In this section, we load four identical drives in a Thecus N8880U-10G rackmount NAS. The system connects to an enterprise network via a 10GbE link. The test system also connects to the same network over 10GbE and runs the workloads remotely.
Most of the tests happen over CIFT (//name_of_NAS/test_folder) but we also run an iSCSI test at the end of this page.
In a NAS, we have much more variability in the performance. This comes from two areas. The first is the nature of moving data over a network with two systems working together. Both systems have an operating system and power management that wakes, puts to sleep, and manages bus speeds.
The second variability is the one we're looking for and testing. When you combine drives in an array, the latency from one drive feeds the others. The latency compounds with the length of the test. Drives with lower latency will process the commands faster and higher latency will rapidly build and present itself as lower performance. Our RAID 5 array with just four drives shows a typical NAS used at home or a small office. More drives show in an array exacerbate the latency difference between products.
Sequential Read Performance
Here we present random read performance in two ways. The first is by block size or the size of the data transferred in each request. The 64KB, 128KB, and 256KB sizes appear well controlled, but as we get to larger sizes, the NAS's buffer play a larger role in the file reads. This is a superior test when comparing NAS against each other.
The second chart shows 128KB blocks coming from the NAS at different queue depths. This is a better test for comparing storage media.
The two new 14TB class IronWolf HDDs are faster than the previous generation. All four IronWolf HDDs are faster than the WD Red 10TB drive array and Red Pro array built with 6TB drives.
Sequential Write Performance
In the block size test, we see less variation than the previous test where we read data from the array.
The sequential write test has much more variability with all of the arrays snaking around one another.
Sequential Mixed Performance
The sequential "sweeps" show the percent of data reads in a workload. The closer you get to a 50% read mix, the lower the performance. On the previous page, we learned the Seagate drives have more trouble with sequential mixed workloads than Western Digital HDDs. That comes through in our test with the media in a NAS.
Random Read Performance
Have you ever seen so little variation in a test before? At the very top of the 32KB area, you can start to see a sliver of the other colors different arrays. This leads us to believe the bottleneck in our block size random read test is the NAS and not the drives. The 4KB performance also shows that compared to what we measured on a single drive.
Moving over to 4KB reads exclusively in the second chart, we see the IronWolf series drives well under the two Western Digital HDDs.
Random Write Performance
The IronWolf drives perform much better in the random write test. All four IronWolf drives are right together with only a small performance difference between them.
Random Mixed Performance
The random mixed workload test shows a lot of variability, but this comes mainly from the NAS and not the array. The DRAM buffer in the NAS causes a lot of issues when comparing different arrays. The IronWolf series is clearly faster than the Western Digital Red drives.
Traditional Server Workloads
In three of the four server workloads, the IronWolf series outperforms the Reds. There is very little difference between the new 14TB generation and the year-old 12TB generation.
Multiclient Office Workload
The multi-client test is my favorite because we built most of it. The test took over a year to build and takes ten dual Xeon servers that cost many thousands of dollars. The workload comes from Microsoft Office and is a mix of Excel, Word, and Power-Point. A single PC doesn't stress a NAS or array with this mostly random workload. As more computers connect to the system, up to 120 in total, the NAS and array start to reach the throughput limits, and the latency begins to increase rapidly.
The Red Pro 6TB array shows a very consistent rise and fall in the throughput graph. The IronWolf drives are faster, but slow less consistency.
PCMark 8 Total Storage Bandwidth Over iSCSI
Our only iSCSI test uses the same PCMark 8 Storage Test that we used on the previous page. The difference here is that the target drives are in an array and tested over a network. iSCSI is one of the best options available for gamers to load a lot of games on a redundant array without placing the drives in the local PC. This places the storage in another location.
The 14TB IronWolf drives, both the Pro and non-Pro, are a little slower in this test than the two previous generations. The difference is small, but we expected the new series to outperform the previous generation. The specifications show the 12TB and 14TB drives with very similar performance.
Last updated: Sep 25, 2019 at 12:27 am CDT
PRICING: You can find products similar to this one for sale below.
United States: Find other tech and computer products like this over at Amazon.com
United Kingdom: Find other tech and computer products like this over at Amazon.co.uk
Australia: Find other tech and computer products like this over at Amazon.com.au
Canada: Find other tech and computer products like this over at Amazon.ca
Deutschland: Finde andere Technik- und Computerprodukte wie dieses auf Amazon.de