Fibre Channel Meets IT Challenges
The emerging network/storage interface may be the only hope for IT managers faced with booming storage capacities and burgeoning management costs.
By Jeff Vogel and Mike Befeler
Joe, you`ve got to take the database server down to add a new RAID subsystem tonight. By the way, we can`t get to the tape library to back up the financial server; it may be that recurring SCSI connector problem. And if you don`t get the CAD server back up, all the engineers will go home, and they`ll fall another day behind schedule. You know who will get the blame for that."
Sound familiar? Systems and network administrators live in a world of constant change, expected and unexpected problems, and the need to provide users with access to valuable information resources by ensuring that they have access to the data.
Data has become the life blood and currency of the corporation. Information technology (IT) organizations are the "physicians" making sure the data circulates to their users. The users, in turn, rely on the availability of this information to get their jobs done and stay competitive in a fast-paced marketplace. Consequently, more organizations are going to 7x24 operations.
In a recent study conducted by Storage Systems Marketing Inc. (SSMI), a market research firm in Boulder, CO, of 270 IT organizations using Unix and Windows NT, two-thirds indicated they either have in place, or are considering, around-the-clock operations. This trend is being driven by more corporations working longer hours, business travelers who require access to their data from remote locations, and by international organizations that span many time zones. This means that the "window of opportunity" to service or reconfigure the enterprise`s information infrastructure is becoming nonexistent. In some shops, customers have set a policy of no downtime.
In addition, storage capacity is growing significantly. The SSMI study found that IT organizations on average expect a 40% growth in disk capacity at their sites over the next year. Capacity will grow in two ways: IT managers will increase disk storage on existing servers as well as add additional servers to satiate the incredible demand for information.
All these factors keep IT managers under constant pressure as they look for new ways to effectively use their current resources while maintaining the levels of service that their companies expect and demand. Without a new approach for managing the growth of data, enterprises will be facing a data glut that, at best, will slow the network to a crawl and, at worst, could slow the adoption and implementation of new applications.
Storage Area Networks
Another major trend is under way in the enterprise: A new paradigm, often referred to as a storage area network (SAN), is emerging in corporate computing.
This next wave after decentralized client/server systems--recentralization--is necessary to more effectively manage the dispersed resources of the enterprise. In the SSMI study, 62% of IT organizations saw a movement toward recentralizing storage management functions; 61% of the organizations were recentralizing servers; and 61% were recentralizing storage subsystems.
Three network events are driving this trend: the Internet has inspired businesses to connect their internal and external networks to speed up overall communications; companies are placing mission-critical applications on their intranets; and companies have learned that the costs of maintaining PCs on networks is expensive. This has resulted in greatly increasing data storage costs as the amount of data moving across networks increases and networks become so important that any downtime directly affects productivity. This compromises the competitive position of the enterprise.
Technology-based businesses have become so focused on their IT assets that many actually calculate their downtime to the penny. A study by NationsBanc Montgomery Securities showed that total Internet traffic will increase from 2,500 terabytes in 1998 to nearly 25,000 terabytes in 2002--a 10x increase. This is driven by the need to "externalize" the enterprise to take advantage of the Internet as an effective means of timely communications.
Fibre Channel Solutions
IT organizations are required to keep spending to a minimum even as increasing network data levels drive the total amount of downtime higher. Many IT professionals believe that planning for system downtime, service, or reconfiguration will become their number one issue by the year 2000.
In the SSMI study, 81% of IT organizations indicated they had implemented or were implementing RAID; 57% were committed to redundant network components; 55% to redundant servers; and 37% to high-availability software. This demonstrates that data center management departments are shelling out large percentages of their IT budgets for reliable, highly available solutions.
In the area of data storage, the technology that addresses these needs is Fibre Channel. The paradigm shift occurring in the market using this emerging technology as the storage network is often described using an alphabet soup of acronyms: SAN, SAS (SAN-Attached Storage), NAS (Network-Attached Storage). Regardless of the terminology, the architecture provides a blueprint for the way servers and storage subsystems can be interconnected and centrally managed to provide high availability, flexibility, and effective resource usage.
Fibre Channel is a shared gigabit storage solution that uses the popular SCSI protocol, but with an improved "network-like" interconnect infrastructure. Fibre Channel-Arbitrated Loop (FC-AL) is analogous to Token Ring or FDDI, where two communicating nodes possess the shared media only for the duration of the transaction and then yield control to other nodes that are connected to the gigabit loop.
