Storage up to 32TB in a 14-inch tower (December 16th, 2013)
Product Manufacturer: Sans Digital
Price: $369 with PCI-E card
- - Quiet - Fast - Cost effective
- - Dust magnet - Ridiculous light pollution
Regardless of what Cupertino would like us to believe with the new cylindrical Mac Pro, there are needs for vast amounts of hard-drive based storage in one convenient array. Where the slab-sided Mac Pro can cram in eight 3.5-inch hard drives in a pinch, its not optimal, and for some, it may not be enough. Sans Digital has both mass storage and speed in mind with its Sans Digital TowerRAID TR8M6G, a fairly massive edifice devoted to maximum transfer speed without Thunderbolt in an external case.
The Sans Digital TowerRAID TR4M6G holds eight tray-mounted hot-swappable 3.5-inch hard drives, with each up to 4TB. Fully populated, the array can hold 32TB of storage. Through the included Windows 7 and greater and OS X (10.6 and up) RAID controller card, the array handles RAID 0, 1, 10, 5, 5+spare, Spanning, and JBOD configurations -- the tower itself doesn't handle RAID allocation or formatting, leaving it to either the host adapter or the OS itself. Enough power for eight drives is supplied by a 250W power supply.
Interestingly, the array is served by a pair of e-SATA connections, both with speeds up to SATA-3. While the array isn't quite universally multiplexing, if the drives in the array are configured properly, one e-SATA cable will handle one drive in a RAID, with the other handling the other. For our tests, we configured a RAID-5 for the first pass and RAID-0 on the second to compare speed.
For the uninitiated (or confused), RAID 0 is also known as a striped volume, which splits data evenly across two or more disks without parity information for speed. No data redundancy is provided with RAID 0. Striping generally provides a speed boost -- the speed gain is seen most in large file transfers as opposed to a series of smaller ones.
A RAID 5 consists block-level striping with distributed parity. Upon failure of a single drive, subsequent reads from the array can be calculated from the distributed parity such that no data is lost. The equivalent of one disk in the array is spread across all the disks in the group. Some performance is lost over RAID 0 from the parity, but the array is still faster than a single drive. RAID 5 is generally not used for applications that require large sequential data transfers -- these types of data transfer can prevent effective load balancing.
Our primary test machine is a 2009 Mac Pro with the PCI-E SATA card installed running both Windows 7, Windows 8, and OS X 10.6 through 10.9. We loaded the case with eight 2TB 7200 RPM SATA-3 drives, with 32MB of cache. This evolution took less than five minutes, including screwing the drives into the trays, and insertion. Formatting the 16TB array (or 14TB, in the case of RAID 5) took a bit longer at about 10 and 12 minutes respectively. Formatting times were consistent between Windows 7, Windows 8, and OS X 10.6. OS X 10.7 through 10.9 were a bit faster at 8 and 11 minutes for RAID 0 and RAID 5.
Following array load, we then commenced first copying 400GB of H.264 encoded video, all over 3GB in size per file. Following the large file test, we copied 400GB of MP3 and AAC files, all around 3MB in size. All tests were performed in each OS, with a fresh format of the array between tests. Our reported speeds were within three percent between all the OS variants we tried.
Speeds delivered by the array were better. RAID 5 sequential transfers of the large files peaked at 460 megabytes per second written to the newly formatted array, with 577.1 megabytes per second reading from to the array. The smaller files were written at 70.2 megabytes per second, and read back at 99.9 megabytes per second.
RAID 0 was more impressive, peaking at 542.2 megabytes per second written also to a blank array with 692.3 megabytes per second read speeds also to a blank array. The smaller files also moved faster to the RAID 0 array, as expected. Write speeds hit 150.0 megabytes per second, with the transferred files read back at 111.1 megabytes per second.
Temperatures were good, with little if any temperature excursion. At idle with no active file transfers going, the drive temperatures at peak never exceeded 24C on a thermistor reading directly on the hard drive metal casing itself in a room with an ambient temperature of 19C. At maximum load, just before completion of the file transfer, temperatures hit 31C at a peak. With the array full of drives, and an extended write perfumed on a single drive, forced by formatting the array in JBOD configuration, drive temps hit 32C after 380GB of data had been moved. The case fan seems to not vary speed by temperature, and maintains what is probably max speed at all times, with noise similar to a Mac Pro tower under heavy processing load.
No product discussion is complete without at least a nod to design. The case itself isn't that dissimilar to a PC case. Seams are sturdy with good welds and attention to detail not found in cheaper alternatives. The burnished exterior metal has a slight texture to it, which, unfortunately, holds onto dust very well. If case appearance is important, it really needs to be wiped down with a moist towel or damp cleaning cloth at least daily in all but the cleanest environments.
The case is lined with perforations on the sides, allowing for decent ventilation, driven by a 120mm fan on the back. Unfortunately, this case and specifically the fan is illuminated with blue light and is best described as a neon testament to mass storage. While we don't have a photometer for the ultimate in LED spill testing, at 2AM in a dark 12 foot by 18 foot room, with no other external light sources from seven feet away the light exuded by the rear fan is nearly bright enough to read to. We'd really like to see some sort of control switch on this light, but a perusal of both the enclosure and the documentation came up empty. This case is not intended for a case modder's enthusiast gathering, and is best suited for a medium-performance application, so the very bright light isn't really necessary.
The boxy Mac Pro is on the way out, like the G5 tower it supplanted. In its place is a svelte cylinder that eschews mass storage for unadulterated speed in both processor and its minimalist storage. This is a shame, because there are many applications which can utilize and demand large volumes of storage. Thunderbolt is expensive. The eSATA spec, while nominally slower, is very good at relatively inexpensively housing large quantities of data in a RAID environment tailored for speed or security, or a hybrid of both. The Sans Digital TowerRAID TR8M6G gives the Mac Pro and other compatible hardware with either a free PCI-E slot or a pair of eSATA connections the ability to store external to the computer up to 32TB of storage in a finely finished enclosure. For the price which includes the PCI-E card capable of driving the case to its maximum, what's not to like?