The $120 grand includes the original 14 142 gig 10K fiber channel SCSI drives as well that we have had in place for around 4 years. I added the other 12 this year for my virtualization project.

The system is almost completely redundant with the exception of the drives. Each drive is dual controller managed, with dual fiber switches and dual fiber network cards in every server. I can lose an entire fiber switch or controller and never notice an impact (thought I will get an alert). With the drives I can lose 2 in each group (2 in the 14 and another 2 in the 12) before I lose any data.

You could start up your own NASA in Cape Breton with it heh.

To the moon Alice!

And if you want to go one step further, the hardware specs of the physical server that is running this virtual machine from our SAN are:

HP DL360 G5
2 x Xeon E5440 2.83Ghz Quad core processors
32GB PC2-5300
No hard drives (booting from embedded VMware Virtual Infrastructure Enterprise)
Quad GigE port network card
4GB fiber channel adapter

I have 3 of these in a VMware High Availability cluster.

HAPPY TIMES!

Well I just had to get an Adaptec 2405 and indulge with Steve.

This was the fastest performance I could get with the GSkill SSDs in Raid 0 and running from the x58 chipset, aka software raid.



The following is the fastest I can now get from the Adaptec 2405 hardware RAID controller.



While the above compare might not be apple to apple due to the cluster size of the RAID, the results are the fastest I could obtain between software and hardware.

IMO worth the money.

So what were the cluster sizes? I'd have to say the 128MB of ram on the controller would be the reason that the spikes go away.

Nice to see the CPU usage go down to 1% from 3% as well (but not surprising).

Should be interesting to see extrapolation with 3 and 4 drives.

The drive and cluster sizes in the image name.

the Adaptec cluster size is 64k, while the X58 cluster was 4k.

Steve, you might try a 32k cluster size for your four disk array and format with a 32k sector size.

I just might, might I?

Whoooops!
http://www.ocztechnology.com/products/flash_drives/ocz_z_drive_pci_express_ssd
Shown as 1600$+tax on CC (other online site). That's about $450 more than I'm spending, and all I'm missing is the nice enclosure.

Funny how the 1TB speeds are actually slower than the 512MB speed heh.

I'm sure this bad boy will be dissected very soon and we will probably see 4 Apex(?) drives (4x64GB, 4X128GB, 4X256GB), hopefully not in hard drive cases. That would be silly having that much extra packaging. I can probably get the same effect with cardboard and duct tape (no I won't actually do that).

looks like fun, now im jealous, some day i'll adventure to it.

I've come to the conclusion that I'm going to use this as a storage drive only. From the many ramblings on the interwebs i've seen it would be best to put the inet cache and temp files on a regular hard drive to save the lifespan. That and it's the super lazy approach, I don't need to reinstall Windoze (hopefully).

Maybe somewhere down the line if they upadte the firmware to make it last longer, I'll give it another shot.

Anywho you should see some numbers in a new thread tomorrow, prepare to be shown the same effects of a "Pan-Galactic Gargle Blaster".
http://en.wikipedia.org/wiki/Pangalactic_Gargleblaster#Pan-Galactic_Gargle_Blaster
Ahh, that ol' Janx Spirit!

Seems like a bit of a waste imho. I know these drives don't have great lifespans with data constantly being written to them, but isn't the main purpose of wanting fast access time to have it when it's bring used the most (ie. with OS files which are constantly being accessed)? If the current SSD's aren't good when running as the OS drive, because of lifespan issues, than I see them as failing at their main purpose so why bother getting one especially when they are much more expensive than conventional drives?

That, and I'm lazy.

The MLC drives are supposed to have a 1,500,000 hour MTBF reliability when based upon 50GB of daily write transactions.

I wonder what the daily total of write transactions are in my system and if the MTBF is a linear inverse function of write transactions? I also assume the the 25 degree C baseline calculations for P/N junction lifespan applies, with reduction of 50% life for every 10 degrees C above 25. I recall reading the explaination for MTBF of mlc drives being greater than that of the P/N junction. The explaination was the way in which the drive controller spreads the data around the drive. It was the same article which recommended not to defrag the SSDs, and to turn off this auto defrag in Vista.

I use my SSDs for the OS and applications, with data mirrored to hard drives, and nightly backup/restore points stored to the hard drives.

I'm use to these measures as I like to over-clock and have experienced data loss and OS drive failures due to this over clocking.

I have yet to measure maximum system clocks with the SSDs, but I do think they would be slightly more robust.

If it doesn't last that long why are people getting these then? lol I'll wait even though I don't have any money but still. they'll get better as they get older & cheaper ha.