FC-AL uses Fibre Channel commands to negotiate loop access and to assign loop addresses. A loop aggregates loop ports in a physical star configuration with each hub typically providing 7 to 12 ports. One FC-AL node is reserved for attachment to a Fibre Channel switch, allowing multiple loops to join into a broader network or switched fabric. For higher performance, Fibre Channel switches typically offer eight to 16 ports, with full gigabit speeds available on each port.
Fibre Channel provides end users with a more reliable solution since storage interconnect systems reuse over 80% of existing SCSI technology, thus mitigating the "new technology" element. In fact, IT can also lower the cost of adoption by migrating from SCSI to Fibre Channel by using Fibre Channel-to-SCSI bridges to transparently access storage regardless of the interface.
Fibre Channel hubs and switches can support up to a 100MBps transfer rate, with a large number of devices attached, over distances of up to 10 kilometers. In practice, a 10km link might be extended on a few ports for disaster recovery, with the majority of ports well under 500 meters. In fact, typical link distance configurations will probably be less than 250 meters (similar to client networking).
Fibre Channel can run on either fiber-optic or copper cabling. Copper cabling is less expensive but more susceptible to electromagnetic interference (EMI), with passive copper providing no signal balancing and a maximum length of 15m. Active copper provides signal balancing and can run up to 30m. Some Fibre Channel network manufacturers support either fiber or copper on interconnect products using a hot-pluggable gigabit interface converter (GBIC).
Fiber-optic cabling is immune to EMI, so it can be deployed within or between buildings and extend storage networking across an entire campus. Multi-mode fiber can run to 500m, and single-mode fiber to distances up to 10km, according to the ANSI standard specification.
The Fibre Channel storage interconnect model addresses IT`s need to have systems that are highly available, that are adaptable to changing environments without incurring downtime--planned or unplanned--and that deliver performance where needed. The storage interconnect effectively addresses both total cost of ownership while providing cost-effective solutions across all resources.
Fibre Channel Meets IT Needs
What are IT managers looking for in a storage network? The storage interconnect infrastructure must conform to high-availability requirements. This includes testing and certifying complete storage systems from servers to hubs to arrays.
High availability is what IT professionals expect as the prerequisite for enterprise-level solutions. This implies that some form of redundancy is required, and that no single point of failure is possible. One way to accomplish this is by creating fully redundant networks by configuring multiple hubs or switches. Redundant power supplies or fans within a single hub or switch do not provide true high availability because this does not remove the single point of failure.
IT organizations are looking for flexible solutions that allow for dynamic, real-time growth. Fibre Channel offers a "hot plug" methodology and "hot swap" capability that allow a system or peripheral device to be taken offline for maintenance or redirecting without system quiescence or taking the system offline.
IT managers expect a growth path that protects their investment as their needs change or expand over time. IT executives don`t want to over invest when they first buy a product by purchasing more hardware than is required. They are looking for a "pay as you grow" solution that matches their investment to the application while maintaining investment protection.
An interconnect product line of storage network solutions must offer a complete device upgrade path, including user-configurable ports; low-cost-per-port unmanaged hubs on the low-end; single-point-management hubs for more complex loops; and switching fabrics (switches) for comprehensive loop/fabric connectivity. A single FC-AL hub is an inexpensive interconnect solution for two- to 12-node configurations that attach multiple servers and storage arrays. IT managers can cascade multiple hubs to build 60+ node networks with a central point of management. At the other end of the price-performance spectrum is a fabric switch that allows for a combination of dedicated gigabit access with shared gigabit access via hubs.
This kind of flexibility can equal big savings. Because IT organizations can now control whether a transceiver module will be fiber optic (longwave or shortwave) or copper, they no longer have to predict the future by pre-configuring their needs. They can respond in a very short planning window, and with no interconnect downtime to any enterprise data location or bandwidth need. This can result in a variety of savings for the organization, including:
- Not having to buy new equipment before it is needed.
- Significantly reducing or eliminating the necessity for system downtime.
- Saving time associated with scheduling, planning, and coordinating system downtime.
- Saving troubleshooting and interconnect downtime by enhancing existing systems rather than replacing them.
While some vendors are pushing either hubs or switches, most IT sites will require both solutions with multiple offerings in each category. Of course, prices escalate along with performance. Because a switch requires more processing power, memory, and microcode at each port to timely route frames, switch-per-port costs are often more than six times the per-port costs of basic FC-AL hubs.
IT organizations want solutions designed to allow for the sharing of system assets. With a demand for increasing storage capacity and data resources, IT managers want a solution that goes beyond the traditional storage/server configuration to the storage interconnect model as offered through Fibre Channel. As one CIO said, "We`re setting up two systems and find that we will have to duplicate everything on both of them. I`d rather have twice as much storage accessible by both systems." Likewise, a vice president of IT reflected, "We need some way to have multiple servers use a single tape library." As IT organizations re-centralize, they view data resource sharing as increasingly important. Another IT executive said he "likes the idea of a whole building full of storage so it looks like it`s right next to any machine we`ve got."
