Implementing an IP SAN for disaster recovery: using iSCSI as an enabler.

Staying competitive in today's business environment requires constant access to up-to-date information. Highly-available access to customer and competitor data, financial documents, online transaction processing and ecommerce applications are essential to ensuring efficient and timely business decisions. In the event of a disaster, natural or otherwise, loss of access to critical data can be disastrous to a company.

Having a thorough, and tested, disaster recovery plan in place can be an immense help if a disaster were to strike. A good disaster recovery plan covers many more areas than just data access. However, for the purpose of this article, only access to information and business-critical data will be considered. By implementing a Storage Area Network (SAN), an organization is able to ensure data high availability and redundancy. Further, IP SAN architectures based on the iSCSI protocol offer affordable solutions to the small and medium-size business (SMB).

Disaster Recovery Challenges

To any business, the major consideration to recovering from a disaster is one of cost. One major factor determining the amount of cost incurred is the time to recover. The longer the recovery time, the larger the revenue loss and the more difficult to financially overcome.

During the disaster recovery process, how will business be conducted? Are there processes or plans in place to allow you to continue business until your systems can be recovered? Will you have application and data availability during this period? These questions and others should all be considered, answered, and assumptions tested prior to a disaster. The primary recovery strategy for most organizations continues to be tape backup, especially in the SMB market segment. There are several shortcomings to this approach.

Implementing a LAN-based tape backup requires streaming data across the LAN to a backup server. During the backup process the LAN experiences significant performance degradation. To circumvent this problem, many companies perform their data backups overnight. The burgeoning amount of data needing to be stored is in direct conflict with the need for decreasing backup windows created by 24X7 business transactions.

Another shortcoming of tape is the media itself. Using tape repeatedly will cause the media to stretch. There is also direct tape contact with the head mechanism, increasing the chance of media contamination.

Finally, the largest downfall of tape is not the time it takes to back up, it's the time it takes to recover from a disaster or an equipment failure. Restoring from tape can take several hours to several days depending on the amount of data to be restored and whether full or incremental backups have been performed. This downtime is simply not acceptable to businesses wanting to stay competitive in today's data-driven business environment.

Despite the drawbacks stated above, tape is still a widely accepted storage medium. Tape archival can be an effective method of long-term data storage. However, if you are using tape as your primary disaster recovery method and you cannot tolerate a large restore window, it would behoove you to rethink your strategy. Current SAN technologies could improve your recovery scenario and help consolidate your storage infrastructure simultaneously.

Introducing iSCSI Storage Networking

Since the approval of the protocol, iSCSI has been steadily gaining acceptance among end users and storage vendors and is proving to be a disruptive technology. More vendors are now offering iSCSI products ranging from interface cards to routers, gateways, and storage devices. Initially there was much debate and skepticism about the speed of block data access over a network. The ability of host systems to process incoming and outgoing IP packets was a subject of major concern and endless debate. With the introduction of TCP Offload Engine (TOE) bus adapters, this concern has been lessened. The removal of the transmission processing from the host operating system and CPU to the HBA is paramount to providing improved performance on the host system and a big step toward iSCSI acceptance.

The Promontory Project has proven that, indeed, native wire-speed IP storage is possible. What this means is that IP storage networking is now viable for the SMB market segment. Fibre channel SAN infrastructures will not become obsolete any time soon--large organizations are very heavily invested in the technology. However, iSCSI offers the SMB segment the benefits of networked data storage without breaking the bank on more costly Fibre Channel solutions. Today, 100MB/s transfer rates are achievable with gigabit Ethernet. This aggregate throughput will increase with the introduction of 10, 40 and even 100Gbps Ethernet technologies.

There are other benefits of iSCSI storage to be realized by end users. First, Ethernet has a very large advantage by being a thoroughly entrenched networking technology. As such, most technical personnel understand the technology, which eliminates the expensive training needed to implement and maintain a large Fibre Channel solution. Another advantage of personnel being familiar with Ethernet is the ability to get their arms around an Ethernet-based SAN quickly. The learning curve is much shorter.

The iSCSI SAN (From the Inside Out)

The first consideration in the implementation of an iSCSI SAN is the storage array to be used. Vendors are offering iSCSI storage enclosures in increasing numbers. Having an enclosure that speaks the iSCSI protocol is advantageous since you'll remove the necessity to purchase an additional translation device. However, it is fine to still use your existing arrays with a device such as an iSCSI bridge to handle the translation.

The fabric level exists above the data storage and consists of Ethernet switches, hubs and gateways. The benefit of the iSCSI fabric over the Fibre Channel fabric is cost and interoperability. In addition, there are still compatibility issues among Fibre Channel vendors requiring users to separate incompatible hardware into separate switch zones.

Speed and network latency are issues that must be addressed in the iSCSI fabric. Since iSCSI provides block-level access to storage over an IP network, 10/100Mbps speed is unacceptable. To maximize efficiency, storage traffic should be separated from LAN traffic and implemented with gigabit Ethernet. Ideally, this would consist of a dedicated LAN strictly for storage traffic, not just a different subnet on the same network. An exception to this would be a small organization using the SAN only as a centralized storage pool.

Outside the fabric resides the host level. The host level is made up of the servers and systems attached to the fabric and initiating the storage transactions. When considering the host (or in iSCSI lingo, initiator) hardware it is important to consider the HBA that will be used. As discussed previously, one of the most advantageous HBAs to use (to reduce host CPU load) is TOE-enabled. Ordinary Ethernet cards will require the processing be done by the host CPU (as in ordinary IP transactions), thereby inhibiting host performance.

Availability, Redundancy and Rapid Recovery

In addition to benefits such as increased storage utilization and data consolidation, storage networks provide added benefits such as data availability, redundancy and the ability to recover data quickly. By providing such benefits, storage networks are the ultimate solution currently available for business processes, mission-critical applications and data disaster recovery.

By utilizing multiple-path-capable software, IT departments are able to meet demanding service level agreements with minimal administrator intervention. By allowing initiators access to data stored on a SAN via multiple paths, you can achieve high data availability and redundancy at the host level. This is done by providing multiple HBAs in each host and multiple Ethernet switches and interconnections within the fabric.

Point-in-time copies, snapshots, remote mirroring and disk-based backup are features all targeted for rapid recovery. In the event of a disaster, the ideal situation is to provide access to a mirrored copy of data stored at an offsite location--either through a second data center or a remote hot-site (storage service provider).

For the business requiring 24X7 data availability, the most attractive option is to provide a remote data center, preferably residing on another power grid. If this is too cost-prohibitive, the next best thing would be to mirror to a hot-site. If this is still prohibitive, local point-in-time copies and/or snapshots coupled with storage device RAID configurations can be implemented within the SAN to affordably provide a rapid recovery solution.

Putting it All Together

When considering today's burgeoning data growth rates created by the need to use information for competitive strategy and business processes, compounded by government regulations such as the Health Insurance Portability Assurance Act (HIPAA) and Sarbanes-Oxley, an IP-based SAN is an effective way of mitigating the risk associated with disaster and recovery. By providing a remote site accessible over hundreds or even thousands of miles at wire speed utilizing iSCSI technology, the SMB can now more affordably protect the investment they've made in their organization.

www.peaktechnologyconsulting.com

Brian Irwin is president of Peak Technology Consulting (Fort Collins, CO)