Inside Networking
June 2001

 

Tony Rybczynski

Storage Steps Into The Light

BY TONY RYBCZYNSKI


Reduce the currency of e-business to its essence, and you are left with the elemental medium of exchange -- information. How it circulates! The amount of information shared each year, according to Berkley, is now at the exabyte level, an exabyte being 108 terabytes. This is a staggering quantity, accounting for all data, audio, and video shared on a global basis. What is truly staggering, however, is that 93 percent of this information is stored digitally.

As vast as this hoard of information may seem, it is only getting larger, and the amount of stored information is bound to grow at an ever-increasing rate, as enterprises augment traditional business models with e-business models, mediating online transactions, interactions, and collaborations -- business to business, consumer to business, and person to person.

The growth in stored information poses several challenges. How do we manage the sheer volume of market, customer, and product information and multimedia content? How do we make all of this information instantly accessible to applications and users, wherever they may be? How do we satisfy the 7-by-24-by-365 availability demands that attend globalization and e-business? How do we meet the need for distributed back-up systems dictated by sound business practices and government regulations (such as in the financial services industry)?

Taken together, these challenges suggest we'll soon face a painful crunch: on the one hand, we'll need more elaborate provisions for storage (redundant systems to ensure critical information continues to be available under failure conditions); on the other hand, we'll have less time to administer over these provisions (drastic reductions in the time available for backups, updates, and maintenance).

Shifting Investments
In 1999, the storage capacities shipped by the top five storage vendors exceeded 10,000 terabytes. Not only does this represent a significant capital outlay, but also attendant management costs have been estimated by Yankee Group to be $12/MB/year. Approximately 80 percent of this storage today is what is called direct attached storage (DAS), dedicated to a particular server or mainframe over interfaces such as SCSI and ESCON. While directly attached, these can be local or remote, for example to an off-site backup data site. The remainder is connected to storage networks, dominantly to Storage Area Networks (SANs).

A SAN is a high-speed network dedicated to information management. More formally, a SAN is a combination of technologies -- including hardware, software, and networking components -- that provides any-to-any interconnection of server and storage elements. The overall trend is towards accelerating the use of SANs and towards networked attached storage (NAS). Today, NAS represents less than 1 percent of the storage market, but it promises to extend the benefits of the IP/Internet networking paradigm to storage applications.

At this point, we might ask why the idea of extending this paradigm to storage should be so attractive. To answer this question, let's take a quick look at the evolution of computing and how it has influenced storage considerations. In the early days, a single computer vendor provided a proprietary solution to a single buyer: the data processing manager. With the minicomputer, the process changed, and departments bought their own computing solution. The market transitioned to multiple solutions sold to multiple buyers, resulting in incompatible, proprietary data processing systems. Over time, users realized they needed to combine all data processing into an integrated environment. This requirement opened the door for open standards-based solutions.

Now, companies are connecting their mainframes with enterprise and department servers for distributed client/server architectures. Distributed computing along with parallel processing has resulted in a significant increase in process-to-process communications. At the same time, the data storage requirements have exploded. This new paradigm only works if data can be moved and shared quickly. The need for very high-bandwidth and extremely low-latency I/O is paramount.

Distributed computing, client/server applications, and open systems give today's enterprises the power to fully integrate hardware and software from different vendors to create systems tailored to their specific needs. These systems can be fast, efficient, and capable of providing a competitive edge. Unfortunately, many enterprises have taken a far less proactive approach with their storage systems.

Strategic Storage?
Storage, unlike a Web application server or a database system, is rarely viewed as a strategic tool for the enterprise. This view, however, is beginning to change. With the explosive growth of e-business, IT managers are working intensely to keep pace with managing the significant growth of data. They are installing high-performance storage systems to meet the demands for smaller backup windows and greater application availability. However, these systems are often complex and expensive to manage. In addition, DAS by its nature creates a performance bottleneck in providing access to data to a broad range of users and applications.

For example, an e-commerce transaction may require information from multiple databases, directly attached to application servers. As the volume of Web hits increases, the network can become choked, slowing response times to unacceptable levels -- not just for the e-commerce applications, but for every application that must send information across that same network.

