agrikk
Gawd
- Joined
- Apr 16, 2002
- Messages
- 933
Our company is building a new data warehouse and has run out of iSCSI storage in our VMware cluster, so it was time to figure out how we wanted to add more storage.
We are currently using a NetApp FAS2050 that I absolutely hate. Don't get me wrong- it's a really solid performing device and has never given us any grief, but the way it is set up seems to be a horrible, horrible waste of disk (twenty 300gb disks (5360gb raw storage). two disk aggregates each with two parity disks and a spare (down to 3192gb), two volumes using 90% of the aggregate (down to 2872gb), multiple LUNs that can use no more than 90% of a volume (down to 2585gb).
Usable space? 48% of raw
and nevermind the lack of SNMP monitoring.
So rather than spend another $20k on another shelf for this thing, I wanted to see what was out there and I found Compellent.
The short version of my love affair with this device is its ability to handle storage across multiple tiers of disk at the block level, deciding on the fly access patterns of a block and determining which tier of disk it should be stored on as well as where on a disk the block is stored. On the fly short stroking of disks sounded pretty rad to me.
Also if a disk block is accessed alot it is written across disks in RAID-10, but as its usage fades away, its stripes will be converted on the fly to RAID-5 in the background and dropped to lower quality storage tiers for long term storage, recovering disk space automatically.
Hell yeah.
So we spent a chunk of change for a fully redundant, dual-head solution with a mix of SAS 450GB 15k drives and 1TB 7K drives.
It arrived last week and we put it all together and racked it today. Next week the Compellent guys show up to install the software and configure the thing.
Here's a photo log of the work done to date.
We are currently using a NetApp FAS2050 that I absolutely hate. Don't get me wrong- it's a really solid performing device and has never given us any grief, but the way it is set up seems to be a horrible, horrible waste of disk (twenty 300gb disks (5360gb raw storage). two disk aggregates each with two parity disks and a spare (down to 3192gb), two volumes using 90% of the aggregate (down to 2872gb), multiple LUNs that can use no more than 90% of a volume (down to 2585gb).
Usable space? 48% of raw
and nevermind the lack of SNMP monitoring.
So rather than spend another $20k on another shelf for this thing, I wanted to see what was out there and I found Compellent.
The short version of my love affair with this device is its ability to handle storage across multiple tiers of disk at the block level, deciding on the fly access patterns of a block and determining which tier of disk it should be stored on as well as where on a disk the block is stored. On the fly short stroking of disks sounded pretty rad to me.
Also if a disk block is accessed alot it is written across disks in RAID-10, but as its usage fades away, its stripes will be converted on the fly to RAID-5 in the background and dropped to lower quality storage tiers for long term storage, recovering disk space automatically.
Hell yeah.
So we spent a chunk of change for a fully redundant, dual-head solution with a mix of SAS 450GB 15k drives and 1TB 7K drives.
It arrived last week and we put it all together and racked it today. Next week the Compellent guys show up to install the software and configure the thing.
Here's a photo log of the work done to date.
Last edited: