Replication is an essential part of the data protection strategy for many organizations. However, before an organization can benefit from data replication, it must determine whether it is better to perform hardware replication or to replicate data at the software level. There are advantages and disadvantages to each approach.
Software-based replication generally costs less than hardware-based replication and offers a higher degree of flexibility. In some cases, an organization might already have everything it needs to begin using software-based replication without having to buy anything extra. Windows Server 2012 and 2012 R2, for example, include a built-in replication feature for virtual machines.
Whether an organization performs data replication at the hypervisor level or as a function of its backup product, the software provides a level of abstraction that allows replication to occur without the need for matching hardware. As such, using software-based replication is a great way to avoid vendor lock-in.
A replication solution that does not take application state into account may lead to problems with data consistency across application servers.
Although some organizations might not be put off by the idea of hardware vendor lock-in, they have to take into account some additional considerations. First, when it comes to hardware replication, storage vendors typically require matching hardware in the primary and recovery sites. This requirement may increase the cost of the replication solution because a storage array that is a good fit for the primary site might be overkill in a disaster recovery site. While the primary site may need a high-performance storage array that can deliver a high level of read IOPS, there might not be a need for high performance in a disaster recovery site.
Another disadvantage to using hardware-based replication is that it undermines the gains that have been made through virtualization. Server virtualization products allow storage to be abstracted, thereby allowing an organization to mix and match storage products in a way that makes the most sense for them. This is particularly helpful when an organization maintains multiple storage arrays in each site and not all of those arrays are from the same vendor. Storage can still be virtualized even when hardware-based replication is in use, but vendor lock-in mitigates the advantages of being able to mix and match storage products from multiple vendors.
Perhaps the biggest advantage that software-based replication offers is application awareness. Hardware replication solutions typically copy storage blocks from one array to another without regard to what those blocks actually are. The problem with this approach is that some database applications maintain transactions in memory for a period of time, and copying storage blocks without regard to memory contents can leave the database in an inconsistent state. Never mind the fact that it is becoming increasingly common for applications to span multiple servers, and a replication solution that does not take application state into account may lead to problems with data consistency across application servers.
Although it may at first seem that software-based replication has a clear advantage over hardware-based replication, there are certain advantages to performing replication at the hardware level. For one thing, hardware-based replication tends to be easier to troubleshoot when things go wrong. Because replication occurs between two hardware arrays, there are fewer pieces to the puzzle then there would be with software-based replication, and that greatly simplifies the troubleshooting process. Furthermore, it is easier to get technical support for hardware-based replication issues because the organization does not have to worry about vendors pointing fingers at one another.
Another advantage to hardware-based replication is that certain processes can be offloaded from your production servers and performed at the storage hardware level instead. For example, most software-based replication offers deduplication capabilities. However, because deduplication is occurring at the software level, the deduplication process may place an additional load on your production servers.
On the other hand, hardware-based deduplication can occur without impacting production workloads. Furthermore, most storage products offer a form of global deduplication that eliminates cross-server redundancy and can go a long way toward making the replication process more efficient.
Neither hardware- nor software-based replication alone is an ideal solution that is a perfect fit for every organization. Each approach has advantages and disadvantages that must be considered prior to making a purchasing decision. One way to overcome the inherent limitations of each approach and achieve a best of both worlds replication solution is to look for a vendor that offers hardware- and software-based replication products. Many storage vendors have backup products that provide capabilities that cannot be implemented natively through hardware. Of course, the disadvantage to using such a solution is that vendor lock-in becomes even more prevalent.
About the author:
Brien M. Posey, MCSE, has received Microsoft's MVP award for Exchange Server, Windows Server and Internet Information Server. Brien has served as CIO for a nationwide chain of hospitals and has been responsible for the department of information management at Fort Knox. You can visit Brien's personal website at www.brienposey.com .