The Disaster Recovery Movement
The business drivers for disaster recovery today are far more pervasive than they were a decade ago. The combined impact of compliance with industry specific rules, governmental regulations such as HIPPA-SOX and global standards has placed increasing pressure on companies of all sizes. At the same time, today’s CIOs must cope with shrinking IT budgets while providing more favorable asset utilization and more efficient use of IT resources to better address the growing demands of their clients. Ensuring continuity of operations, adhering to service level agreements, maintaining customer satisfaction and business agility all continue to challenge the enterprises of today’s global economy. However, in this changing world where negative media criticism spreads rapidly through the digital circuit and incites public outrage every time a company loses data, disaster recovery has become an escalated concern for CIOs everywhere – particularly those running Fortune 1000 operations.
The Evolution of Disaster Recovery Technology
Disaster recovery has gone through some dramatic changes over the last decade. We are barely out of the “sneakerware” era, where running memos down the hall was once the norm. At the close of the 20th Century, fledgling technologies such as local and remote hardware-based replication began appearing in enterprises. These were typically difficult to implement, and the high price point meant that that only a few could afford these functions and fewer had the expertise to implement them. These technologies have improved dramatically in terms of deployment, performance and manageability. Today, most storage vendors have disaster recovery capability and offer a solid, easy-to-manage solution with their own storage products. Because they are based on proprietary technology, they lock customers into a single vendor. Licenses must be purchased for each storage device, and pricing can be complex (e.g., based on number of TB).
A number of pure software products also evolved over this period. These were effective solutions designed for less robust disaster recovery implementations since they were slower than other alternatives at that time. Software options have become more robust over time; however, they still fall short in critical areas such as performance and interoperability.
A third approach focused on network-centric disaster recovery solutions. This approach carried strong attributes such as storage flexibility. However, due to earlier network bandwidth limitations, it was not as widely explored as the alternative solutions. As networks became faster, the network-centric approach began gaining considerable traction, especially as a new breed of intelligent switches emerged with built-in storage management and virtualization.
The dramatically changing landscape across today’s IT organizations is a call to action for companies to factor these new technology advances into their disaster recovery strategy.
Today’s Most Common Disaster Recovery Implementations
While there are many vehicles to implement disaster recovery, including application-based replication and database replication, the three most common implementations are array-based, host-based and network-based. Each has a distinct set of advantages and drawbacks. Array-based replication, the most mature of the three, includes benefits such as the ability to offload overhead from the server, better performance and typically no need for agents on the server. On the flip side, array-based replication can be expensive, may create vendor lock-in and does not always work well with storage arrays from disparate vendors.
Host-based replication for disaster recovery is a notable option. It supports both internal and external storage, does both block- and file-level replication, and can replicate between different OS file systems. Unfortunately, host-based software requires a license at the source and target, must be compatible with the server OS, and most products are Windows-based.
Network-based replication for disaster recovery has many architectural advantages. Network-based replication, delivered through an intelligent switch, can use the current server network infrastructure, and is relatively inexpensive. All three of these major replication architectures have a place in the environment and can address business requirements for disaster recovery replication architectures. The key questions to ask: (i) what are my business requirements? and, (ii) which one is best suited to address those needs?
Integrating Disaster Recovery with Server Virtualization
Server virtualization is the hottest thing since the iPod. More and more businesses are incorporating server virtualization as a key component of their strategic IT direction. The reasons are many and include: the ability to consolidate, great economics and impressive disaster recovery capability. However, on the storage side, the momentum has not been as crisp. That is changing. In a storage virtualized environment, you can now derive the same benefits of consolidation, great economics and impressive disaster recovery capabilities. With a virtual network-based solution for storage virtualization, you can create a framework whereby all storage is visible whether it is fibre channel or iSCSI SAN. By choosing this approach, replication becomes easy to implement regardless of storage type.
In a virtual machine implementation you have the capability to move logical machines dynamically and non-disruptively. In the event of a disaster you have total protection at a logical server level. So the recovery point and recovery time is instantaneous at the server level. With most disaster recovery storage replication implementations, the failover is not instantaneous, and manual intervention is required. Therefore, you lose the benefit of instantaneous recovery.
To address this discrepancy between the server and storage, you can implement instantaneous failover by utilizing a network-based intelligent switch architecture, as depicted in the illustration below.
This implementation demonstrates the complementary capabilities of virtualized storage and servers. In the event of a disaster at one location, the virtual server has the ability to failover to the server at the remote location without any interruption. On the storage side, the storage cluster switch maintains a heartbeat and if one suffers from a disaster, the other dynamically responds to the virtual server. That means zero hours RTO (recovery time objective) and zero RPO (recovery point objective). The new generation of virtual servers and storage virtualization delivered via an intelligent switch offers a way to effectively and efficiently implement disaster recovery without the expense and need for proprietary storage architectures.
Implementing a disaster recovery plan is much less complex and more economical today than it was a decade ago. What has changed the paradigm from one of complexity and high cost to one of simplicity and low cost is network-centric disaster recovery based on intelligent switches. Intelligent network switches are designed to work with both storage and the network. By comparison, storage array-based disaster recovery was retrofitted to provide this capability. Intelligent switches are storage vendor agnostic and support any storage type or brand – two very key factors that diminish complexity and increase flexibility on the network. Intelligent switches are also server OS agnostic. Users can select a storage or server type that best fits their disaster recovery requirements without the dependencies or cost of host or array software.
Kim Tchang is vice president of marketing for SANRAD, a leading supplier of Storage Area Network (SAN) connectivity and management solutions. Pervious to her position with SANRAD she held executive marketing positions at NetApp’s StoreVault Division, Hitachi Data Systems and several new start-up businesses at Hewlett-Packard. Tchang holds a B.A. from Yale University and an M.B.A. from the Wharton School of Business.
"Appeared in DRJ's Summer 2008 Issue"