Why Zeus was experiencing crashes was not as important as the possible consequences of a definitive hardware failure. I used to stores on it thousands of personal photos and videos, as well as all ripped CD/DVD, softwares, PC backups, many VMware VM, etc.. And I was using a RAID 5 based on onboard controller.

1. In case of hardware failure, I would have been unable to access my data and had to find a new motherboard with exact same controller to be able to rebuild the RAID 5.
2. In addition, although an (not true) onboard “hardware” RAID 5 was offering good access performances in normal conditions, it was deadly slow after a crash as it has to check all the data.
3. Finally, I was not able to put more than 12 disks in Zeus’ case.

So, I decided to build a new home server,

1. with a software RAID to not depend on any hardware. The idea is that I could replace dead pieces, if any, with any other one (possibly not identical chipset, …) and still be able to access the data
2. with a very large case where I could add new disks whenever required without relying on special internal multi-bays (like the Icy dock’s ones..) or external drives.

From the hardware point of view (For my needs; home server = not true server = desktop hardware):

• For the case, I bought a case like the Norco Case RPC 4224 with 24 hot-swappable Sata/Sas III drive bays sold by X-Case. They sell that case in two versions, one for home server (RM424s) and one for servers (RM424). I took the server version as the home one was not available immediately (My brother bought one too… Both cases did a very long delivery trip 🙂 ).
  • 4U rackmount design
  • Support EEB (12″x13″), CEB(12″x10.5″), ATX (12″x9.6″), Micro ATX (9.6″ x 9.6″), Mini-ITX (6.7″ x 6.7″) motherboards
  • 24x hot-swappable Sata 6Gb/s (compatible Sata 3Gb/s, 1.5Gb/s) / SAS drive bays
  • Six internal SFF-8087 Mini SAS connectors support up to twenty-four 3.5″ or 2.5″ Sata 6Gb/s or SAS hard drives, mounted on horizontal backplanes for better ventilation.
  • Hot-swappable HDD tray with special designed power-off and lock mechanism + LED indicators for power and activity on each HDD tray.
  • 3 Ball-bearing cooling fans for better ventilation in the case and 2 cooling fans
  • Two front USB ports
  • Redundant 4pin molex PSU connectors support redundant power supply
  • Screwless top cover
  • Smooth border prevent lacerating your skin

I decided next to reuse the motherboard, RAM and CPU of my current PC (Chaos) — a good opportunity to buy a new motherboard supporting new processors/RAM/devices to update Chaos

• an old Asus Striker II Formula with
  • CPU: Intel Socket 775 (for Core 2 Quad/Core 2 Extreme/…)
  • Chipset: NVIDIA nForce 780i SLI
  • Bios: version 2042
  • Front Side Bus: 1333/1066/800MHz
  • RAM: 4 x DIMM DDR2 (Max 8 GB) – Dual Channel Architecture.
  • Storage: 1 x UltraDMA 133/100/66/33, 6 x Sata 3Gb/s ports NVIDIA MediaShield RAID with Raid 0/1/5/10/JBOD.
  • Slots: 2 x PCIe 2.0 x16 (mode dual x16), 1 x PCIe x16 (mode x8), 2 x PCIe x1, 2 x PCI 2.2
  • LAN: 2 x Gigabit LAN
  • Audio: 8 channels HD (SupremeFX II Audio Card ADI 1988B 8 plugged in one of the PCIe x1 slot).
  • IEEE 1394: 2 x 1394a ports (one at back panel, 1 onboard)
  • USB: 10 x USB 2.0 ports (6 at back panel, 1 onboard), 12 x USB 2.0 (6 at back panel, 6 on MB).
• 1 old CPU Intel Core 2 Quad Q6700 8MB Cache (LGA 775) – 2.66Ghz.
• 1 old Zalman CNPS9500 LED CPU Cooler
• 4 old x 1GB Ram DDR2-800 CL5 (5-5-5-15 at 333MHz) Kingston HyperX Blue in Dual Channel mode.
• 4 new x 2GB Ram DDR2 1066 CL5 (5-5-5-15-2N) G.Skill F2-8500CL5D-4GBPI-B in Dual Channel mode.
• 2 old x 150 GB HDD Western Digital VelociRaptor (WD1500AHFD) Sata 1.5Gb/s 10.000RPM 16MB Cache (Read 128MB/s, Write 142MB/s) in Raid 0 for the OS (Windows Server 2012 Essentials)
  • They are controlled by 2 onboard Sata controllers
  • They are mounted inside the case, next to the motherboard, but not in one of the 24 bays.
• 1 old ATI Radeon 9600 256MB 128-bit DDR AGP

• To control 24 HDD, I will
  • reuse a Promise FastTrak TX4660 Sata/Sas II controller which was installed in the old Zeus.
  • use a new LSI SAS 9201 with 16 Ports Sata/Sas 6G.
  • use a new LSI SAS 9211 with 8 Ports Sata/Sas 6G.
  • Notice: I could not reuse the 4 remaining onboard Sata controller with a (reversed) cable SFF-8087 Mini SAS to 4 Sata :(.

• Disk (named Unit-of-Risk: UoR) will be formatted with NTFS and FlexRAID is going to be used  to create a smart RAID system protecting data with a snapshot model.
  • Disks with data (named Data-Risk-Unit – DRU), once removed, will be readable from any other PC using any other kind of Sata controller.
  • Parity (stored on disks named Parity-Protection-Unit – PPU) will be computed only once a day (Snapshot model), not slowing down the data access (Notice: Real Time model is also supported). Notice: PPU must be as large as the biggest DRU.
  • New DRU can be added at any time: data won’t be erased.
  • Support for multiple parity levels (E.g.: with parity level 3, no data will be lost if maximum 3 disks fails at the same time).
  • So:
    • Failure of one UoR does not affect any other UoR in the array.
    • If you lose more UoR than supported by the parity level, you only lose those “extra” devices. All your remaining devices will be healthy, and the data on them will be fully readable.

• Disk (named Unit-of-Risk: UoR) will be formatted with NTFS and tRAID is going to be used  to create a smart RAID system protecting data in real time.
  • Disks with data (named Data-Risk-Unit – DRU), once removed, will be readable from any other PC using any other kind of Sata controller.
  • Parity (stored on disks named Parity-Protection-Unit – PPU) will be computed in real time with some impact on the performances but with a high protection level. Notice: PPU must be as large as the biggest DRU.
  • New DRU can be added at any time: data won’t be erased.
  • Support for up to 2 PPU, so data are not lost if maximum 2 disks fails at the same time.
  • Data can still be accessed in real time although up to 2 disks have failed thanks to a live reconstruction feature.
  • So:
    • Failure of one UoR does not affect any other UoR in the array.
    • If you lose more UoR than supported by the parity level, you only lose those “extra” devices. All your remaining devices will be healthy, and the data on them will be fully readable.

Here above “old” only means “re-used from the previous lifes” of Zeus or Chaos…