Reads are way off on the right hand screen shot...
I assumed it has to do with my 16GB of RAM. If I had 24 or 32GB, it would probably still scream on the 16GB with cache enabled, right?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Upgrading RAID for video editing - which card?
- Thread starter wonderspark
- Start date
- Sort by reaction score
I assumed it has to do with my 16GB of RAM. If I had 24 or 32GB, it would probably still scream on the 16GB with cache enabled, right?
I don't understand what's happening on the graph. These shorter tests show a close-up of what happens... There is this burst at the beginning of every write, then it slows down, then it does this thing where is spikes to the floor repeatedly. If I do only 1GB tests, both read and write are insane... up in the 2000+MB/sec range. Then it does this spike to the floor.
Does that make sense to you? Do you know what's happening?
Also, on the cache disabled side, the read has this wide band of speed which is averaged by the red line, whereas it's a narrow band with the writes and the cache enabled reads.
Would it be safe to guess this is related to the 64MB cache on the drives somehow?
Does that make sense to you? Do you know what's happening?
Also, on the cache disabled side, the read has this wide band of speed which is averaged by the red line, whereas it's a narrow band with the writes and the cache enabled reads.
Would it be safe to guess this is related to the 64MB cache on the drives somehow?
Attachments
Both are related to cache. When it's on, the the blue lines (writes) remain higher than when off, as the data is moved off of the system faster (resides in cache until it's actually written to the platters).
Same with the red line in the case of the reads (card's cache doesn't help here, but the disk cache does - think read ahead).
The downward spikes in the blue lines is much wider than the spikes in the green lines, and I'm guessing that's because the write cache from the card is the 1GB, and the read cache from the disk is only 64MB, hence much more frequently having to "reload" as it were. Have I got this figured out correctly?
What's the difference between "write back" and "write through" setting on Volume Cache Mode? It seems like "write back" gives better performance.
Does write through put data on both cache and disk, and write back only puts data to disk?
Does write through put data on both cache and disk, and write back only puts data to disk?
Last edited:
In simplest terms, write back (aka write behind), is usually faster.
For further information, take a look at the following links (Yes, I'm being lazy
- but they are good explanations that should answer your questions). Cache Wiki (specifically, Write Policy section)
jboss.org
Wiki Tangosol (Oracle is the original source I think)
Thanks, that answers my questions.
I've added another 2TB drive to my RAID, and I decided to delete and create a new 7-disk one, rather than expand the old 6-disk one. Should have done 7+hotspare to begin with, but I needed to use one of the bare drives as backups during the test phase. After I finish this project and can delete a few terabytes, I'll build that RAID5 for more testing fun, but right now the RAID3 is working so well that I don't want to change it yet.
I've added another 2TB drive to my RAID, and I decided to delete and create a new 7-disk one, rather than expand the old 6-disk one. Should have done 7+hotspare to begin with, but I needed to use one of the bare drives as backups during the test phase. After I finish this project and can delete a few terabytes, I'll build that RAID5 for more testing fun, but right now the RAID3 is working so well that I don't want to change it yet.
