Personal computing discussed
Moderators: renee, mac_h8r1, Nemesis
Scrotos wrote:So I misunderstood overprovisioning. I thought some utility could change the SSD to report a lower capacity and BOOM I'm done. But no, the process is to just make a partition smaller than the full size of the drive. That leads me to the question of whether or not having something in a RAID will utilize it 100% even if the partition made from that RAID is smaller than what's fully available?
Flying Fox wrote:The Samsung software coming with the SSD runs on Windows and you can configure overprovisioning there?
lilteap0t wrote:It sounds like you aren't really needing a high performance database for this. An 8 SSD RAID array will be fine for most needs, but if you need really fast queries, you may want to investigate 4 SSDs in 4 RAID 1 arrays instead (1 + hot spare). For example, let's say you have a query that pulls millions of rows for two tables and two indexes. By carefully arranging the data storage - each table / index on a seperate RAID 1 array - you could really maximize the IOPS and the real benefits of ultrafast seek times of each drive. By effectively merging all drives into one big drive, you lose many benefits the SSDs have in order to simplify your storage.
mattshwink wrote:One note about the P400i - it should require 512MB cache to enable RAID-6. HP is a little strange with its support of RAID-6. Usually they require a cache upgrade to enable RAID-6 on the older controllers or the purchase of the Array Advanced pack for the newer ones.
EDIT: Also, it appears the P400i only supports 6 drives (see this http://h18000.www1.hp.com/products/quic ... 0_div.HTML)
This link also shows that a P400i in RAID-6 cannot have a hot spare assigned: http://itcoop.wordpress.com/2011/02/02/ ... y-trouble/
Matt
mattshwink wrote:Also, your vendor is nuts about the storage performance needed. I would ask for benchmarking numbers backing up the claim (and with that they should be able to give you the parameters the scenario ran under). As already been stated, though, even a single SSD would blow 8 x 15k SAS (or even FC) drives out of the water.
just brew it! wrote:@Scrotos -
Your statement that you're "running the business" on "some dumpy old homebrew server" concerns me. Perhaps this is an area that ought to be addressed first, instead of getting this one user off of his old DOS-based app?
just brew it! wrote:Also... you seem to be implying that there are standard SATA commands to tell a drive to over-provision itself. While this would be a cool thing to have, I was not aware it existed; additional info, please?
just brew it! wrote:Upon further reflection, I take back what I said about the over-provisioning. The RAID controller is going to issue writes across the entire array when the array is initially built, so over-provisioning on top of the RAID probably doesn't work. You need to tell the RAID controller not to use the entire capacity of the drives up front (assuming it lets you do that). This should accomplish the same thing as over-provisioning at the raw partition level.
Scrotos wrote:just brew it! wrote:Upon further reflection, I take back what I said about the over-provisioning. The RAID controller is going to issue writes across the entire array when the array is initially built, so over-provisioning on top of the RAID probably doesn't work. You need to tell the RAID controller not to use the entire capacity of the drives up front (assuming it lets you do that). This should accomplish the same thing as over-provisioning at the raw partition level.
Yeah, but does it really? I thought most RAIDs just marked the MBR (or whatever) and then in a 5 or 6 just wrote parity as it striped? As long as the container/array/whatever isn't filled, why would the controller be writing to the entire drives?
Scrotos wrote:I realize that's talking about an 830, but I figure it's similar enough. I don't know that I want to try issuing SATA commands to artificially limit the drive size shown to the RAID BIOS/ORCA.
# hdparm -N p78165360 /dev/sdx
/dev/sdx:
setting max visible sectors to 78165360 (permanent)
max sectors = 78165360/78165360, HPA is disabled
Scrotos wrote:You have to love any man page with so many "This is EXTREMELY DANGEROUS and will very likely cause massive loss of data. DO NOT USE THIS COMMAND" and "VERY DANGEROUS, DON'T EVEN THINK ABOUT USING IT" and "EXCEPTIONALLY DANGEROUS. DO NOT USE THIS FLAG!!" and "VERY DANGEROUS, DATA LOSS IS EXTREMELY LIKELY."
Looks like they have this version of pain for Windows, too: http://disablehddapm.blogspot.com/
I'm still somewhat undecided on what to do with this. Any thoughts?
mattshwink wrote:EDIT: Also, it appears the P400i only supports 6 drives (see this http://h18000.www1.hp.com/products/quic ... 0_div.HTML)
This link also shows that a P400i in RAID-6 cannot have a hot spare assigned: http://itcoop.wordpress.com/2011/02/02/ ... y-trouble/
Matt
physicaldrive 1I:1:5
Port: 1I
Box: 1
Bay: 5
Status: OK
Drive Type: Data Drive
Interface Type: SATA
Size: 256.0 GB
Firmware Revision: DXM04B0Q
Serial Number: S12RNEACC72439Z
Model: ATA Samsung SSD 840
SATA NCQ Capable: True
SATA NCQ Enabled: True
Current Temperature (C): 40
Maximum Temperature (C): 70
PHY Count: 1
PHY Transfer Rate: 1.5Gbps