Network Appliance—The Case for a Gorilla (Part I--Bowling Pins)  We shall examine NTAP's value chain, proprietary open architecture, and switching costs. This will be done in at least three parts. This is Part I. Throughout this presentation, links to supporting marketing, technical, press releases, and product descriptions on NetApp's web site will be provided.  I would like to field questions about each Part of this presentation before I publish the next Part. Suggestions for subjects to be covered or formatting will be considered.  TFRM, Page 52  Proprietary Open Architectures with High Switching Costs “The power of the gorilla is based on its control over a value chain.”  Let us first examine NTAP's value chain. In this installment, we will examine the value chain as it relates to NTAP's bowling pin strategy, which was successfully executed before the tornado began. In later installments , we shall examine NTAP's value chain as it relates to other technology platforms and providers.  The First Bowling Pin—ECAD NTAP recognized the need to focus on a few niche markets in 1994-1996 when they deliberately went after the engineering CAD market (ECAD). They sought and won platform certification by Parametric Technology's Pro/Engineer® and Pro/PDM™ and from Cadence Corp. for IC design tools. These early endorsements, which required considerable engineering changes, testing and certification protocols, resulted in important new wins in the mid-1990's with some prestigious firms.  NetApp decided to take advantage of the early adapters of NAS who were using equipment from NetApp's archrival, Auspex. Users of Auspex NAS servers were getting the benefits inherent in NAS, but were suffering from high upgrade costs and some problems inherent in Auspex's proprietary hardware platform. Some of these important wins are described in the following case studies:  3-COM case study (ECAD): netapp.com .”Network performance and reliability has improved dramatically with his implementation of a hierarchical topology. NSD's engineers are freed from coping with miniature servers on their desktops. Important design data is no longer dispersed across local disks. And localized file service -- using NetApp filers -- delivers on David's goals of high performance and scalable storage within single large file systems, with unsurpassed data availability, responsible data management (backup, restore, archiving, and reallocation as needed), and contained systems and network administration costs.  Cirrus Logic case: netapp.com .Cirrus Logic's Fremont, CA, facility houses over 500 semiconductor design, test, and manufacturing engineers who depend on NFS file servers to provide reliable access to their home directories, project data, and application software tools (e.g., silicon compilers, VHDL and Verilog systems, design rule checkers, SPICE simulators, etc.). Including several smaller satellite sites, Cirrus Logic has almost 900 Sun workstation NFS clients accessing over 400 gigabytes of data on two dozen Network Appliance filers; legacy servers from Sun and Auspex hold another 200 GB which were part of Cirrus Logic's network computing environment prior to their first purchase of a NetApp FAServer in 1993.  Cisco Systems netapp.com "We realized early on that there were many advantages to putting our workgroups on multiple servers rather than on one large system," said Josh Goldenhar, senior systems administrator at Cisco. "We wanted to move away from the single point-of-failure model in place at the time. The logical and cost-effective replacement option was a multiple NFS server configuration. The proven performance and reliability of Network Appliance confirmed that this was the right way to go."  Western Digital Corp: netapp.com Completing large simulations at least twice as fast enables more design iterations, leading to quicker completion of projects and/or better-refined results. By providing fast, simple, and reliable data access, NetApp filers are helping WDC realize its goal of increasing customer satisfaction by getting products to market faster, with better quality, and at lower cost.  Motorola netapp.com "My staff has increased to match the growth in the Advanced Messaging Systems Division but our administration and support load is much more manageable," according to LoCascio. It's really dramatic. I now can say we have an extremely reliable network computing system."  John Deere: netapp.com “Since moving Pro/E and Pro/PDM onto Network Appliance's data servers, user productivity has increased, downtime associated with costly disk crashes have been eliminated, and data file storage and access has been greatly simplified. In addition, overall system performance has improved, and John Deere's "power users," who previously required the PTC software and data files to be local on their individual workstations to maximize performance, have maintained comparable performance accessing their files on the NFS-mounted filers.  Cadence Corp: netapp.com . "Our workstations are now able to focus on application performance and run much faster. In addition, the workstations are now able to run multiple applications where they couldn't before due to the network topology."  According to Irv Stafford, Cadence project lead engineer, the performance afforded by the NetApp Filers, coupled with the improved manageability, make it a very compelling solution. "My response time using the NetApp Filers is as fast, if not faster than if my data was stored on local workstation disk," he said.  "As a systems administrator, one of the big advantages in going to a centralized file server approach is when I go to a client's workstation to perform an upgrade," DeBord added. There aren't as many crucial production files that you need to retrieve off the workstation, making upgrading and maintaining client workstations much easier."  