• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

How does hard drive buffer size work?

kevman

Diamond Member
ok,

so I have this new HD thats 8mb buffer as opposed to the old 2m buffer. Why does the box on the new one say that it's only 20% faster then the old 2m buffer? shouldn't it be 4 times faster?

How does hard drive buffer work?


Kevin
 
the way I understand it is, it sort of acts like cache memory..

the communication on the bus, will "anticipate" data that will be needed, thus being stored in the 8mb buffer, being available for use.

so its basically:

PIII 256k Cache -VS- PIII 512k Cache

40 GB Hard Drive 2 MB Buffer -VS 40 GB Hard Drive 8 MB Buffer

Relate those two comparisons, and you should be able to figure it out

 
Oddly enough, the IBM page doesn't discuss the area that larger cache helps the most which is in writes, not reads. Known as write-back caching, it involves storing data being sent to the disk in cache until the drive is free to write the data to disk. By storing multiple write requests in cache, the drive does not have to stop what it is doing to write, and it also reduces the numbers of writes that are required because they are done in bunches rather than one at a time. The downside to write-back caching is if you have data waiting to be written to disk in your cache and the computer crashes, you lose the data. For this reason WinXP and Win2k+SP3 default to write-through caching which performs the write as soon as the request is sent. The terrible write performance that some people are experiencing win WinXP and Win2k+SP3 with SCSI drives is due to this issue and shows how much performance can be gained with caching writes.

The 20% faster quoted on the HD boxes is a bit optimistic and not fully realized by typical users doing typical tasks.
 
Originally posted by: Pariah
Oddly enough, the IBM page doesn't discuss the area that larger cache helps the most which is in writes, not reads. Known as write-back caching, it involves storing data being sent to the disk in cache until the drive is free to write the data to disk. By storing multiple write requests in cache, the drive does not have to stop what it is doing to write, and it also reduces the numbers of writes that are required because they are done in bunches rather than one at a time. The downside to write-back caching is if you have data waiting to be written to disk in your cache and the computer crashes, you lose the data. For this reason WinXP and Win2k+SP3 default to write-through caching which performs the write as soon as the request is sent. The terrible write performance that some people are experiencing win WinXP and Win2k+SP3 with SCSI drives is due to this issue and shows how much performance can be gained with caching writes.

The 20% faster quoted on the HD boxes is a bit optimistic and not fully realized by typical users doing typical tasks.

Pariah, mind posting a KB article or some tweaking article that describes this write-through caching performance issue for us? I was not aware of this issue and probably need to check and correct it?

techfuzz
 
I'm not sure if there is an article, as I don't know if MS acknowledges this as a bug. There are a couple of huge threads on this on the StorageReview messageboard.

Edit: I couldn't find anything in KB. I don't think this is a bug though as MS did it intentionally knowing the consequences. According to the folks over at SR, you can fix the problem by converting discs to dynamic discs.
 
For this reason WinXP and Win2k+SP3 default to write-through caching which performs the write as soon as the request is sent
This is not so. At least with xp. It is very clear whenever I unrar a file that the write back cache size limit on XP is about 40MB. After that, XP begins committing the cache. If any OS used no write back caching at all, disk writes would be mind bogglingly slow. I've read the storagereview threads and XP only has write through caching for disk metadata. The size of the write cache is much bigger (unlimited?) if the performance options on the computer are set to optimize for system cache.
 
Pariah,

I reread your response and I must see that I missed your statement about SCSI drives being affected, but no indication of IDE. After reading some articles over at SR I finally reread your post and it made sense. I was beginning to wonder what I was missing. Doesn't apply to me at all I don't believe since I thought we were talking only about IDE. My bad...

techfuzz
 
thanks for the reponses all.


Pariah,

Is converting the disk to dynamic disk essentially using WIN XP's software based RAID? Also would the write-back caching apply in a IDE RAID scenario? I'm using IDE raid with a controller card.

Kevin
 
Back
Top