the latter involves continuous sessions with many packets sent over high capacity WAN links. Data applications can typically recover from lost or out of order packets by retransmitting the lost data. Performance might suffer, but the results are not catastrophic. Data replication applications, however, do not have the same luxury. If packets are lost, throughput can be decreased so significantly that the replication process cannot be completed in a reasonable timeframe - if at all.

This paper will discuss why packet loss and ordering is becoming a bigger problem in today's WANs, and what can be done to overcome these packet delivery challenges. To achieve this goal, this document will integrate theory and practice. In terms of theory, this document will include a widely accepted model that identifies the impact of packet loss on the throughput of a TCP stream. In terms of practice, this document will include input from two IT professionals who have a lot of experience with data replication. One of these professionals is a technical business consultant at a storage company. The other is a manager of network services for a company that does hosting in multiple data centers. These two professionals will be referred to in this document as The Consultant and The Manager respectively.