> View this White Paper now

Published on: 09/01/2008
Type of content: White Paper
Format: Adobe Acrobat (.pdf)
Length: 4
Price: FREE

Overview
Data replication across geographically dispersed data centers with minimal risk of downtime is a growing requirement for many businesses. Distances between data centers continue to expand due to data center geographical placement considerations, increased regulations, and risk management. There is an increased business sensitivity to data loss, and risks that might have been acceptable ten years ago are no longer acceptable. In addition, the volume of data being replicated is growing dramatically, leading to increased data replication traffic being transmitted over the WAN. Add to all of this the ever-present budget limitations, and it can make data replication solutions quite complicated.

The performance of most data replication solutions depends heavily on the performance of the WAN. WANs, however, can be notoriously fickle with varying latency, packet loss, and congestion levels over time. As a result, most companies feel the need to over-purchase their WANs, which can help with performance, but can result in much higher expenses than necessary. According to a recent Forrester Research report, WAN bandwidth can represent as much as one-third of the total replication project's ongoing costs. It's hard for business to anticipate the need for increased bandwidth, and replication solutions can require bandwidth to be sized to the maximum volume of data activity at any given time, rather than the average volume, leaving a lot of bandwidth unused.

In addition, businesses expect tighter and tighter RPO and RTO, which puts a heavier burden on existing WAN infrastructure and often creates a bottleneck to improving data replication performance and reliability over the WAN. Dedicated WAN replication links are becoming less common, so important replication traffic often competes for bandwidth with less important types of traffic.

The key characteristics that affect WAN performance include latency, packet loss rates, congestion, and amount of available bandwidth. Unfortunately, these characteristics can change over time and adversely affect the performance of what was originally a sound DR plan. In addition, when the DR application shares the WAN with non-replication traffic, file transfers, and other application traffic, the RPOs and RTOs that were previously attainable can become out of reach. Many customers resign themselves to living with lower than desired performance. One fix often attempted is to replicate only the most critical data and therefore reduce the amount of data replicated. Another common solution is to lease additional bandwidth. Neither is ideal because they both either significantly increase business risk or costs.

> View this White Paper now