In a storage box is an IDE architecture
slower then FCAL performance?
If we compare IDE to FCAL it doesn't make any difference whether the storage is physically in a box or spread over several yards, so the box itself has no implication on the performance.
First let me clarify what IDE stands for. IDE is the abbreviation for "Integrated Drive Electronics" and is the drive part that matches ATA on the controller side. FC-AL means "fibre channel arbitrated loop" and is a way of interconnecting devices in a fibre channel environment.
That means we must look at the ATA/IDE pair and the fibre channel side. ATA/IDE was developed with the goal of reducing costs. It's meant to be a low-cost, moderate performance interface primarily used in desktop PC's. For this reason, ATA/IDE doesn't perform very well when it comes to high throughput and multi I/O from an architectural point of view. (High throughput and multi I/O certainly weren't objectives for the development of ATA/IDE).
On a single drive ATA/IDE can handle a throughput of up to 16.6 Mbps. FCAL on the other side goes up to 100 Mbps. ATA/IDE does not support overlapping operations and is not well suited for high throughput environments with many disks. FCAL Hubs (I/O controllers) on the other side are capable of multi I/0 and the data between controller and devices is transferred with the 100 Mbps technology. The only disadvantage of FCAL compared to a fabric is the FCAL works similar to Token Ring Networks, where only two devices can exchange data at the same time (that is one controller/device connection at a time). But FCAL is well above the speed of ATA/IDE and the performance gain is greater when more disks are connected in an array. A practical maximum for disks in an FC-AL is around 15-25.
I would suppose therefore that an ATA/IDE box with the same amount of disks is remarkably slower than a FC-AL attached architecture.
This was first published in October 2003