The Second Bowing Pin—ISPs The internet was exploding. NetApp recognized that ISP's would benefit from the simplicity, reliability, and speed of NetApp filers and decided to attack that market. An early win was Netscape:  Netscape case study: netapp.com Dan Mosedale describes the F330's performance in Netscape's environment, even under their heaviest loads, to be "as fast or faster than the best performance of [Netscape's previous (conventional, Unix-based) file server] under a light load." And he is particularly pleased with NetApp's Snapshot™ file system feature, which facilitates full-integrity backups of online, actively-accessed data and allows Netscape users to recover their own accidentally-deleted or overwritten files without recourse to system administrators and backup tapes. Most of all, though, Netscape appreciates the reliability and consistent behavior of their F330 filers.  By allocating file service to each department or workgroup, Netscape compartmentalizes risk. No longer can a single file server bring down all of Netscape. Furthermore, the reliability Netscape has experienced with their NetApp F330 filers has had a noticeable impact. Bob Plummer, Netscape's Director of Information Systems, succinctly summarizes their experience after implementing two F330 filers: "The things which used to cause us pain are no longer painful."  GTE Case Study netapp.com GTE Internetworking has teamed with Network Appliance, Inc. to develop a highly scalable Windows NT® data access solution to scale to millions of consumer and small business users of its planned dial-up access service. Adding NetApp® filers to its Windows NT and UNIX® environment has improved system performance and reliability, while greatly reducing administrative costs. The system also provides GTE with the ability to manage disk-space quotas on a large scale, a benefit unavailable on other operating systems or platforms. This goes hand-in-hand with GTE's goal of providing "phone service quality" for Internet products to GTE.net residential customers. Test results indicate that GTE sustains a 99.9% uptime and unparalleled reliability when compared to systems costing three to four times as much.  Again, NetApp's engineers made special software provision to solve the most vexing problem facing ISPs at the time—e-mail and news feeds. Below are two articles which describe the approach and benefits of NetApp filers to solve these problems:  netapp.com Directories with a large number of entries are becoming more common. For example, an ISP (internet service provider) may have tens of thousands of users' mailboxes in a single /usr/spool/mail location. Directories with large numbers of entries pose a performance problem for generic file servers. Traditionally, Unix directories are linearly searched so that even if a name cache is present, a name cache miss results in a costly search of a large directory. Note that file creations always miss the name cache, since the file to be created does not yet exist. This paper describes a method used by Network Appliance for greatly reducing the time needed to search a directory using a space-efficient hash table on the entire directory.  netapp.com The surprisingly fast performance of the Network Appliance filer in the netnews benchmarks made the decision regarding SLAC's netnews server obvious. More important, though, the results served as a strong reminder to avoid preconceptions. Benchmarks can produce surprising results, which presumably is why many people run them in the first place. Finally, NFS, despite its age and weaknesses, can still do a remarkably good job.  The result of NetApp's successfully solving most of the e-mail and news feed performance and management problems and focusing its sales force on the ISP market place was a deep and broad penetration into hundreds of ISPs throughout the US and Europe, including the very largest ISPs on both continents.  (That experience also pointed NetApp in the direction of solving another general performance problem that the ISPs were experiencing, which was bandwidth to their points of presence (POP). The NetCache product was the result, buy I shall not address that product here, as we are focusing only on NTAP as the Gorilla in the NAS market.)  The Third Bowing Pin—Databases  Unknown to NetApp, several of its very largest customers had ported their database tables to NetApp filers in a Unix NFS environment. To all of the major database vendors, Oracle, Informix, Sybase, this was an unsupported configuration. NFS was known for its slow access speed and its lack of reliable transaction rollback/recovery for database tables. NetApp's customers understood NetApp's architecture and knew that files were not subject to these problems and so were not reluctant to move their databases from their Sun, H-P, and IBM Unix-based servers to the NetApp filers.  Oracle found itself under pressure by some of its largest, fastest growing customers to formally support NetApp filers as NFS mounts of Oracle databases. Fortunately, Oracle was already a customer of NetApp's in its software development labs, and there were some key interpersonal relationships between upper management of the two companies. Within just a couple of months, Oracle prescribed some minor modifications to the NetApp ONTAP operating system that were implemented and tested by the two companies.  These changes were not designed to correct any reliability issues, but were designed to detect the unlikely event of non-volatile RAM (NVRAM) failure on a reboot after a catastrophic failure, and to notify the Oracle host of the failure so that its rollback recovery system could detect “stale handles” and recover the transactions that were lost in NVRAM in such an event.  The result is that NetApp filers are the only NFS mount environment supported by Oracle:  netapp.com In the past, Oracle has not supported NFS due to concerns about its reliability, especially when accelerated by NVRAM. However, after careful analysis of the possible failure modes, Oracle and NetApp determined that the unique design of the NetApp filer (file server appliance) makes it well suited for database storage and together designed techniques to make the NetApp filer even safer. This allows Oracle users to take advantage of the flexibility that network storage can offer and to take advantage of the speed, simplicity and reliability of NetApp filers. Informix and Sybase soon followed suit:  netapp.com |