Issue link: https://insights.oneneck.com/i/1190444
9 DISASTER RECOVERY GUIDE – Powered by ZERTO Available DR Solutions A number of disk-to-disk based DR solutions haven been introduced over the years, none of them fully virtualization-aware. In a short overview we will summarize their structures and their shortcomings in a virtualized environment. Array-based Replication Array-based replication products are provided by the storage vendors and deployed as modules inside the storage array. They are single-vendor solutions, compatible only with the specific storage solution already in use. The relationship between the VM and storage is fixed and the entire LUN is replicated, whether it is 40% or 90% utilized. • Hardware-defined – Array-based replication is designed to replicate physical entities. It doesn't "see" the virtual machines and is oblivious to configuration changes. • Not independent – Though optimized to work with the existing storage array, it locks in the organization to a single vendor. • More management points – In addition to the physical storage array's management console, IT needs to manage virtual assets from a virtualization management console as well. • Growth and change – The relationship between the VM and storage is fixed, eliminating the flexibility of virtualization and removing the ability to respond to evolving business needs. • Granularity – Replicating the entire LUN, array-based replication lacks the granularity needed in a virtual environment. • Costs – The entire LUN is replicated, whether it is 40% or 90% utilized, increasing power, cooling, networking and storage costs. • Single point for recovery – Many array-based solutions do not have the ability to store a history of the performance of the LUN. With that limitation, if the last data point was corrupted, that is what the business has to use for recovery, rendering the DR solution useless. • Time – Recovery is very time-consuming and complicated as there is no automation so the VMs and applications have to be built from scratch. Appliance-based Replication Appliance-based replication solutions are also hardware-based and specific to a single platform. The main difference is that replication runs on an external, physical appliance instead of inside the storage array itself. This makes it more flexible and less- consuming in array resources. But the disadvantages are more or less the same as for array-based replication. • Hardware-defined – It is also designed to replicate physical entities rather than virtual entities. • Not independent – Though it is more flexible than array-based replication, it is still specific to a single platform. • More management points – Appliance-based replication requires dual points of management: the physical management console and the virtualization management console. • Growth and change – It doesn't "see" configuration changes. As a result, BC/DR plans will be out of sync with the current production environment, eliminating the flexibility of virtualization and removing the ability to respond to evolving business needs. • Granularity – Appliance-based replication focuses on the logical unit rather than the virtual machine. This lack of granularity conflicts with the requirements and promise of virtualization. • Costs – Since the entire LUN is replicated, costs for power, cooling, storage and networking increase. VM VM VM VM VM VM VM VM VM VM VM VM VM VM VM VM VM VM VM VM DR site Production site Figure 1. Array-based and appliance-based replication requires coordinating two replication products, both for the physical and the virtualized environment. This increases management complexity and undermines the investment made in virtualization.