I am building a chassis (of sorts) to house an Optiplex 390 SFF board with a 35W CPU and a four disk Storage Spaces volume. The choice of mobo and CPU was a matter of opportunity+cost. I have a free PCI-e 2.0 x16 and x1 slot. I will use the x16 slot for a non-RAID controller card.
I will not consider an external housing for the volume. I know that I could release captive cash and fulfill the design requirements. The mobo and disk array will reside in an oak cabinet with sliding drawers.
I am using a four disk array to provide a slightly faster I/O. I am currently using a a (3+3)TB + (3+3)TB resilient array formatted ReFS. I will migrate to (5+5) + (5+5) after the platform/chassis is operational. The choice of 5TB disks is driven by the fact that I already own a 5TB disk. I would prefer 6TB disks but I will not need that much capacity at twice the disks' warranty periods. The archived data is 14 years old and will not grow quickly.
The mobo is an H61 chipset and does not support SATA3.
I want a four port PCI-Express 2.0 x4 controller card, but I think an x2 card will suit this application. x4 = $120, x2 = $25. I think I can keep the hundred dollar difference and not lose throughput.
I am going four disk to provide a little more throughput vs a two disk array. Currently available HDDs have a capacity that will outlive the hardware (unless somebody repeals DMCA).
The platform will serve solely as an archive host and will seldom be read (and written to at weekly/monthly intervals).. If I have my head around Storage Spaces (a four HDD resilient, or mirrored, volume), I will have headroom remaining with an x4 card, and will not saturate an x2 card. It is the I/O at x2 that I am confused by.
USB3.0 is a common interface for 4 disk external RAID but caps at 5GBs.
+
Numerous reports of disk dropouts over a port multiplier with Marvell x1 chips. I attribute this to an untrapped ATA communication error between the disk and the controller card related to saturation, but this is hypothesis. The reports seem to occur only in four disk arrays and not two disk arrays.
=
I assume the USB3.0 protocol handles the saturation problem before the signal is passed to the disk.
USB3.1 overcomes the 5GBs limit, but I do not want to introduce the complication of mounting an interface between a USB3.1 controller card and the four disks inside the chassis.
So, an x4 card seems like the best case option, but it costs $100+ to play there.
An x2 card seems to be the solution, and I can play for $25.
tldr?
The $100 Question:
Will an x2 controller card feed the four disk volume as fast as the volume can eat?
I will not consider an external housing for the volume. I know that I could release captive cash and fulfill the design requirements. The mobo and disk array will reside in an oak cabinet with sliding drawers.
I am using a four disk array to provide a slightly faster I/O. I am currently using a a (3+3)TB + (3+3)TB resilient array formatted ReFS. I will migrate to (5+5) + (5+5) after the platform/chassis is operational. The choice of 5TB disks is driven by the fact that I already own a 5TB disk. I would prefer 6TB disks but I will not need that much capacity at twice the disks' warranty periods. The archived data is 14 years old and will not grow quickly.
The mobo is an H61 chipset and does not support SATA3.
I want a four port PCI-Express 2.0 x4 controller card, but I think an x2 card will suit this application. x4 = $120, x2 = $25. I think I can keep the hundred dollar difference and not lose throughput.
I am going four disk to provide a little more throughput vs a two disk array. Currently available HDDs have a capacity that will outlive the hardware (unless somebody repeals DMCA).
The platform will serve solely as an archive host and will seldom be read (and written to at weekly/monthly intervals).. If I have my head around Storage Spaces (a four HDD resilient, or mirrored, volume), I will have headroom remaining with an x4 card, and will not saturate an x2 card. It is the I/O at x2 that I am confused by.
USB3.0 is a common interface for 4 disk external RAID but caps at 5GBs.
+
Numerous reports of disk dropouts over a port multiplier with Marvell x1 chips. I attribute this to an untrapped ATA communication error between the disk and the controller card related to saturation, but this is hypothesis. The reports seem to occur only in four disk arrays and not two disk arrays.
=
I assume the USB3.0 protocol handles the saturation problem before the signal is passed to the disk.
USB3.1 overcomes the 5GBs limit, but I do not want to introduce the complication of mounting an interface between a USB3.1 controller card and the four disks inside the chassis.
So, an x4 card seems like the best case option, but it costs $100+ to play there.
An x2 card seems to be the solution, and I can play for $25.
tldr?
The $100 Question:
Will an x2 controller card feed the four disk volume as fast as the volume can eat?