Sata High Capacity Storage / Hybrid Storage
High performance multiuser filerserver/ archiveserver in Windows AD environments with Windows ACL/SID support,
NFS or iSCSI for mediastorage or backups
Use a19" Storagecase with 3,5" backplanes from Chenbro, DataOn, Dell, HP, Norco or SuperMicro like the 24bay SuperMicro SC846A-R1200B with up to 144 TB Raw (24 x 6TB) or a Chenbo 50bay Case up to 288 TB Raw capacity. Both support expanderless operation with HBA controller and offer a redundant power supply.
Add a serverclass mainboard (socket 1150 with max 32 GB ECC RAM or socket 2011 with up to 128 GB ECC RAM), 1 to 6 LSI HBA controller in IT mode and best a 10GbE Intel nic. If you are using hardware from DataOn, Dell or HP, check either for compatibility with other Solaris based distributions or if they use LSI (or rebranded from IBM or Dell) HBA diskcontroller without hardwareraid like a LSI 9207 or LSI 9211 with IT firmware - otherwise replace these two critical parts.
The price of a working 24bay system without disks is possible up from about 1000 USD (Norco) to over 5000 USD (Chenbro 50bay case). To improve performance of slow Sata disks, add as much RAM as Read cache as possible (use max 128GB) otherwise use SSDs as readcache (hybridstorage).
A professional storageserver based on a SuperMicro SC 847 with a mainboards from the SuperMicro X10 line with 32 GB ECC RAM and 48 TB RAW-capacity and HBAs from LSI (9207) or IBM (M1015) costs up from about 5000 USD. Usable capacity depends on selected raid-configuration and amount of local disks.
A ZFS server is usually built from local disks with one or more SAS HBA disk adapters and optional an SAS expander. ZFS builds raidarrays (vdevs) and pools from these disks. If you want to increase capacity or performance you can add more vdevs from local disks.
For special demands on performance, capacity and availability you can separate ZFS file and server-services from local disks by using one ore more ZFS server as storagehead where you do not need local disks beside bootdisks. Disks were provided from independent ZFS storageboxes called storagenodes offering their capacity via FC, IB or iSCSI as a blockdevice target. From storagehead view they are working like a huge and ultrafast local disk. They can be used to build a basic pool from a single target or you can build mirrors or Raid-Z from them. You can connect your storagehead(s) with your storagenodes directly or over a switch either a dedicated SAN network or your regular LAN. Storagenodes can be located several kilometers away from your storagehead(s). An optional spare storagehead can take over functionality manually within a very short time (keep it simple, a preconfigured backuphead must only be powered on) or automatically when using avtive/active HA.
- 19" Storage Case + hotplug backplane (up to 50 local disks without expander, more with expander)
- Serverclass mainboard with i3 or Xeon, ECC RAM
Socket 1150 or Single CPU 2011 ex http://www.supermicro.nl/products/motherboard/Xeon3000/#2011
- Intel Nic 1G but prefer 10 GbE (example Intel X540-T1 or T2)
- opt 10 GbE Switch ex Netgear XS708E
- one or more LSI HBA ex. LSI 9207 or IBM 1015 (reflashed 9211-IT mode)
- optional SAS expander or a storage head solution