Worlds of Terror
Today I looked on horrified as I read how "shocked" some delusional people were at the fact that drive failure rates do not magically stabilize with age.
Here comes the Cap'n:
Captain Obvious says: It's not so much the drive platters but the
disk bearing which contributes to a lot of drive failures. It's
a mechanical component which wears out.
Okay so maybe not everyone reading this spent a week working with QA guys from Seagate but I did so I'm rather underwhelmed at all of this. One humorous highlight amongst the charge of tin foil hat brigade came from the Zerowait blog where Mike wonders why we haven't seen people present this stuff at storage conferences (Maybe they were busy finishing their research first?) before asking if it's because vendors have something to hide? Well FAST 07 was a storage conference (File and Storage Technologies. FAST. Get it?) and if the vendors were doing their utmost to suppress the information they've done a really bad job since most of them sponsored the event if not the very award the paper in question won. Hell, the vendors are even thanked in the acknowledgments if anyone was bothered to read that far:
We thank the members and companies of the PDL Consortium (including APC,
Cisco, EMC, Hewlett-Packard, Hitachi, IBM, Intel, Network Appliance,
Oracle, Panasas, Seagate, and Symantec) for their interest
and support.
I'll make this quite clear, when it comes to overthrowing governments and suppressing information the PDL Consortium wouldn't be the folks I'd call. You're talking socks with sandals & beards worth remembering.
Moving off that I was busy today building an EMC Avamar system. I dutifully ordered equipment from DELL which met the requirements and instead of packing things with FC disks I went with SAS drives, thereby leaving me in a situation with no spares to hand when inevitably a drive fails (Newsflash: Your drives will fail eventually. Don't panic.), but it's not for production usage and I wanted to mix things up and see what that SAS magic was like.
What it was like was just like any other bunch of disks with a single LUN created on it before it left the factory, which means they're probably all still spinning away while they bind the LUNs I created after I blew away the factory shipped ones. Even with things binding in the background as I was loading the code eventually you get to a point where your wrist watch overrides your need to see what's over the next hill and you go home for the night.
I suppose it's best not to start these things fifteen minutes before quitting time.
I'll make this quite clear, when it comes to overthrowing governments and suppressing information the PDL Consortium wouldn't be the folks I'd call. You're talking socks with sandals & beards worth remembering.
Moving off that I was busy today building an EMC Avamar system. I dutifully ordered equipment from DELL which met the requirements and instead of packing things with FC disks I went with SAS drives, thereby leaving me in a situation with no spares to hand when inevitably a drive fails (Newsflash: Your drives will fail eventually. Don't panic.), but it's not for production usage and I wanted to mix things up and see what that SAS magic was like.
What it was like was just like any other bunch of disks with a single LUN created on it before it left the factory, which means they're probably all still spinning away while they bind the LUNs I created after I blew away the factory shipped ones. Even with things binding in the background as I was loading the code eventually you get to a point where your wrist watch overrides your need to see what's over the next hill and you go home for the night.
I suppose it's best not to start these things fifteen minutes before quitting time.