In addition, there are the complications of managing information in a heterogeneous environment. Limited integration between products from different vendors requires separate management tools, and the variety of operating systems and user interfaces adds to the burdens shouldered by administrative staff. Basically, in a heterogeneous environment, the complications include investments in redundant skills and management software, which often lead to labor-intensive processes, as staff are obliged to perform the same functions multiple times across multiple servers. The results: a limited ability to rapidly adjust to new business initiatives, and a higher total cost of computing.

To improve data access and reduce costs, IT managers are now turning to storage networking, and specifically SANs, to provide the foundation for the development of the corporate information utility. The move to SANs is motivated by the need to manage the dramatically increasing volume of business data, and to minimize its dampening of network performance. Key factors include:

  • E-business -- Securely transforming internal business processes and improving business relationships to expedite the buying and selling of goods, services, and information via the Internet.
  • Globalization -- The extension of IT systems across international boundaries.
  • Movement to real-time -- The need to exchange information immediately for competitive advantage.
  • Business agility -- The ability to continually adapt, while immediately accessing and processing information to drive successful business decisions.

Storage Area Networks
A SAN is a dedicated, centrally managed, secure information infrastructure, which enables any-to-any interconnection of servers and storage systems. SANs facilitate universal access and sharing of resources, providing the flexibility to satisfy the new business requirements.

SANs are flexible because they eliminate the one-to-one relationship between individual servers and critical business data. In addition, they separate information management from information processing. SANs can be configured to provide servers in different locations with direct access to huge amounts of shared storage resources. SANs also enable direct storage-to-storage connectivity -- for example, between multiple disk arrays or between a disk array and a tape library -- allowing management activities such as backups and archiving to take place independent of any server.

SANs provide a faster, more effective way to deal with rapidly increasing volumes of information. With a separate information management network, additional capacity can be "plugged in" as needed with minimal impact on the performance of application or transaction servers, LANs, or WANs.

Performance is further enhanced by improved backup and recovery capabilities. With a SAN, backup and recovery can take place without involving either the existing LAN or WAN, or individual application or transaction servers. With information no longer tied to any one server in particular, failure of a single server is less likely to degrade information availability. In addition, SANs support advanced multi-server clustering solutions for new levels of information availability and business continuity.

Management also is greatly simplified. A SAN permits the use of a common set of tools and a single point of control to manage a centralized pool of information. Finally, SANs also permit near-real-time updates of remote disaster recovery sites for higher levels of disaster tolerance.

SANs On A Fiber Diet
SANs are most typically based on a "fabric" of Fiber Channel hubs, switches, and gateways connecting storage devices -- such as disk arrays (RAID), CD-ROM libraries, optical disk or tape libraries, or JBOD (Just a Bunch of Disks) -- servers and workstations on a many-to-many basis. Application and transaction servers are attached to both the SAN and to LANs or WANs, creating what appears to be a massive pool of data.

Fiber Channel is a family of ANSI standards, which defines a common, efficient transport system supporting multiple protocols, including IP, SCSI, HIPPI-FP, and video. It can also support raw data transport using native Fiber Channel guaranteed delivery services.

Fiber Channel, as a channel/network standard, contains network features that provide the required connectivity, protocol multiplexing, and reach. It also supports traditional channel features for simplicity, repeatable performance, and guaranteed delivery. Profiles define interoperable standards for using Fiber Channel for different protocols or applications. It runs at speeds from 33 Mbps to 2 Gbps, and 10 Gbps in the future, supporting multiple cost/performance levels, from small systems to supercomputers.

The transmission is isolated from the control protocol, so point-to-point links, arbitrated loops, and switched topologies are used to meet the specific needs of an application. A high-end SAN can scale to a total system bandwidth on the order of terabits per second. Fiber Channel delivers a high level of reliability and throughput with efficient bandwidth utilization over metro dimensions. Fiber Channel has also been applied in applications as diverse as high-performance CAD/CAE networking, movie animation and post-production, and imaging.

Beyond SANs
There is clear advantage in moving storage systems beyond the local area. This can be done over multiple lower speed point-to-point circuits through expensive conversion/gateway devices or ‘natively' over multiple dark fiber. However, two other significant storage networking trends are emerging, one based on NAS, the other on optical MAN's.

