Hot Spots: Step one for DR: Server virtualization

When it comes to disaster recovery (DR), the 80/20 rule typically applies: 80% of a DR budget is allocated to 20% of the systems in the environment. While many of the applications run by organizations are business critical, often only the top tier of servers are protected by failsafe mechanisms such as clustering, synchronous replication and other continuous availability solutions. This leaves the majority of lower tier systems underinsured.

Interruptions such as a server or disk failure, software fault, data corruption, computer virus, natural disaster or manmade errors can occur at any time. These interruptions affect not only the bottom line but the IT organization, which has to allocate time and resources to get the system up and running as quickly as possible. Acquiring replacement hardware, rebuilding the system and recovering data is a time-consuming process, especially if recovery is dependent on tape-based backup.

Server virtualization technology can help. It offers considerable benefits for organizations that want to implement remote replication for DR purposes, including reducing recovery time objectives (RTOs) and the costs associated with remote site server infrastructure, and improving ease of implementation of remote DR. Research from Enterprise Strategy Group (ESG) finds that 26% of organizations are replicating virtual machine images to a remote DR site and another 39% plan to do so.

The economics of DR
Server virtualization is typically equated with running multiple virtual machines on a single physical server for improved resource utilization and hardware consolidation. Improved business continuity and lower data center operating costs are additional, compelling benefits that are fast becoming the primary drivers of server virtualization deployments. Today, the closest IT organizations can come to guaranteeing a service level for a specific application in a physical server environment is to deploy infrastructure and applications as dedicated silos that are isolated from any potential outside impact. However, this type of implementation leads to poor utilization and the likely purchase of additional hardware for high availability. But server virtualization can enable the sort of business resilience you need without the compromises that come with silos. When talking server virtualization and business continuity, there are a few key concepts to understand.

1. Hardware independence. Server virtualization provides a layer between the physical hardware and the operating system and applications, allowing virtual machines to run on any approved x86 hardware.



2. Isolation. Multiple virtual machines run in an isolated environment, independent from one another. If one virtual machine goes down, it doesn't affect others running on the same host. Because multiple operating systems can be run on one physical machine, hardware utilization is improved. This creates a cost-effective means of establishing failover targets in DR strategies.



3. Encapsulation. Virtual machines are stored as hardware-independent files. Encapsulating the virtual machine into a single file enables mobility and allows multiple copies of the virtual machine to be created for DR purposes.

In a physical-to-virtual (P2V) configuration, a production application running on a physical machine is replicated to a virtual machine running on a secondary system. A virtual-to-virtual (V2V) scenario has production and recovery applications running on virtual machines. By periodically or continuously replicating over a LAN or WAN to a virtualized standby system, a higher degree of synchronization is achieved. With greater frequency of replication, data loss is minimized in the event of an outage.

A virtualized DR site
Consolidating physical servers at the DR site eliminates the need for one-to-one mirrored physical machines and can deliver savings in power and cooling, management and maintenance. The portability of virtual machines also means that failback can occur to the original location or to a new one: either a virtual machine or a physical system with dissimilar hardware.

Another advantage is the ability to test DR more efficiently. Regular DR testing--confirming failover and failback processes, and testing documentation--is necessary to ensure DR preparedness. In server virtualization environments, DR tests can be performed on a virtual machine without affecting another machine or the primary production system. Tests can be run in more realistic situations (during peak hours, with peak-time staff and on actual DR hardware) without impacting normal operations and with less risk should something go awry.

P2V- or V2V-based replication offers a big improvement over tape-based backup/recovery, especially when it comes to recovery point objectives (RPOs). With tape, an organization could lose up to a full day's data because tape backup usually occurs once per day. With replication, incremental changes are captured as they happen, and can then be replicated to one or more local or remote virtual machines. Tape-based backup is more affordable, but the recovery process is slower, more manual and error-prone. Replication also lets you effectively meet budget and recovery requirements.

More sophisticated clustering solutions take immediate action by starting up apps at the DR site and making the application available to users. Clustering meets aggressive recovery objectives, but can be more complex and costly to implement and manage. Clustering solutions have moved into the virtual world and can protect virtual machines in much the same way they protect physical machines.

With server virtualization, remote replication becomes cost-effective for organizations that haven't previously implemented DR or for tiers of application servers that may not have justified aggressive RTO objectives, but would still be painful to manually recover.

Categories of replication strategies
Host-based replication solutions, such as CA XOsoft's Replication (formerly WANSync) or Double-Take Software's Double-Take, run on servers (or virtual machines in server virtualization environments) and asynchronously replicate data over an IP network from/to the disk arrays of choice. The challenge with this approach is that host-based agents may put an unnecessary load on the production systems the replication application protects, especially those in high transaction-intensive environments.

Network-based approaches, such as EMC's RecoverPoint or InMage Systems' DR-Scout, leverage a network-resident appliance or intelligent switches. The benefits are heterogeneous hosts and arrays (more flexibility and less vendor commitment), scalability and no performance impact on production systems. Disadvantages include the cost of the appliance or the investment in SAN fabrics that support embedded solutions.

Array-based products, including EMC's SRDF and Hitachi Data Systems' TrueCopy, reside on disk storage systems and replicate data between arrays without impacting host systems. Characterized as costly and array specific, array-based products don't allow source and target hardware to be mixed and matched.

Some replication products offer continuous data replication (CDR). These solutions maintain historical replication data--multiple, remote replica versions at a DR location. This capability protects against situations where a corruption is replicated and allows IT organizations to roll back to a known consistency point in time. Vendors offering CDR capabilities include CommVault and Double-Take.

Other replication products offer continuous data protection (CDP) as a feature. This capability captures data continuously, offering infinite points of recovery to a previous point in time. CDP also delivers more fine-grain recovery points than a CDR solution, as well as granular recovery of data (individual items, such as a message, mailbox or a database table). Replication vendors offering CDP capabilities include BakBone Software, CA XOsoft, EMC, FalconStor Software and InMage Systems.

When deciding on a strategy or vendor, remember that the goal of remote replication for DR is to mitigate risk. Preventing downtime, as well as responding to a disaster, is revolutionized with the introduction of server virtualization in DR environments. Server virtualization technology reduces the cost and complexity traditionally associated with remote replication for DR and DR testing, enabling companies of all sizes to introduce DR where it didn't exist before.