Scaleoutme

Enterprise storage as we know it is dead. SAN and NAS devices have had their day, but that day is done, and I will be the first one to dance on their graves. I am done with these behemoth monolithic devices. HP, NetApp, Hitachi, EMC it doesn't matter, they all share a simple design principle: Put a ton of disks behind a couple servers. This principle perhaps isn't dangerous on it's own, but mix in misplaced CYA with the storage sales engine and you're left with chaos. We pay 3-4x what a disk should cost. We then take all our corporate data and put it on these devices. We in-turn leverage all kinds of silly feature that shouldn't be done on a storage target and tie business processes to it. Short and curlies anyone? In addition to the aforementioned reasons, another main cause of our pain is most well supported storage protocols suck at scaling. There have been some recent cool ideas to band-aid the shortcomings, ie  Coho Data will use the network (SDN) to make ...

My basic goal of doing this was to create a branded/managed OpenDNS enabled guest network on VRF-Lite. I didn't want to put any dependencies on the global routing process, I also wanted to avoid using any kind of OpenDNS Update Client for my cheapo adsl drop. In case anyone is curious this is a working config on c2900-universalk9-mz.SPA.153-3.M1.bin. Not sure why anyone else would ever try to do this, but this may be helpful if you do. ip vrf GUEST_VRF ip dhcp excluded-address 192.168.124.1 ip dhcp pool GUEST_NETWORK vrf GUEST_VRF network 192.168.124.0 255.255.255.0 dns-server 208.67.222.222 208.67.220.220 default-router 192.168.124.1 Create a vrf and guest network dhcp pool. ip domain lookup source-interface FastEthernet0/0/0 ip domain name guestvrf.net ip name-server 208.67.222.222 ip name-server 208.67.220.220 ip ddns update method DDNS_OPENDNS HTTP add https://USERNAME:PASSWORD@updates.opendns.com/nic/update?hostname=SITECODE interval maximum 0 1 0 0 I wanted...

Local write performance on the server is quite good. The disks are all owned by one cluster node. The storage spaces layout is a 66 disk storage pool with 4 hot spares. The volume tested is triple mirror with 22 columns and a 256KB interleave (may play with this setting). I tested with 128KB to match backup software write size. blank spots are the 4 hot spares ;)

Our new backup target has a little space :)

Another little script to initialize the TPM and enable bitlocker for local volumes. Again, no hard coding of drive letters. Initialize-TPM Get-BitlockerVolume | Where VolumeType -eq "OperatingSystem" | Enable-Bitlocker -TPMProtector -UsedSpaceOnly -SkipHardwareTest Get-BitlockerVolume | Enable-Bitlocker -RecoveryPasswordProtector -UsedSpaceOnly -SkipHardwareTest Get-BitlockerVolume | Where VolumeType -ne "OperatingSystem" | Enable-BitlockerAutoUnlock Get-BitlockerVolume | Resume-Bitlocker

A little powershell computer startup script to backup bitlocker recovery keys to AD. Wanted to see if I could make it work without hard coding drive letters. $BLVOL = Get-BitlockerVolume | Where ProtectionStatus -eq "On" | Foreach { New-Object PsObject -Property @{ MP = $_.MountPoint KPID = ($_ | Select -ExpandProperty KeyProtector | Where KeyProtectorType -eq "RecoveryPassword").KeyProtectorId ;} } $BLVOL | Foreach {Backup-BitlockerKeyProtector $_.MP $_.KPID}