Fault Management Tools
IT organizations require effective management tools as part of a complete storage interconnect solution. Fibre Channel storage networks deliver significant value in being able to predict, isolate, quickly identify root causes, diagnose, and recover from problems, all from a centralized location. As one operations manager said, "With our organization, I`d like an alarm scenario so that if there is a point of failure, there is a notification back to us." Another IT manager stated, "I want something that screams at me when it`s not working." In today`s complex data center, the ability to detect and isolate a fault condition represents up to 95% of the total time it takes to recover from it.
Network performance continues to be a main concern in many data centers. With the increased amount of storage on the network and with an ever-shrinking backup window, the gigabit speed of the Fibre Channel storage interconnect offers IT organizations a needed boost. One IT executive summarized the current problem: "Although software vendors tout great backup performance, in reality we`re not seeing that." Another stated, "If you use the network to interconnect between disparate servers, then the network becomes the bottleneck. If you can connect the storage to multiple servers directly, then you eliminate that performance problem."
As IT organizations invest in larger data-sharing solutions and heterogeneous computing, Fibre Channel storage interconnects will help address total cost of ownership. Said one executive, "I would love to take six servers and attach them to a single redundant array system so I can spread that investment."
The Fibre Channel storage interconnect also eliminates the problem of loose or disconnected SCSI cables bringing entire systems down. It can attach a simple hot-pluggable transceiver module to the server, hub or switch, and storage subsystem. For example, one systems integrator said, "We have customers with sites all over the country. When the SCSI cable comes disconnected at, for example, an auto dealership, they have no idea what`s wrong. It just doesn`t work. We had one customer down for two weeks with just a bad terminator on a SCSI bus." A fault management capability embedded within a FC-AL solution would have cut this customer`s downtime to two hours or less, as opposed to weeks.
Fibre Channel has the ability to offer a higher level of visibility to the entire interconnect topology than previously available with SCSI configurations. New fault diagnostic tools are coming to market that identify problems in the storage interconnect and signal that a problem has occurred, isolate the problem, and facilitate fast recovery. These tools improve user services and reduce total cost of ownership by taking the guesswork out of problem management.
While management beyond basic status LEDs and node-based auto-bypass circuitry isn`t necessary for configurations with small-port-count networks, higher levels of port density require a more comprehensive, single-system view. Even basic device-level enclosure management only provides information on hub status--such as power supply, fans, and ports--and cannot provide useful information on the health of the loop. For loops with more than 10 nodes, loop-level management is a critical tool. It provides information on aggregate loop status such as an event log to track and record changes or events, and the identification and isolation of any problem that disrupts the loop.
Additionally, hubs should have a level of intelligence that will automatically restore the loop to a functional level by systematically interrogating the looped devices until it zeros in on the offending node. This type of interconnect architecture will help IT managers optimize application service levels, thereby ensuring that in the event of a fault condition, all users sharing the data infrastructure are not affected for any lengthy period of time.
The next step in storage interconnect management will be to add predictive capabilities so that the storage network can anticipate impending problems and provide an alert to take preemptive action. These types of tools improve the effectiveness of the IT staff by focusing their attention on problem resolution before the problem happens, thereby maintaining uptime for end users.
The Fibre Channel storage network sets the framework for enabling a wholesale shift in the way storage is managed. Moving from an event-driven (fault conditions) environment to one that is increasingly more predictive will require a different approach to storage management.
IT professionals understand that in a data-intensive world where access and availability are the underlying principles that rule the enterprise, there is the need for a change in the storage architecture. Fibre Channel-based storage networks will take center stage over the next decade in bringing about a level of user service that is not possible with today`s configurations.
In a survey of 270 systems/network administrators, conducted by Storage Systems Marketing Inc., respondents said that reliability was the most important factor in selecting network-based storage products.
In the SSMI survey, the majority of systems/network managers said they are centralizing servers and storage systems, as well as the management of those devices.
More than 80% of systems/network administrators rely on RAID for data availability, while 37% have implemented high-availability software.
The next step in storage interconnect management will be predictive capabilities that allow the storage network to anticipate impending problems and provide alerts to take preemptive action.
Jeff Vogel is vice president of marketing and business development at Vixel Corp., in Bothell, WA. Mike Befeler is vice president of Storage Systems Marketing Inc., a Boulder, CO, consulting firm specializing in the data storage market.