PR engine's at full throttle -- check out the author!
-----
Electronic Engineering Times
March 08, 1999, Issue: 1051
Section: System Design -- Focus: Data Management
Net appliance keeps database current
Bruce Clarke, Director of Database Marketing, Network Appliance Inc., Santa Clara, Calif.
Data-storage demands are exploding with the information technology revolution. Where once databases were the purview of payroll and inventory, they are now the crown jewels of the business enterprise.
The database imperative has always been reliability. Not only has that become even more pressing as they move from the guarded enclaves of the corporate IT department to the freewheeling environment of networks, but simplicity and performance demands are also increasing. And data-bases, once restricted to business use, have become fundamental to scientific computing, to the design of embedded systems and to any project that requires the correlation of a vast amount of data.
Database management may not be rocket science but it is certainly essential in today's scientific community. For example, scientists are using them in the growing field of bioinformatics, which tracks biological data, such as the Human Genome Project-the complete documentation of the human gene sequence. And biologists working on biodiversity and environmental protection are beginning to look for ways to access far-flung museum collections and species accounts.
Data warehouses, where all corporate data is in one place, now often store 1 to 20 Tbytes. A few years ago, examples of multiterabyte data warehouses were hard to come by-Wal-Mart is a well-known one. Now, even mom-and-pop businesses can store hundreds of gigabytes and Fortune 500 companies are commonly in the multiple-terabyte range. Even today's 18-Gbyte disk drives require large arrays in order to hold this explosion of data.
CAD work generates massive amounts of data that must be stored and kept accessible if the requirements of today's design environments are to be satisfied. Design reuse requires data that is stored in a reachable, reusable manner. Drawings, simulation input, testing output: All must be captured both for use today and for reference in the future.
The computing dilemma is that of providing very scalable solutions, which can be effectively managed, grown and protected. The amount of data to be stored is growing not by leaps and bounds, but by terabytes upon terabytes. Raw disks, although tedious, were used because of the need for complete database reliability, but they are difficult to manage. What is needed in the networked database environment is not more disks, but more storage.
In a 24-hour, seven-day-a-week world, where availability must be constant on a worldwide basis, it is not acceptable to take databases offline in order to back them up, which used to be the way to maintaining reliability. Today, the need for availability is approaching the scale of the need for reliability: Users want a device that functions so smoothly that the interaction is as simple and nonintrusive as possible, one that fulfills the needs of database storage in the way Cisco's routers simplified the challenges of data networking.
Database availability can be enhanced when the needs of the users are divided into two groups: those that simply want to access information and those that need to change existing information. The requirement for immediate query is generally greater and those users often can be served through access to the last-but-one version of backed-up data.
Performance question
The final important element is performance, which is generally less important than scalability and availability. No CIO has lost his or her job because a database performed slowly. However, unreliable database access could create career-limiting events.
Managing the databases associated with high-end CAD environments is a major new challenge and is expanding out of just the engineering department into the venue of the corporate MIS department, where it is treated the same as any other critical corporate information resource. However, performance still is important in that design turnaround times can be affected by poor database performance. So keeping up the performance is certainly of great interest to these customers.
Cisco and other companies revolutionized the networking industry by replacing general-purpose computers with special-purpose routers. Replacing a large, complex computing system designed to service multiple applications and users with a single-function device increased the performance and reliability of networking and largely helped usher in today's modern Internet.
Applying this same concept to the problem of data storage, Network Appliance shipped its first storage "filer" in 1993. Instead of raw disks, use of a special-purpose storage system takes care of all the thorny storage-management concerns. Its specialized functionality is designed to cope with the complex data-management problems arising in an Internet-linked world.
Storage appliances deliver tangible performance and storage-management benefits with their patented WAFL file systems, optimized to reduce I/O bottlenecks. The appliances off-load the overhead of file management from the CPU running the application, freeing it to tackle database-specific tasks. Unlike raw I/O storage systems, which are difficult to expand after configuration, appliances operate as single file systems that can be easily grown simply by adding new disk drives to the system.
No duplicate data
The NetApp filer also offers greater availability. Using a Network Appliance feature known as Snapshot, which maintains an online read-only image of the file system, users can cut their exposure to online backup from hours to minutes. Equally dramatic improvements in database recovery are also realized.
For data warehousing, the read-only snapshot also provides a static database image that can address most query situations. Traditionally, this requires a copy to be made of the database, but NetApp's snapshot approach doesn't require duplicate data, thereby dramatically saving disk space, decreasing costs and improving ROI. This image can stay online 24 hours each day while the database itself is refreshed in the NetApp active file system. When appropriate, a very quick operation can be performed that removes the old database image, snapshots the current da-tabase and changes appropriate pointers. In fewer than five minutes, the data warehouse is back on line again with refreshed data.
The increased emphasis on these matters is clear from recent industry announcements such as that of Sybase and Oracle Corp.'s decision to certify NetApp's approach to storage as well as Oracle's own recent unveiling of its Raw Iron appliance. Separating functions onto specific platforms is becoming more common as networks become more complex.
This strategy also applies to managing data on the Internet. For example, mirror sites are often maintained to speed local delivery of data. There are now mirror sites that are starting to perform "proactive content caching" to manage the process faster and more efficiently.
Just as we don't have one device to do all the chores in our modern kitchens, information processing is also turning to appliances. Routers, workstations and print servers were the first information appliances. Now file servers and database servers are coming into the market. Appliances are inherently more reliable, simpler to install and manage, and faster. Large general-purpose systems will always be with us as complex problems require tightly connected resources. Since a modern server system is not used as a router or a print server, why should it be used to store information? Wouldn't an appliance naturally do a better job? Whether storing CAD drawings or database information, a dedicated storage appliance will do a better job. |
