SAN versus NAS The Holy War Not Worth Fighting

Transcription

1 THE CLIPPER GROUP Explorer TM Published Since 1993 September 27, 2000 SAN versus NAS The Holy War Not Worth Fighting Management Summary Analysts: Mike Kahn and Anne MacFarland In the beginning, more than 40 years ago, computer vendors created the storage devices that we now call disks and the world seemed complete; in retrospect, at least simple. Disks gave computer systems a place to store and retrieve data randomly, at speeds and capacities that seemed perfectly adequate. Both disk and tape coexisted then as today via one-to-one cabling, an important parallel to our discussion. In this Land of Eden, surrounded by glass, information systems prospered through nearly three decades of data growth, until four forbidden fruits realigned the I.T. cosmos forever. First was the PC. Second was the local area network (LAN). Third was the Internet, which linked everything together, insatiably creating demand for instant access to everything. Fourth was the realization that the most valuable asset was not the computer, but the data stored on those disks, even those distributed through the enterprise. The I.T. community became disrespectful and rebellious after they were tossed out of Eden. Gradually, new ideas sprang forth, giving rise to two additional ways to connect data to the users (and their inseparable servant, the "computer", reincarnated as the "server"). The need to share certain departmental files as well as alternative network-based operating systems [NetWare, UNIX, and Windows NT] gave birth to what we now call network-attached storage, or NAS for short. More recently, from the land of too many servers, exploding database terabytes and a voracious demand for high-speed transactions came SANs, or storage area networks. A holy war, of sorts, has been raging between the NAS and SAN vendors, with both sides incessantly trumpeting victory. At The Clipper Group, we have grown tired of the war. The customer is confused by the onesidedness of each camp s claims, and rightly so. The answer is not an either or proposition. There is a place for both NAS and SAN in most, larger enterprises. In this issue of The Clipper Group Explorer, we sort out the reasons for each approach with a focus, first, on satisfying application requirements and, second, on efficiency of management. From our analysis we conclude that the real issue is how best to do both NAS and SAN. We have long been a proponent of consolidation of many smaller boxes into fewer, much larger footprints, whether servers or storage. The answer to the holy war is a quest for a consolidated storage solution that does not yet exist, but is clearly on the horizon. Read on to sharpen your vision and begin to act before your terabytes turn into petabytes. 4 Navigating Information Technology Horizons SM IN THIS ISSUE Prime Directive: Business Requirements Drive Storage Requirements...2 Applications Are Sensitive to Storage Architecture...2 The Holy War Revisited...3 Convergence Ahead...4 Conclusion...4 The Clipper Group, Inc. - Technology Acquisition Consultants Strategic Advisors 888 Worcester Street Suite 90 Wellesley, Massachusetts (781) (781) FAX Visit Clipper on the Internet at Send comments to editor@clipper.com SM

2 September 27, 2000 The Clipper Group Explorer TM Page 2 Prime Directive: Business Requirements Drive Storage Requirements You are probably tired of hearing this, but business requirements are the prime directive of wise storage acquisition. Today s successful enterprise relies on the availability of information, regardless of where it is stored, the time of day, or method of access. Your imperatives continue to be: Insuring availability, Managing the entire heterogeneous enterprise (across operating systems, databases, applications, and SAN and NAS), Simplifying the operation, and Speeding deployment, access to and distribution of information. It is a 24x7 world and your storage solutions (hardware and software) must reflect it. Applications Are Sensitive to Storage Architecture The confusion of the storage holy war has much to do with applications and platforms. Some vendors propose NAS as a solution to everything, while others propose SAN as a catchall. In fact there are common-sense guidelines that should be followed to determine whether an application is best suited for SAN or NAS. Forcefitting an application into an ill-suited solution will only create storage management nightmares. The optimal solution is to deploy the application-appropriate storage technologies including both SAN and NAS. Applications That Manage Their Own Data Spaces A storage area network (SAN) is the technology of choice for those highperformance applications, such as database management systems and transaction processing systems, which manage their own, typically very large data spaces. It s all about flexibility regarding physical placement and a focus on fastest direct access to a specific record or small set of records. These applications do not need universal access, from many environments, at the file level. They care more about block level performance and control. Applications that write to volumes (like database management systems or DBMSs) demand SAN connectivity. It s all about application availability and performance. It s also about meeting capacity demands. Typically, SANs are required to meet the rapidly growing capacity and management requirements, replacing older direct-attached storage (DAS), which has limited scaling capabilities. So a SAN should be used for DBMSs like Oracle, SQL Server, and UDB, and applications such as ERP, data warehousing and CRM, which use databases as the underlying structure. Many years of engineering have gone into designing these DBMSs to run in a block-oriented environment; databases are built for sequential reads and writes. One must consider that DBMSs are constrained with TCP/IP-based file access (including NAS), because it doesn't guarantee delivery of packets, as do SCSI and Fibre Channel protocols. With TCP/IP, you have to wait for all the packets to arrive and be reordered into SCSI (channel I/O) data. Once this occurs, the result is sent back to the database server, which goes through the same reassembly process. For example, the SQL communicates through bus I/O. If you are using SQL with NAS, the I/O needs to be translated to the TCP/IP stack, which creates enormous latency since every packet has to be translated. Add this latency to the additional latency from reordering the packets and performance can suffer dramatically, just because the wrong connectivity was used. In summary, SAN is for high-performance database access and transaction processing.