Network Attached Storage
NAS provides a common pool of storage that can be shared by multiple servers and clients, regardless of their file system or operating system. NAStorage is designed to separate storage resources from network and application servers, to simplify storage management and improve the reliability, performance, and efficiency of the network, thus increasing the overall productivity of the organization.

NAS appliances have a streamlined architecture designed for one function: to serve data files to clients in heterogeneous network environments. Powered by an operating system optimized for file I/O activity, file-serving performance is greater than that of a general-purpose server, which is designed to perform a multitude of functions. A file system is located and managed on the NAS device, and data is transferred to clients over industry-standard network protocols (for example, TCP/IP over an Optical Ethernet) using industry standard file sharing protocols (SMB/CIFS, NCP, NFS, AFP, or HTTP). This intelligence on the NAS device enables true data sharing among heterogeneous network clients.

The NAS approach offers several advantages. Separating storage from the server reduces the file serving activity, relieves I/O bottlenecks, and increases server bandwidth. CPU cycles can then be dedicated to handling application requests, resulting in improved client response time.

Because the NAS approach separates storage resources from the server, NAS decreases both the number of components and the amount of file I/O activity. These decreases reduce the probability of server downtime, and increase the reliability of the network and application servers. Because NAS servers operate independently of network servers and communicate directly with the client, files remain available in the event of network server downtime.

By enhancing reliability, the NAS approach helps minimize costs attributable not only to storage problems, but also to server failures. Storage-related problems, according to Dataquest, account for over 60 percent of server failures. And network downtime resulting from server failure costs organizations thousands of dollars per hour.

Optical MANs
Based on Dense Wave Division Multiplexing (DWDM), optical MANs can substantially simplify metropolitan networking environments for DAS extensions and SANs, as well as for NAS, providing unprecedented levels of scalability. And scalability may be improved both in terms of bandwidth and also application technology, with delivery being accomplished with carrier-grade reliability.

Optical MANs can support channel extension and storage networks running within the Enterprise System CONnection Architecture (ESCON at 200 Mbps) or via FIber CONnection (FICON at 1 Gbps) -- for "big iron" DAS environments; Fiber Channel SANs; inter-router links between campus LANs running long-reach Ethernet; Point-to-Point Protocol (PPP) over 45-Mbps links or Asynchronous Transfer Mode (ATM); and high-quality video links for employee communications and training.

Optical MANs can accept any of the above optical signals in native mode (from 16 Mbps to 2.5 Gbps), without hardware changes, and then map them onto wavelengths. Consider the example of a change from a DAS to SAN or NAS architecture (or, equivalently, from a multiprotocol to a universal IP on optical Ethernet environment). Since such a change can be accomplished without hardware changes, the investment in high-performance, ultra-high-reliability optical MAN infrastructures may be preserved.

The wide deployment of DWDM systems is also enabling the option of subscribing to a storage networking service, variations of which are being planned by some service providers. These services could provide remote mirroring, disaster recovery, centralized disk-on-demand, business continuity, and centralized tape backup. Such services could offer the benefits of extended reach, consolidation of fibers, reduction of equipment, carrier-grade reliability, a forecast tolerant architecture easily adapted to unforeseen changes in traffic, and centralized monitoring and control.

Summary
With storage predicted to account for over 75 percent of all computer hardware expenditures over the next five years, it's increasingly important for today's enterprise to have an effective strategy for storage networking to make data available whenever and wherever it's required. As networks have become faster, the bottleneck has shifted from the network to the server and its DAS. Server, storage, and networking vendors have implemented significantly different product designs to take advantage of parallel processing technology for DAS, NAS, and SAN products. Optical DWDM systems, with its convergence path to optical Ethernet, have emerged as a common high-performance foundation with the ability to support all three architectural approaches.

Tony Rybczynski is director of strategic marketing and technologies for Nortel Networks' Enterprise Solutions unit. For more information, visit the company's Web site at www.nortelnetworks.com. E-mail questions or comments to tonyryb@nortelnetworks.com.

[ Return To The June 2001 Table Of Contents ]