Transitioning from direct attached storage helped National Semiconductor slash backup times, reclaim bandwidth and realize savings that would make any CIO proud.
A few years ago, National Semiconductor of Santa Clara, CA, used only direct attached storage (DAS). In recent years, the company has gradually transitioned to a combined Storage Area Network (SAN)/Network Attached Storage (NAS) architecture. This has resulted in more rapid backups, less bandwidth being consumed by storage and overall lowered storage costs.
But National isn't resting on its storage laurels. It is in the midst of further advances, creating a hybrid storage network that utilizes a combination of iSCSI and Fibre Channel (FC).
"I see great promise in iSCSI in solving the dilemma of speed versus distance i.e. being able to manage storage over long distances while maintaining data transfers at high speed," says Chief Information Officer (CIO) Ulrich Seif. "However, the technology is not quite there yet."
Established in 1959, National Semiconductor combines analog and digital technologies — from complete systems on a single chip to high-performance multi-chip products. With headquarters in Santa Clara, California, National reported sales of $1.67 billion for its most recent fiscal year. The company has 9500 employees worldwide.
National's data center originally consisted of an IBM mainframe in tandem with Sun, NT and AIX servers using DAS. These servers were backed up over a 100 Mbps Fiber Distributed Data Interface (FDDI) network. However, a 36 hour backup was required over the network, server storage needs were out of control, network performance suffered and costs multiplied.
Therefore, the company moved from DAS to a centralized storage architecture, maintaining the FDDI network to connect end users to its servers which were hooked up to one centralized storage repository via SCSI. The mainframe also connected to the same storage pool. This storage consolidation project saves the company $3.2 million and reclaimed 18,000 square feet of data center space, according to Seif. Further, one backup server was now able to take care of all backups in a fraction of the time it used to take during the old DAS days.
However, problems remained in several areas of the storage environment. National had to keep adding NT and Sun boxes to keep up with storage demands and the steady stream of new applications being implemented at the company. Network speed continued to slow backups, and IT staff ran out of connections to centralized storage. The company decided to add a Fibre Switch to interface between the data servers and several EMC centralized storage boxes working together as a single SAN. This architecture resolved connectivity issues, permitted dual pathing, better resource utilization, load balancing and simplified storage management.
Page 2: Boosting Network Performance
Boosting Network Performance
But network performance issues persisted. To combat this, the company added a cluster of NT application servers connected to a network switch. According to Seif, this reduced TCO per desktop from $8000 in 1996 to less than $1400 by 2002.
Next came a Network Appliance-based NAS. This offered networked storage for the NT cluster. This resulted in better Windows-based file sharing capabilities and a greater ability to manage distributed applications.
"NAS is easier to manage than a SAN and provided huge benefits in terms of NT storage and backup performance," says Seif. "SAN requires very careful planning whereas with NAS you can drop it in and make it work in a relatively small amount of time."
While the SAN and NAS proved worthwhile projects in their own right, National now considers it double work to manage both separately. The next stage in the evolution of the National data center, then, is the investigating the best means of achieving of a seamless convergence between SAN and NAS. The plan is to be able to eventually manage storage across multiple platforms without effort. By doing so, National's CIO believes the SAN/NAS management nightmare and the need for separate storage pools will vanish. But that hasn't been achieved yet.
In the meantime, further upgrades have been carried out. FDDI has been replaced by an all Fibre Channel Switched (FC-SW) network that encompasses HPUX, Sun, IBM RS/6000 (running AIX) and clustered Dell servers running NT/Windows 2000. These connect via a fabric switch to several EMC boxes. The company is also investigating iSCSI as a possible alternative to FC.
National is also moving in the direction of a virtual data center concept that manages all of the company's activities around the planet. Seif envisions the use of iSCSI for speed over distance and FC for local speed.