3 September 27, 2000 The Clipper Group Explorer TM Page 3 Applications Driven by Universal Access to Files Many applications are driven by a need to share access to the same files, usually for read-only access. Many internet-based applications require file sharing with web page serving is the prime example. Shared data access is typically from many, heterogeneous servers running concurrently, each wanting access to the same web pages. Think about media serving, where potentially thousands of copies of the same movie or video clip are being delivered to an equal number of users at the same time, but not necessarily at the same point in the clip. It will take many servers to deliver this same clip over and over again, in most cases pointing to a single file. So if you are serving up a large collection of web pages, or video streams, to many users at the same time across a variety of client environments, then network-based file access is required. This can be done via storage attached to a general-purpose server, say Windows NT, or by putting storage directly on the network. Let s be clear here. Disks are never directly plugged into the network. There is always a server of some sort managing that storage that appears to sit on the network. These servers range from general purpose (as with NT mentioned above), to configurable special-purpose file servers, to manufactured file-serving appliances. Sometimes they are physically separate from the array of disks; other times they are collocated in the same enclosure. The question to ask is where does the sharing occur? If it is at the file level, this is the primary reason for NAS - many, sometimes large, files need to be shared, typically by diverse servers. Examples of files that need to be shared can be found everywhere network drives in an NT environment, workgroup CAD files, and web pages shared by many web servers. Today, the fastest, easiest way to meet increasing web pages and web servers performance and capacity requirements is to use NAS. Benefits include scalability (no need to clone data to individual servers, just plug a new server into the network and you have instant access to consistent data) and managability (fewer copies of data reduces storage staffing needs). Some NAS storage solutions have pointin-time copy capability to allow for scaling the number of copies of data to be accessed allowing systems to be run in parallel. You should consider using the parallel storage to scale up for, say, the big Super Bowl ad, and then redeploy that same storage next February to meet ongoing growth demands. You can also use NAS to consolidate servers. In many environments, one highly available and scalable NAS solution can replace twenty general-purpose servers that have been dedicated to network file serving. This makes it easier to ensure that these business critical applications are available all of the time. The Holy War, Revisited Proponents of the holy war believe, falsely, that there is one answer for your enterprise s hybrid needs. They are the open access proponents versus the control advocates, each seeking to optimize to their own view of the world. Today, the former swears by NAS and the latter by SAN. Are they rational in their thinking? Yes, each in their own way. Is each the only view required by their enterprise? No! Instead of choosing sides, your focus must be on building and growing your SAN and NAS storage capabilities that provide: Full testing, integration and support of SAN and NAS applications within a common storage infrastructure. This infrastructure should be capable of including both centralized and distributed information;

