Emerging SAN management models

Management is a major concern among SAN adopters. Domain management provides the necessary control for optimum performance, security, and data availability.

By Bill Prescott

Storage area networks enjoyed a yearlong coming of age party in 1999. Despite rapid growth, a surge of market awareness, and several spectacular IPOs from pure-play SAN companies, a large portion of SAN activity has been confined to pilots, leaving the proving grounds of live production data centers yet to be covered. Now, as the Internet drives radical changes in the demand for data delivery, SAN technology is expected to advance into the line of fire. The question is how will it stand up in prime time.

While interoperability, or basic connectivity, was the first great challenge facing SAN adopters, the top concern now is management. Few doubt that management capabilities will hold the key to success as SANs go live. Of the multiple approaches to SAN management, neither traditional server-centric software suites nor feature-enriched infrastructure hardware offer ideal solutions. A hybrid "domain management" approach uses a combination of optimized hardware and software to create and manage domains, i.e., physical subsets of the larger SAN with specific control points for management. This hybrid approach complements and enhances the enterprise management suites and infrastructure equipment, enabling SANs to truly meet corporate needs.

At the most basic level, companies simply want assurance that:

  • Their data is forever safe and secure.
  • They can read and write data fast, wherever they need it, and without interruption.
  • They will always have enough capacity.
  • They will expend a minimum amount of time, effort, and money accomplishing the above.

SANs offer scalability, performance, storage consolidation, 24x7 availability, and disaster protection-business needs not addressed adequately by traditional server-attached storage. However, by itself a SAN topology only provides a high-performance set of connections between servers and data. Without additional intelligence and management layered within and on top of the SAN, IT managers are left with many of the same issues. Domain management capabilities address reliability, performance, disaster protection, security, and data-sharing needs in SANs. The table on p. 26 lists a sample set of enabling technologies, features, and mechanisms for domain management.

Without these management features and controls, the full benefit of SAN technology cannot be realized. One advantage of this model is that it solves the problems and provides the benefits as described above with remarkable configuration flexibility. The model enables IT administrators to deploy storage resources in a variety of domains-from small subnets to the entire SAN-and to organize these domains into a cohesive hierarchy.

Furthermore, in addition to the underlying controls, the model presents a unified and consistent (read open) management interface so users do not need a different interface for each vendor's devices and management features.

Applying domain management

SANs consolidate storage resources into a common pool. Data pathways connect any two nodes, so in principle each server can see every storage device in the SAN. This capability promises benefits in terms of data sharing, clustering, hot-plugged capacity additions, heterogeneous support, and centralized management. Yet, it also poses security and organizational problems as well as the potential for bottlenecks and single points of failure-all of which could affect a large number of servers and, hence, users.

Click here to enlarge image

In practice, it is usually necessary to add structure to SANs, that is, to subdivide them into logical subnets. Each subnet has control points or gateways that offer a logical place to apply domain management. Generally, there are five types of domains: the entire SAN or subnets connected by SCSI, Fibre Channel, LAN, or WAN interfaces (see figure).

The whole enchilada

One obvious place to apply storage management is between all of the computers (i.e., the "consumers" of storage, the application servers, database servers, desktop clients, etc.) and the rest of the SAN. In this case, the entire SAN, not just a SAN subnet, is managed at a single control point, called either a central domain manager or a storage domain manager.

Click here to enlarge image

A domain manager that serves as the sole entryway to the SAN must support all SAN traffic, so it has to have extremely high-bandwidth and it must be an intelligent manager with switching capabilities and high-availability design. No matter how fast, such a device places a practical limit on the maximum size of the SAN below it. Other designs for central domain management avoid this limitation by weaving central domain managers into a hierarchy that shares information and supports cascading, fail-over, and enhanced scalability in a cooperative network.

SAN subnets

Often an enterprise SAN is made up of departmental SANs, or smaller SANs, that are stitched together, usually via a switch that serves as a gateway. In some cases, the goal is to create a single, centralized SAN; in others, it is to keep distinct departmental SANs or subnets, while allowing limited, managed cross-connectivity among them. In either case, the departmental SANs can be handled as distinct subnet domains, with the gateway device acting as the control point for storage management.

SCSI subnets

Legacy disk and tape devices must be able to participate in a SAN or companies will find the migration from direct attached to network topologies too large a step to take. Devices that connect SCSI devices to Fibre Channel SANs-such as bridges, routers, gateways, and domain managers-allow up to 60 SCSI devices to be attached to a Fibre Channel SAN. Multiple devices and hence subnets can coexist in the same SAN. Each is a potential control point for management of the storage beneath it, as long as the device has the intelligence and software to provide or support storage management for the SCSI subnet. Products such as domain managers and gateways offer a degree of storage management capability.

Cross-LAN subnets

Virtually every organization capable of deploying SANs has already invested significantly in IP LAN backbone infrastructures. Therefore, in some cases, it may be more practical to extend a SAN or to join two distinct SANs across the IP backbone than it is to build a Fibre Channel link between them.

In such cases, Gigabit Ethernet connectivity in a switch or router presents a gateway control point for management of the remote domain. Such management needs to be bi-directional, for each side can view the other as the "remote" SAN.

Cross-WAN subnets

Similar to the logical extension of remote subnets across a LAN, geographically remote SANs can be joined across a WAN link, such as ATM. Likewise the WAN connection points create the control gateway for storage management of the connected resources.

Hierarchical domain management

Ultimately, through a combination of central and subnet domain managers, an infrastructure for the SAN takes shape. For the sake of organization, it is helpful to organize these domain managers into a hierarchy. This type of hierarchical domain management architecture enables central domain managers to take advantage of the management features in the subnets below it. By delegating some of the storage management tasks to lower levels of the SAN, performance and the granularity of management can be improved.

Without storage management of domains, SANs cannot adequately support the business needs for data protection, security, integrity, reliability, and availability. As the number of SAN deployments in live production environments increases, management will be the key to success. To that end, emerging architectures for SAN domain management and hierarchical domain management will be crucial.

Bill Prescott is vice president of sales at SmartSAN Systems (www.smartsan.com), in Santa Clara, CA.

This article was originally published on February 02, 2000