Figure 19 compares the speed of the DS-500 to probably the most common fast external storage solution used today, Firewire 400. The solid black line is the average write speed of a striped DS-500 array over 22 replications. The blue line is the average speed of a Maxtor external firewire drive over 6 replications. The dingy orange line is the speed of the Ubuntu RAID 5 NAS I built in this article. The broken black line is the write speed of the DS-500 in a RAID 5 configuration.
Figure 19: DS-500 Write Comparative Speeds
Clearly, RAID 0 directly attached storage (DAS) kicks NAS, Firewire 400, and directly attached RAID 5at least in the case of the DS-500. The difference between the solid and broken black lines in Figure 19 are the first solid data that I've seen on the costs of RAID 5 in software. It would be interesting to see a directly attached hardware RAID 5 implementation for comparison. The Ubuntu RAID5 NAS system I built used an LSI Logic hardware RAID card but, as a NAS, the Ubuntu system was not directly attached.
Figure 20 displays the comparative read speeds of the DS-500. Again the solid black line is the DS-500 in a striped configuration, the broken black line is the DS-500 in RAID5, the blue line is the Maxtor external firewire drive, and the dingy orange line is the Ubuntu RAID5 NAS.
Figure 20: DS-500 Read Comparative Speeds
The surprise in Figure 20 is that the Firewire 400 drive narrowly beat the directly attached stripe array. But the striped array is 1.13 TB in size while the Firewire 400 array is 279 GB. There are a lot of jobs for which you can buy an external firewire drive that is big enough and fast enough for all your needs. But there are jobs, with very large data sets, where the DS-500 would provide really big, amazingly fast storage.
Flashback: My Ph.D. was written using UPC scanner data from A.C. Nielsen and Information Resources Inc. In 1991 when I was crunching numbers, Nielsen had 1 TB of storage for all grocery stores, all UPCs, and all weeks in the USA. Because of cost and limitation of storage at that time, Nielsen used many tricks to save disk space. One trick was to track decimal points implicitly. Another was to encode dummy variables in hexadecimal. Another trick was to store only 3 years of data online. The Norco DS-500 gives home users today the ability to store and process massive quantities of data. The Nielsen data set, for example.