Scaling storage networks demands careful consideration

In the Fibre Channel world, scaling often means adding more and faster switch ports to extend the fabric's bandwidth and connectivity. But IP storage (mainly iSCSI) is a growing area of SAN expansion, using ubiquitous Ethernet network technology to pass storage data between storage devices. This requires the use of IP switches and routers, and can involve the deployment of iSCSI TOE host adapters to offload iSCSI traffic from the local server's CPU. This Buying Guide is intended to help clarify the principle considerations involved scaling storage networks and each chapter offers a set of buying points and product specifications that can help readers identify prospective new scaling products.

Storage area networks (SANs) must scale to accommodate the added demands of more users along with a proliferation of more sophisticated enterprise applications. In the Fibre Channel world, scaling often means adding more and faster switch ports to extend the fabric's bandwidth and connectivity. For example, adding 4 Gb FC ports supports more interconnected storage devices and servers, while enhancing their performance and availability. But IP storage (mainly iSCSI) is a growing area of SAN expansion, using ubiquitous Ethernet network technology to pass storage data between storage devices. This requires the use of IP switches and routers, and can involve the deployment of iSCSI TOE host adapters to offload iSCSI traffic from the local server's CPU. This Buying Guide is intended to help clarify the principle considerations involved scaling storage networks and each chapter offers a set of buying points and product specifications that can help readers identify prospective new scaling products. Before we look at specific product categories, let's start by identifying the core concerns when scaling any storage network.

@35604 Consider the rate of storage growth. Storage growth rates will have a profound impact on the technology that is selected. Slow growth often favors traditional monolithic storage systems and fixed port switches where there is a single capital expense and only infrequent upgrades. Rapid growth needs typically favor modular storage systems or blade-based servers and switches which can be expanded quickly. Growth also influences bandwidth, and storage network growth should eventually include upgrades to switch port speeds. For example, it may be prudent to implement 4 Gb FC switches to scale the storage fabric's bandwidth and avoid bottlenecks -- even if some servers and storage devices are still operating at 1 Gb or 2 Gb speeds. Similarly, deploying TCP/IP Offload Engine (TOE) cards can enhance the performance of an iSCSI SAN.

Consider the time table of any expansion. Analysts note that overall storage costs (e.g., disk, port and HBA costs) are declining over time, and more features are always being added. This means a massive scaling project implemented in the near term may actually wind up costing more and offering fewer features than a systematic upgrade effort that takes place periodically over time. This can be true even when factoring in potential savings of upfront bulk purchases. Once you have an accurate knowledge of growth rates, compare product costs and see what makes more financial sense for your organization.

Migration should be included in the scaling process. When storage capacity is scaled with more disks or additional storage systems, some data may need to be moved in order to take advantage of faster disks, more capacity, or faster connectivity. Time is also essential when an older device is being decommissioned because the longer that migration is delayed, the more your company must pay to operate and support both systems. Migration is a mission-critical task, but it's typically overlooked until new hardware is sitting on the data center floor. Data classification can help ease the migration process, especially when using data classification and migration automation tools to classify and move data according to your existing policies.

Remember to also scale backup and disaster recovery. Any time that storage and storage networks are added or expanded, it's important to consider any adverse side-effects on data protection policies and practice. Increased storage will demand larger backup windows. Changes in tiered storage may alter backup or snapshot policies -- not all data may need to be protected in the same way. In extreme cases, improved backup technologies may be required. For example, tape may give way to disk (e.g., virtual tape library or disk-to-disk storage) and space-saving tactics like data deduplication. Similarly, backup changes will impact disaster recovery tasks such as remote replication. If backup software licensing is priced "per terabyte", increasing storage can dramatically increase backup costs.

Evaluate changes to management requirements and processes. Scaling a storage network can also influence established and documented procedures such as backups or provisioning. For example, implementing iSCSI in a workgroup or integrating an iSCSI SAN with an existing FC SAN can cause substantial procedural changes that should be considered during any scaling project. Also weigh the management overhead that may accompany an upgrade. Experts note that additional storage capacity rarely adds significant management overhead, but adding new technologies or processes to the storage environment may require more labor. For example, another 10 terabytes of disk space may be readily manageable with existing staff, but implementing a new remote replication scheme for data protection might require additional attention.

Track performance before and after scaling. It's important for an enterprise to monitor network, server, and storage fabric performance before and after any scaling project. For example, industry experts suggest watching inter-switch local and long distance links to identify points of congestion and ensure required service levels. Testing allows IT staff to draw a performance benchmark prior to an expansion, then measure the benefits, optimize any new hardware, and spot any unexpected bottlenecks which can be addressed or tabled for future upgrades.

Return to the beginning

Read more on Networking hardware