Networked storage–NAS or SAN–can facilitate resource consolidation and centralized management.
By Tim Rasmussen
Data warehousing, data mining, CRM, ERP, e-business, databases, e-mail. All of these applications contribute to the massive amounts of information that businesses need to store. Rather than maintaining a direct-attached storage environment where there is a one-to-one relationship between each server and its disk subsystem—and where there is a continuous need to add storage capacity to handle the increasing amounts of data—companies continue to turn to storage consolidation.
Storage consolidation centralizes and shares storage resources among numerous application servers. It addresses the limitations associated with direct-attached storage and solves the growing data demands generated by information-rich business environments.
The benefits of storage consolidation include
- Centralized management of storage resources;
- Ability to manage higher capacity per storage administrator (10 to 20 times as high as direct-attached in some cases);
- Storage investment spread across many servers;
- High-capacity utilization;
- Reduced administration costs; and
- Ability to meet quality of service (QoS) requirements, including capacity, performance, and availability.
Three methods of consolidation
Currently, there are three dominant architectures through which a company can achieve the benefits of storage consolidation: enterprise RAID, storage area networks (SANs), and network-attached storage (NAS). Although these architectures provide common benefits, each one has unique features and different environments for which it is most ideally suited.
Figure 1: An enterprise RAID system supports heterogeneous servers with multiple direct-attached host connections.
Enterprise RAID and SAN architectures both rely on block-level transfers between the operating system of the application server and the storage subsystem. An enterprise RAID system (see Figure 1) supports heterogeneous server environments with numerous—64 or more in some cases—direct-attached host connections. These systems are often implemented in data-center environments where there are a fixed number of mission-critical servers.
SAN storage consolidation architectures (see Figure 2) leverage enterprise RAID subsystems by connecting them to a dedicated storage network, which is typically based on Fibre Channel. With a RAID storage subsystem connected to a SAN, many more servers can use its capacity than the number of physical connections would normally allow.
A SAN provides benefits to companies in virtually any industry. Typically, SANs are best-suited for companies that have large capacity requirements spread across many servers. The larger the data storage requirements are, the more dramatic the ROI on a SAN will be.
The primary difference between using an enterprise RAID array in a SAN environment versus having it directly attached to servers is that the physical connections to the RAID system can be shared by multiple servers. This capability not only allows more servers to participate in the storage consolidation solution, but also may reduce the number of physical Fibre Channel interfaces required on the RAID system. Different classes of RAID systems can be connected to a SAN, allowing capacity to be provisioned based on QoS requirements of individual applications.
NAS consolidation architectures (see Figure 3) are different in that they use standard IP-based networks to provide file-level access to stored information. NAS is a good fit for applications that can leverage file-based access to data and where management complexity needs to be minimized.
One concern that businesses may have in implementing storage consolidation architectures is that they're putting "all their eggs in one basket." That is, with all the mission-critical information kept in one storage subsystem, access to all that information would be lost if the subsystem failed.
Figure 2: SANs enable storage consolidation by connecting disk arrays in dedicated storage networks, or fabrics.
This is addressed by using enterprise RAID arrays that have redundancy built in to handle any type of physical failure. Redundant paths to each host ensure access to information is not compromised if a cable, switch, or host bus adapter (HBA) fails. In addition, maintenance such as upgrades in capacity, cache, or firmware can be done without system downtime. Clustering technology provides a similar level of high availability for NAS solutions.
Recently, new technologies have begun to emerge that provide additional functionality and cost justification for storage consolidation infrastructures. Among the more noteworthy are the iSCSI protocol, virtualization, storage resource management (SRM) software, and NAS-SAN convergence.
iSCSI uses existing Ethernet
iSCSI works by loading a SCSI intercept driver on the host, which captures a SCSI command and encapsulates it into IP packets. The SCSI command is then routed through the IP communications network rather than a dedicated physical cable such as Fibre Channel. The main benefit of iSCSI is that servers that could not previously be connected to the storage subsystem because of distance limitations or prohibitive cost can now be connected.
Drawbacks to iSCSI include reduced performance (compared to Fibre Channel) and, at least for now, a limited number of vendors supporting the protocol. However, performance penalties are being overcome by faster CPU processors and dedicated Ethernet NIC cards. Once the protocol becomes standardized, large storage vendors will most likely begin offering iSCSI interfaces directly on their storage subsystems, eliminating the need for routers to convert from Fibre Channel to IP.
Virtualization provides benefits
Different vendors have varying definitions of virtualization. Fundamentally, it can be defined as creating a logical representation of physical disk drives, a so-called abstraction layer. As a result, physical disk drives or RAID subsystems can be logically combined to create one large pool of storage.
The benefits of virtualization include the ability to mirror across heterogeneous storage; provide remote replication to heterogeneous storage; use legacy storage subsystems in SANs; leverage less expensive storage subsystems; increase management efficiency; and provide dynamic LUN expansion and assignment.storage resource management
Storage resource management (SRM) software can be categorized into two types. With the first type, a central management server gathers information from each client and performs monitoring functions. This allows system administrators to ensure only appropriate information resides on the storage consolidation infrastructure, thus reducing the amount of capacity required. The second category focuses on the storage infrastructure, including hardware components such as switches, HBAs, and storage subsystems. This type of software monitors the infrastructure, but does not necessarily look at the type of information being stored.
As SRM software evolves, solutions that incorporate both types of functionality will emerge, allowing storage administrators to not only monitor their hardware components, but also to ensure the information stored meets the company's standards for storing data.
Until recently, NAS and SAN storage architectures were two completely different types of storage consolidation architectures. But emerging technologies now incorporate the functionality of both architectures. (For more information, see "Who needs NAS-SAN convergence?", InfoStor, October 2002, p. 24.)
This gives storage administrators flexibility to provide capacity to application servers and clients based on the most beneficial file access method (block-level or file-level). Either choice will leverage the single storage consolidation architecture, contributing to efficient management.
While storage consolidation is not a new concept, the need for consolidation has never been more apparent than now. For more information about storage consolidation, visit consolidator.datalink.com and download the white paper, "Storage Consolidation: Simplifying Storage Infrastructures."
Tim Rasmussen is a senior product manager at Datalink (www.datalink.com) in Chanhassen, MN.