Farewell to Data Loss: Understanding Data Replication
Data protection and continuous availability are top priorities of IT managers and C-level executives alike, but as companies allocate additional resources to ensuring 24x7 access to data and applications, the question of what strategy and products to employ emerges. With that in mind, this article will help you better understand the role of off-site replication technology in a data protection solution, the difference between synchronous replication and asynchronous replication and how asynchronous replication works.The Transition
Relying on tape back-up alone is no longer adequate. Despite increases in megabytes-per-minute speed, tape back-up technology has not kept up with the growth in server size causing further increases in restoration times. For example, years ago, with traditional tape-backup technology a typical 50-gigabyte server could be restored in about eight hours at a rate of around 10 MB/min. Even with the best tape-backup available today, restoration for a 1-terabyte server at a 900 MB/min or 52 GB/hour rate will still take approximately 19 hours, meaning the application and server will be unavailable to users during that time. Such a situation could potentially threaten service level agreements (SLA) and customer relationships.
Recognizing the restrictions of only relying on tape back-up, companies today are integrating replication technology to maintain real-time copies of data and applications at one or more off-site locations. A company headquartered in New York, for example, may transfer data to both a New Jersey location that employees can reach relatively quickly and to a rural desert location hundreds of miles away. Validating this evolving protection strategy, analyst firm Gartner Group forecasts that by 2003 75 percent of large enterprises will be combining disk-based data replication and tape-based technology for rapid application recovery.Technology Alternatives
As companies research and evaluate products to augment their tape-back up strategies with replication technology, it is important to understand the level of protection a business's applications require. If Automated Teller Machine (ATM), stock or other critical transactions cannot be lost under any circumstances, and cost is not a concern, synchronous replication is the route to go. It is vital to note, however, that the "zero loss" associated with synchronous replication is expensive and can reduce overall application performance.
Maintaining "zero loss" through synchronous replication requires a two-phase commit approach where each write to a disk block is written to and acknowledged by the target drive and then written to the source drive, and finally committed to both before any subsequent read or write input/output (I/O), the transfer of information between devices, can be processed by the disk subsystem. The performance penalties that may emerge can become proportionately greater as the distance between the systems increase because communication speed is limited by the speed of light, in the best case. Often times once network protocol and routing latency are factored in, it is much slower.