Q&A: Disaster Recovery Considerations for Large Enterprises
Increasingly complex data center environments put an extra strain on your disaster recovery plans. How should you address performance, capacity, and management requirements and ensure your protection strategies are effective?
As the complexity of your data center grows, so does the complexity of your disaster recovery plan. To learn how IT is addressing key issues of performance, capacity, management, and effectiveness of its data protection plans, we contacted Jay Kramer, vice president of worldwide marketing for SEPATON, a data protections specialist.
Enterprise Strategies: What challenges and requirements do large enterprises face?
Jay Kramer: Like their counterparts at smaller organizations, enterprises need a fast, efficient way to keep critical business operations running in the event of a variety of downtime scenarios, ranging from brief power outages to full site-wide disasters. However, for enterprises, this challenge is made significantly more difficult by the sheer volume of data and complexity of business requirements they must meet. For example, an SMB can simply replicate data from a local data center to a remote facility over a WAN. However, moving petabytes of data over a WAN is too costly and time-consuming for an enterprise to consider.
How are enterprises currently addressing the performance, capacity, and management requirements they need to deliver effective disaster recovery (DR)?
Until recently, the only cost-effective way for large enterprises to protect their backup data from disasters was to copy it onto tape and ship the tapes off site. While, this strategy is cost-effective in the short term, restoring data from physical tape is a slow, linear, highly manual process. It also requires the enterprise to transport data outside the security of its own data center -- an intrinsically risky proposition. In addition, performing disaster recovery testing in a physical tape environment is expensive and time-consuming. As a result, many organizations do not perform testing frequently enough nor in a completely realistic scenario to gauge the effectiveness of their DR strategy.
What should enterprises be doing to address these issues?
Backup systems that were designed for smaller data volumes cannot deliver the performance, scalability, or deduplication capabilities that are needed in an enterprise data center. These systems are typically limited to a single processing node and a fixed capacity. To add either performance or capacity, you need to add an entirely new system. Although these limited systems are adequate for SMB and departmental environments, they require an enterprise with larger data volumes to divide its backup among dozens of individual systems.
For that reason, more enterprises are adopting enterprise-class virtual tape libraries with remote replication and data deduplication. These systems back up petabytes of data to a single system. As data is backed up, systems concurrently deduplicate and transmit the data to the remote site. As a result, they cut the amount of data to be transmitted to a remote site by orders of magnitude while maintaining the fast backup performance needed to meet enterprise backup windows.
Disaster recovery planning includes meeting both recovery point objectives (RPO) and recovery time objectives (RTO), yet RTOs are often overlooked. What are some of the technology issues enterprises need to consider to ensure RTOs are met?
Many technologies, including disk-based solutions, make restoring data slow and cumbersome. There are several impediments to restore performance. For example, some virtual tape library technologies can only restore data through the same port that was used to back up the data. Even if several ports are free to perform a restore, the backup has to wait. Other technologies perform deduplication in a way that requires significant processing time to rebuild the data before it is restored. These solutions use the first backup as a baseline. They break up each subsequent backup into more pieces -- new data and pointers to this baseline data.
Over time, these systems require longer and longer restore times. These technologies were designed for smaller data volumes and work well in SMB environments. However, enterprise deduplication technologies keep the most recent backup ready for instantaneous restore without the need for reassembly. Investigate the restore capabilities of the technology you choose. Ask whether you can restore through any port and whether you can restore deduplicated data without waiting for deduplication reassembly.
How does data deduplication affect DR planning?
Data deduplication enables enterprises to store more data online and retain it there longer. As a result, deduplication enables enterprises to save on power, cooling, and floor space both locally and remotely.
When used with a remote replication capability, deduplication enables enterprises to consider setting more aggressive recovery time and recovery point objectives. In addition, by using a virtual tape library (VTL) with deduplication at local and remote sites, enterprises can test more frequently and more accurately because VTLs eliminate the cumbersome tasks of loading tape cartridges and reading them back linearly.
Efficient data deduplication minimizes the amount of bandwidth required for a replication implementation, which saves on network costs and makes DR solutions more attractive from an ROI perspective.
New technology is available to move large volumes of data off-site for disaster protection, but bandwidth remains an issue. What should enterprises consider before adopting remote replication?
Consider the following factors before choosing a remote replication technology:
- Scalability: Many deduplication technologies do not enable you to scale performance or capacity. A scalable solution enables you to start with a system that meets your current needs and lets you add performance and or capacity in increments as you need them.
- Backp performance: Investigate how much the technology is likely to slow your backup performance. Some "inline" deduplication technologies may not be able to process a large-volume backup within your backup windows.
- Restore performance: As described above, some deduplication technologies can't restore data fast enough to meet enterprise class RTOs.
- "Tunability": Many deduplication technologies require "all-or-nothing" deduplication. Choose a deduplication solution that lets you choose which data you want to deduplicate and the level of deduplication you want to apply (byte-level, object-level, etc.) for maximum efficiency.
- Management console: Choose a technology that provides a clear view of where your data is in the process from backup, deduplication, and remote replication.
Enterprise data is growing almost exponentially and business and regulatory requirements are constantly changing. How does an enterprise ensure that the system they pick today will meet their requirements next year, in two years, or in five years?
To stay competitive in today's business environment, enterprise data centers need to adapt quickly to respond to changing IT needs, including mergers, acquisitions, new regulatory requirements, and changing business continuity requirements. Ensure that you can scale both performance and capacity of your solution modularly. Choose a solution that works with mainframe, open systems, and mixed environments and enables you to add, reduce, or change virtual devices (virtual drives and libraries) as you need to.
Additionally, choose a VTL system that lets IT analyze capacity growth and performance efficiency over time to give an accurate prediction of future requirements. Detailed reporting of backup growth, deduplication efficiency, and capacity requirements are essential to eliminate over buying, while ensuring your system has sufficient resources to meet your needs.
How can companies ensure their disaster protection strategies will be effective?
To help ensure your disaster protection plan is effective, you must include a complete test of systems, procedures, and processes in as realistic a manner as possible. This testing has to be structured to identify weaknesses in data movement, network delays, bandwidth limits, and, potential human errors to ensure that applications and data can be restored within acceptable RTO targets.
To perform DR testing with a tape-based system requires recalling tapes from archive, loading them in a remote-site tape library and restoring data from the tapes over a network. Because this process is time-consuming and costly, it is typically performed only annually or semi-annually to meet minimum regulatory requirements. Companies also perform a variety of steps beforehand, such as preloading tapes, notifying staff, and reducing network traffic to make the testing process faster and more efficient. These steps can make disaster testing less accurate.
Where does SEPATON fit into this market?
SEPATON develops enterprise-class, disk-based data protection solutions to meet the data backup and disaster recovery needs of large, data-intensive organizations. Its SEPATON S2100-ES2 virtual tape library is designed to back up and store petabytes of data in a single system. SEPATON DeltaStor deduplication and DeltaRemote remote replication software add fast, bandwidth-optimized replication of enterprise-class backup volumes. Unlike other technologies that have to locate and reassemble deduplicated data before they can begin to restore it, DeltaStor's DeltaCache Recovery feature keeps the most recently backed up data intact and ready for immediate restores without the need for reassembly, eliminating the data fragmentation that occurs over time with other solutions.
Our DeltaRemote software also makes DR testing of a SEPATON VTL faster and more efficient by eliminating the need for tape handling and pre-test preparation. Data is stored in cartridge format on disk. Restores to the primary site are immediate and the advanced management console enables you to track a wide range of replication metrics to test and tune the efficiency of restore parameters.