Page 4: Testing and Use, Conclusion
Testing and Use
We mentioned that the included RAID management software for Windows didn't work so well. It was buggy, it has some language errors, and doesn't seem to be that polished. Do yourself a favor and setup the TR5UTP with the jumpers. Not only will it work without software, it takes the hassle out of setup. You have the option to have a number of different array setups. RAID10/RAID1/RAID0/RAID5/RAID3/CLONE/JBOD are the options are get. CLONE is basically RAID1 with more than two drives.
We tested the TR5UTP in a variety of cases. The first was using the included RocketRAID 622 card in an average Prescott Pentium 4 system. The system was using PCIe version 1 so the throughput should be limited to around 200MB/s. What we found was that it was limited to well below that even on a 5 drive RAID0 array. The second case involved the RR622 card installed in a Asus P6T based Core i7-920 system. The RR622 card was installed in one of the 8x slots. These are PCIe v2.0 and should provide enough bandwidth to saturate the drives. Sadly, the RR622 still underperformed.
We elected to use the internal to external SATA connection and that's where the TR5UTP performance started to shine. We well present four benchmarks. Each benchmark was done on Ubuntu 10.04 with the Drive Utility benchmark program. The benchmarks basically goes through the entire drive surface (or logical surface for a RAID array) and finds the transfer speeds. The drives used were 1TB WD RE3 enterprise level drives.
This benchmark shows what the fastest possible combination would be. It is a 5 drive RAID0 array. Don't ever run a RAID0 array if you value your data. What you can see is that the controller doesn't need to be Serial ATA 6GB/s to get to the maximum speed. Granted, if you used faster drives, you might hit the wall, but not for these enterprise drives. These drives would be typical for this type of setup. Average reads of 232MB/s and writes of 201MB/s are nothing to sneeze at, though. These are very fast speeds.
Next up is a 4 drive RAID5 array. This is probably the most typical setup. It allows for a single drive to be placed for a hot spare and gives you good speed and data security. Average reads of 210MB/s and write of 180MB/s are very good for a RAID5 array. The controller handles everything onboard the TR5UTP. There is no CPU offload.
We took the same RAID5 array and unplugged the hot spare and one of the drives in the array to make it degraded. This benchmarks shows what happens if a drive fails and you have no hot spare. Write performance drops to less than 60MB/s. This isn't too terrible, but it is a very large drop. Thankfully, most operations are reads which stay near 210MB/s.
If you had the hot spare online, this benchmark shows what happens when the array is degraded and is being rebuilt. The access times and transfer speeds increase as the controller attempts to reconcile the data. This is why the graph is all over the place. It still can keep up with a load, though. It is always good to have a hot spare in if you have a drive available.
After the long rebuild time, the TR5UTP happily went back to healthy RAID5 performance. It takes hours to build a large array. That's generally the case with multi-terabyte arrays. All in all, the TR5UTP handles the test cases without issue and performs well. It is good to see a product work as advertised.
Conclusion
The TR5UTP retails for about $340 (Newegg Link, Amazon Link). This is a pretty good price for a quality 5-bay enclosure. It is more expensive than the cheaper commodity type units that offer less features and offload the parity calculation to the CPU. The TR5UTP will give you fast transfer speeds along with the redundancy of whatever RAID level you want. The hot spare feature is icing on the cake. If you're out of internal drive bays and need more redundant storage, the TR5UTP is an excellent choice.
HardwareLogic would like to thank Sans Digital for making this review possible.
We mentioned that the included RAID management software for Windows didn't work so well. It was buggy, it has some language errors, and doesn't seem to be that polished. Do yourself a favor and setup the TR5UTP with the jumpers. Not only will it work without software, it takes the hassle out of setup. You have the option to have a number of different array setups. RAID10/RAID1/RAID0/RAID5/RAID3/CLONE/JBOD are the options are get. CLONE is basically RAID1 with more than two drives.
We tested the TR5UTP in a variety of cases. The first was using the included RocketRAID 622 card in an average Prescott Pentium 4 system. The system was using PCIe version 1 so the throughput should be limited to around 200MB/s. What we found was that it was limited to well below that even on a 5 drive RAID0 array. The second case involved the RR622 card installed in a Asus P6T based Core i7-920 system. The RR622 card was installed in one of the 8x slots. These are PCIe v2.0 and should provide enough bandwidth to saturate the drives. Sadly, the RR622 still underperformed.
We elected to use the internal to external SATA connection and that's where the TR5UTP performance started to shine. We well present four benchmarks. Each benchmark was done on Ubuntu 10.04 with the Drive Utility benchmark program. The benchmarks basically goes through the entire drive surface (or logical surface for a RAID array) and finds the transfer speeds. The drives used were 1TB WD RE3 enterprise level drives.
This benchmark shows what the fastest possible combination would be. It is a 5 drive RAID0 array. Don't ever run a RAID0 array if you value your data. What you can see is that the controller doesn't need to be Serial ATA 6GB/s to get to the maximum speed. Granted, if you used faster drives, you might hit the wall, but not for these enterprise drives. These drives would be typical for this type of setup. Average reads of 232MB/s and writes of 201MB/s are nothing to sneeze at, though. These are very fast speeds.
Next up is a 4 drive RAID5 array. This is probably the most typical setup. It allows for a single drive to be placed for a hot spare and gives you good speed and data security. Average reads of 210MB/s and write of 180MB/s are very good for a RAID5 array. The controller handles everything onboard the TR5UTP. There is no CPU offload.
We took the same RAID5 array and unplugged the hot spare and one of the drives in the array to make it degraded. This benchmarks shows what happens if a drive fails and you have no hot spare. Write performance drops to less than 60MB/s. This isn't too terrible, but it is a very large drop. Thankfully, most operations are reads which stay near 210MB/s.
If you had the hot spare online, this benchmark shows what happens when the array is degraded and is being rebuilt. The access times and transfer speeds increase as the controller attempts to reconcile the data. This is why the graph is all over the place. It still can keep up with a load, though. It is always good to have a hot spare in if you have a drive available.
After the long rebuild time, the TR5UTP happily went back to healthy RAID5 performance. It takes hours to build a large array. That's generally the case with multi-terabyte arrays. All in all, the TR5UTP handles the test cases without issue and performs well. It is good to see a product work as advertised.
Conclusion
The TR5UTP retails for about $340 (Newegg Link, Amazon Link). This is a pretty good price for a quality 5-bay enclosure. It is more expensive than the cheaper commodity type units that offer less features and offload the parity calculation to the CPU. The TR5UTP will give you fast transfer speeds along with the redundancy of whatever RAID level you want. The hot spare feature is icing on the cake. If you're out of internal drive bays and need more redundant storage, the TR5UTP is an excellent choice.
HardwareLogic would like to thank Sans Digital for making this review possible.