Friday, July 31, 2009

The improved reliability of backup has created, to some extent, a false sense of security. Many customers are simply unaware of the time it takes to do a recovery from a backup system. There is an assumption that because the data is on disk that recoveries will be fast enough. For single file or folder recoveries, that is often the case.

When you need fast recoveries the most, however, the time required to restore that data through the backup application, across the network and onto a waiting server may not be fast enough. With any number of potential issues–a corrupted database, a downed server or a lost site, for example–this process can take hours or even days. A big part of why this scenario takes so long is because the data as it is stored by the backup application is not in a usable format. Instead, it’s in the application’s proprietary format–even if it’s backed up to disk. Recoveries, therefore, have to go through the backup application before the data is sent to the client server’s storage.

On the opposite end of the spectrum are array-based mirroring or replication tools. These are expensive solutions that require nearly identical systems at the DR and primary sites. And in most cases, these protection points are not granular enough. Typically, you can choose between real-time replication, where the secondary site is updated in near real-time, or replication based on snapshots. These snapshots usually are not taken frequently enough since most arrays limit the number that can be maintained. In either case, storage is consumed to save these protection instances and it must be done on a ‘like’ storage platform. As a result, these solutions soon become too expensive to be used throughout the data center and only a few applications can justify this expense.

In our briefing, Overland Storage discussed its new solution for solving this problem: the REO Business Continuity Appliance (BCA). The unit is designed to solve what we call the “Data Protection Gap”–or the space between real-time mirroring or replication and standard backup. This system provides real-time data protection by copying data as it changes from the protected server to the recovery server through the BCA.

This gap is especially problematic for small- to medium-sized businesses (SMBs), which typically cannot justify investments in real-time replication or a second ‘like’ storage system that often is required. Overland's BCA focuses on filling that gap and addressing the “WHEN” of data recovery in a cost-effective manner by not requiring identical storage system or additional expensive software.

In addition, REO BCA addresses some of the application and server awareness issues that standard replication solutions suffer from, such as complexity and cost. At SMBs, the IT staff often wears many hats and doesn’t have the time to assemble all the components typically required to handle this problem. Even if the SMB is purchasing through a reseller, there is limited budget for professional services. Therefore, SMBs need a solution that can be installed quickly and painlessly.

To that end, Overland has created an appliance that simplifies much of the work. REO BCA will ship in two configurations, which are based on the number of servers that need protection. The appliance comes with all the ports needed for connectivity and initial storage to begin storing the redundant data.

On the protected servers, two lightweight agents need to be installed. They provide the volume level CDP, application consistency and application failover support. Additionally, the agents work with the BCA to provide centralized scheduling and failover, byte level differencing for space optimization and group scheduling that allows for pre- and post-workflow support.

In the diagram, each of the three production servers require installation of lightweight agents to begin replication to the REO BCA. The unit will then use a logging method to provide point-in-time protection of the environment. Also, in this diagram is a local recovery server that, as an example, is a virtualization host. On failure, it can assume the role of any or all of the protected production servers.

The most important part of this scenario is how the BCA can solve the "WHEN" question. The "WHEN" is now. If the solution is designed as described in the diagram, then recovery is near-instantaneous. This means no data needs to be read into the backup application and restored across the network. The recovery server can be immediately available with the last application consistent copy of the replicated data. 

Downtime is just as critical to SMBs as enterprise data centers. No matter the size of the business, the need to be prepared for a disaster will not wait until economic conditions improve. The turnkey nature of the Overland Storage BCA solution is something that any SMB should consider as it brings enterprise-class data protection to the mid-market and helps make fast DR affordable.