4 September 27, 2000 The Clipper Group Explorer TM Page 4 The ability to add storage without interruption to on-line information processing regardless of application type, and the ability to redeploy those same assets to either architecture in the future as computing requirements potentially change; and A single administrative interface to manage the combined, extensible, mixed-type storage infrastructure. Convergence Ahead There are two interesting and opposing trends in storage, both invoked in the name of managability. First is the "storage appliance", whose theme is "plug it in and go". Here the focus is on simplicity of installation. When you need more capacity, plug in another one. (You may end up with many!) The other is the behemoth array (increasingly being called an enterprise storage server). Here the focus is on optimization of storage resources. While the first may be sized to a terabyte (or maybe two), before having to go out and buy another, the largest arrays are now approaching 40 TB and beyond. The holy war is as much about the way to achieve scalability, as it is about connectivity. Over this decade, larger enterprises have to worry about scaling to hundreds, if not thousands, of terabytes. Most require both SAN and NAS in their infrastructure. The question is how to best do both. Proceeding with small storage appliances will likely require, over the long run, much more IT staff involvement than with a smaller set of larger arrays. Storage management software may mitigate this, but there is no getting around the fact that it is harder to manage many devices rather than fewer. In addition, the ongoing worldwide challenge of finding and retaining staff increases this complexity exponentially. For now, few large organizations will be able to rely solely on enterprise storage servers. They will deploy SAN and NAS storage where they need it, mostly centralized in location, but not always, but hopefully all of it centrally managed. The Clipper Group expects that the specific application requirements for shared file access or record-level performance will not change. Instead, we expect storage vendors to move toward unified solutions, which bring both SAN and NAS access to the same very-large array. It would accept multiple types of connectivity: IP for NAS and Fibre Channel for SAN, at least. The array will be self-optimizing, with a single management interface. We believe that a bigger unified storage server is better! Conclusion Be wary of the piece-meal approach to storage procurement. This will only bring more difficulties down the road. The Clipper Group recommends that you find the most manageable, effective, and unified storage solution to satisfy your enterprise s diverse application requirements, in that priority order. Push your storage suppliers to address all three and base your procurement decisions on their experience, references, and the coherency of their answers and plans.

5 September 27, 2000 The Clipper Group Explorer TM Page 5 About The Clipper Group, Inc. THE CLIPPER GROUP, INC. is an independent consulting firm specializing in acquisition decisions and strategic advice regarding complex, enterprise-class information technologies. Our team of industry professionals averages more than 25 years of real-world experience. Our capabilities are augmented by a team of staff consultants with significant experience across a broad spectrum of applications and environments. The Clipper Group can be reached at (781) and found on the Internet at About the Authors Mike Kahn is Chairman and a cofounder of The Clipper Group. Mr. Kahn is a thirty-year veteran of the computer industry. For the vendor community, Mr. Kahn specializes on strategic marketing issues, especially for new and costly technologies and services, competitive analysis and sales support. For the end-user community, he focuses on mission-critical information management decisions. Prior positions held by Mr. Kahn include: at International Data Corporation Director of the Competitive Resource Center, Director of Consulting for the Software Research Group, and Director of the Systems Integration Program; President of Power Factor Corporation, a Boston-based electronics firm; at Honeywell Bull Director of International Marketing and Support; at Honeywell Information Systems Director of Marketing and Director of Strategy, Technology and Research; with Arthur D. Little, Inc. a consultant specializing in database management systems and information resource management; and, for Intel Corporation, Mr. Kahn served in a variety of field and home office marketing management positions. Earlier, he founded and managed PRISM Associates of Ann Arbor, Michigan, a systems consulting firm specializing in data management products and applications. Mr. Kahn also managed a relational DBMS development group at The University of Michigan where he earned B.S.E. and M.S.E. degrees in industrial engineering. Reach Mike Kahn via at MikeKahn@clipper.com or via phone at (781) Ext. 21. Anne MacFarland is Research Manager of Enterprise Solutions with The Clipper Group. Ms. MacFarland specializes in the strategic solutions being offered by enterprise systems and storage vendors. She joined The Clipper Group after a long career in library systems, business archives and research, including work for Connecticut Historical Society, Stowe Center, Aetna Life and Casualty, and Travelers Insurance. Ms. MacFarland earned a Bachelor of Arts degree from Cornell University, where she was a College Scholar, and a Masters of Library Science from Southern Connecticut State University. Reach Anne MacFarland via at AnneM@clipper.com or via phone at (781) Ext. 28. Regarding Trademarks and Service Marks The Clipper Group Navigator, The Clipper Group Explorer, and The Clipper Group Observer are trademarks of The Clipper Group, Inc., and the clipper ship drawings, Navigating Information Technology Horizons, and teraproductivity are service marks of The Clipper Group, Inc. The Clipper Group, Inc. reserves all rights regarding its trademarks and service marks. All other trademarks, etc., belong to their respective owners. Disclosure Officers and/or employees of The Clipper Group may own as individuals, directly or indirectly, shares in one or more companies discussed in this bulletin. Company policy prohibits any officer or employee from holding more than one percent of the outstanding shares of any company covered by The Clipper Group. The Clipper Group, Inc. has no such equity holdings. Regarding the Information in this Issue The Clipper Group believes the information included in this report to be accurate. Data has been received from a variety of sources, which we believe to be reliable, including manufacturers, distributors, or users of the products discussed herein. The Clipper Group cannot be held responsible for any consequential damages resulting from the application of information or opinions contained in this report.