Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Disk Drive Sector Discussion Forum -- Ignore unavailable to you. Want to Upgrade?

To: Dwight E. Karlsen who wrote (1472)	11/10/1997 7:35:00 AM
From: Mel Spivak	Read Replies (1) \| Respond to of 9256

Any opinion on HMTT & how they are positioned?

To: Dwight E. Karlsen who wrote (1472)	11/10/1997 9:26:00 AM
From: David A. Lethe	Read Replies (1) \| Respond to of 9256

>if info from a particular file is requested simultaneously from four different sources, the four users would get their info faster on a non-RAID system? Not necessarily. Lets explore this. First, SCSI's arbitration problems come into factor. Only one disk can read/write at a time on a SCSI controller. All of the cache in the world cannot get around this. SCSI-based RAMDISK cannot get around this either. The only technology available today that doesn't have arbitration is SSA. SSA-based RAID-5 performs very differently than SCSI-based RAID-5. Secondly, you have to factor in cache on the disk, in the operating system buffer(s), and in the controller card (RAID or even non-RAID). If the OS is able to keep the data in its cache, then the request will never even go to disk. This will yield the fastest result. Example: A RISC-based HP3000 computer and the HP9000 computer are the same hardware, only difference is the O/S. Well, the HP3000 can traditionally do about 3x more "work" in I/O intensive applications because of the way the OS interacts with the disk subsystem. Not 10-20% mind you, but 300+ percent. Wow! Finally, it makes a big difference if you are doing reads or writes, random or sequential.. cache is part of it, but the big penalty is that a RAID-5 write does multiple writes .. one to each disk in the stripe, then to the parity disk. Don't even get me started on block size, and the impact of multiple controllers. I could speak volumes on this, as I am doing some design work in this area. I just wanted to interject into this thread because so many people think they understand how THROUGHPUT and various RAID technologies interact. So many variables, so much bad information.