FTP Online
 
 

Storage Methods That Prepare You for the Worst
Use storage virtualization, mirroring, and replication techniques to develop an effective disaster recovery plan.
by Nils Odhner

Your company needs a safe remote backup or disaster recovery (DR) solution. In our post-9/11 world, and with recent events such as the power failure that caused a major blackout in the northeastern U.S., companies of all sizes—particularly utilities and other critical service providers such as telecom carriers—must make sure they are prepared for the worst.

According to the META Group, more than 30 percent of businesses that suffer a catastrophic disaster such as such as fire, flood, or earthquake will never reopen their doors for business. Faced with such a grim reality, the data storage industry has responded by developing techniques and technologies that can help you create a sound and effective disaster recovery plan or strategy for your company, using such common techniques as storage virtualization, mirroring, and replication as a foundation.

Virtualization, Mirroring, and Replication
Three of the more common backup methods used in today's IT world include storage virtualization, mirroring, and replication.

Storage Virtualization refers to the process of dividing or aggregating available storage devices and their capacities into "virtual volumes" without consideration of the actual physical storage-volume layout or topology, which can include various storage elements such as disk drives and redundant array of independent disks (RAID) subsystems. Storage virtualization typically requires storage area network (SAN) management software for centralization, which can assist you in allocating volumes and reassigning storage resources. There are several techniques for carrying out storage virtualization; two of the more common techniques are data mirroring and replication.

Mirroring (or Data Mirroring) is the process of copying production or other business-critical data on disk storage devices from one location to another as a "mirror image" of the information at the production site. Essentially, data mirroring creates a secondary or duplicate copy online, replacing the primary data in the event of an emergency. A mirrored copy eliminates the need for a daily, manual backup.

Replication (or Remote Replication), on the other hand, simultaneously duplicates data to a secondary system to ensure continuous access in the event that the primary system fails. The differentiating factor in remote replication is the use of a secondary system, rather than a virtual online "mirrored" copy.

Both data mirroring and remote replication eliminate the long wait time associated with loading and restoring data on backup tapes following a disaster. Both techniques also can replicate damaged or corrupted data as it enters the system, copying damaged databases in the process.

Storage virtualization, or the "blueprint" used to carry out data mirroring and replication, is essential in assessing and determining the parameters of your disaster recovery or business continuity plan. But how do these three processes figure into the implementation of an effective disaster recovery strategy?

Perhaps you should look at it more simply: storage virtualization is similar to the process that a general uses in mapping out his platoons before battle into their various groupings or "clusters," and when he gives the command to go into battle, his plan is "virtualized" into action. Though more crude than SAN management software, the general serves as the centralized control for his troops. Unfortunately, you cannot actually replicate the soldiers once they're gone, but data can be replaced and restored to the appropriate systems based on the storage virtualization process.

Both mirroring and remote replication simply copy data; key processes used the actual backup. But without the initial virtualization or "mapping" of the physical storage terrain, however, a successful mirroring and/or replication cannot take place.

Tactics for DR Plan Implementation
Now that you've considered the importance of storage virtualization, mirroring, and replication, or the "nuts and bolts" of disaster recovery, let's outline a sound strategy that you can execute efficiently and effectively when an emergency occurs. Using these steps, you'll see how both methods and their variations can help you achieve sound business continuity practices.

Step 1: Addressing interoperability issues

When working with a SAN, which is what you'll typically be using in architecting a disaster recovery plan, ensuring interoperability between each component's software, firmware, and hardware is crucial. You should conduct exhaustive and detailed testing on each of these components, and their revision levels in particular.

You'll also need to test the overall DR framework to make sure it's compatible with intended applications. The bottom line is that you have to develop and document detailed processes and methodologies to achieve sound results: be scientific rather than just throwing together a plan at the last minute, without seeing proof that it works.

Step 2: Prioritize data and match data types with recovery level

If your business involves a significant amount of online transaction processing, as many traditionally brick-and-mortar companies do, achieving a full backup through synchronous disk-to-disk data replication over a safe distance is one possible option to consider. This means that you need to implement server clustering, remote data replication, and mirroring solutions. This combination will create a transparent failover from the production site to the backup site, regardless of whether the failure occurred on the server or storage device level.

One issue to consider is latency, which can occur in synchronous data replication if the backup site is more than 100 miles away from the primary production site. In the case of the recent blackout, wireless service, for example, could not be restored immediately because the disaster recovery sites were located within the blackout region.

That is where asynchronous data replication comes in. Because this data copying method is less susceptible to latency, it can enable successful backup between primary and DR sites for distances over hundreds or thousands of miles, well past the center point of the potential or actual disruption.

Step 3: Accommodating data replication in heterogeneous storage environments

Heterogeneous operating environments have historically caused problems in effective data backup processes. In such environments, host- or appliance-based data replication involves movement of data between one vendor's primary storage and another vendor's secondary storage (think: the accounts receivable of one business and the invoices received of another). Solutions used for heterogeneous data replication typically involve multiple data copying processes, in which the software converts block data to files for transport over IP networks.

Using new storage over IP technologies, you can combine disk-to-disk and disk-to-tape copying to achieve a successful backup. Key protocols used in this process include Fibre Channel over IP (FCIP), which tunnels data replication traffic within a WAN link, and Internet Fibre Channel Protocol (iFCP), which routes storage in a native IP format. This latter protocol also enables fault isolation between connected SANs to ensure that no data is lost in the backup over longer distances.

There are features you will need to ensure the full optimization of available WAN bandwidth to perform data replication via a Storage over IP setup. These include optimized buffering, data compression, support for jumbo frames, and the use of technologies such as Nishan Systems' Fast Write Algorithm. These will allow for disaster recovery to be extended up to hundreds or thousands of miles, making null and void the 10-kilometer limit that used to be imposed by Fibre Channel extensions.

And if your budget is limited because of your status as a small- or medium-sized business, the use of affordable and available IP network services will allow you to take advantage of a disaster recovery strategy using Storage over IP.

Plan to Play It Safe
According to Contingency Planning Research, retail brokerage and credit card firms will lose an estimated $6.45 million and $2.6 million per hour, respectively, in the event of a disaster. Even if you operate a small business, the numbers can be just as astounding and damaging, and you could be put out of business if you don't develop a sound disaster recovery strategy, and follow through with the right architecture and solutions that will best serve your business.

Using storage virtualization, mirroring, and replication techniques should be the nucleus of your plan, but you should also follow the other suggestions outlined here to ensure that business downtime never occurs, even in the event of a catastrophic disaster.

About the Author
Nils Odhner is a Senior Editor and Analyst at Faulkner Information Services, an IT analysis and advisory firm, and a freelance technical writer, analyst, and contributing author to .NET Magazine's Security Strategies Column. Please e-mail him at nils_odhner@yahoo.com for feedback to this article or related topics.