Speeding Storage Access With Magneto-Optical Jukeboxes

Automobile air bag manufacturer AutoLiv requires that its data be accessible for at least 20 years. But the company's storage system was lacking security, and the retrieval process was taking two days. Now, employees can access 20-year-old data within minutes.

It was like searching for a needle in a haystack. And, according to AutoLiv, the search was simply taking too long.

AutoLiv is an automobile air bag manufacturer for the United States and Canada. The company continually collects data, such as process variables, attributes, workstation setup, serial number tracking, and shipping information. The company also requires that this information be accessible for at least 20 years. (According to AutoLiv, it's this data that sets them apart from the competition.) However, there were two problems – the storage system and the retrieval process.

AutoLiv was writing data to 8 mm and half-inch tape. The first step in the retrieval process was to search for the correct tape. The company then had to free disk space on the production system and move the data from the tape to a disk. Many times the company loaded the wrong tape, or the data was not in the right format. It often took two days to retrieve the data for one request – and the company got several requests each day. According to one AutoLiv representative, this was unacceptable.

The Real Search Begins
AutoLiv began searching for a new system to provide archival and permanent storage. The company wanted to migrate to an Oracle database – hoping to take advantage of data warehousing. More specifically, AutoLiv was looking at an Oracle Version 8 that provides a partitioning option. But nobody offered drive mapping the way that AutoLiv wanted. Because of the A to Z limitation, the company was not interested in using the drive letter when accessing drives. The drive letter can change, without prior notice, as the configuration of the server is modified (either by adding or removing a hardware component). AutoLiv preferred a native NT file format for optical platters. However, each integrator that AutoLiv spoke with, offered systems that read the data from a magneto-optical (MO) drive. They then moved it to a local drive to be cached. Nobody would work with Oracle or make the data and disks look like local drives.

Never Say "Nobody"
Finally, AutoLiv discovered KOM, Inc. (Kanata, ON). AutoLiv asked that someone from KOM visit the site to help install the system. KOM agreed.

The entire installation took two days to complete. Reseller IKOM Office Solutions (Salt Lake City) provided AutoLiv with a Dell 6300 (with dual central processing units). KOM integrated an Oracle database (running version 8.1.5) to access data on the hard drive and MO. The hard drives are Fibre Channel drives. The system also included a Hewlett-Packard 1200 EX with six readers and writers, and a jukebox – capable of holding 238 MO platters. AutoLiv was also able to use Microsoft's NT 4.0.

There was very little integration involved. However, AutoLiv did admit to having some problems with the system. Oracle needed a few patches. "AutoLiv could run the system without the patches, but it was not as optimal until the patches were provided," commented Kamel Shaath, director of engineering at KOM. So, KOM fixed the problem.

AutoLiv is now using drive names instead of drive letters. This allows the security of being invisible to the local system utilities, such as Disk Administrator, and File Manager (while maintaining full access to the data coupled by file system security). The drive name is associated with the virtual disk and is retained on the media. It remains unchanged, regardless of the location. This expands the capabilities of Windows NT. "We have calculated this to allow 5064 potential drives on one server," added Shaath.

AutoLiv also mentioned that it is waiting for an upgrade to its software. The company wants to be able to fill up the jukebox, then move data to an electronic filing cabinet. But for now, AutoLiv is happy with the installation. Data that used to take two days to retrieve, now takes only minutes.