WD Red SMR v. CMR Part I: The Reasonably Similar
We are going to start with some general benchmarks to try and place the WD Red (WD40EFAX) performance in a larger context.
In read tests the SMR drive performs fairly similarly to the CMR based WD40EFRX.
Write tests tell a mostly similar story. As an individual drive, the WD40EFAX is performing pretty well in these benchmarks.
In PCMark8, the WD40EFAX manages to outperform the CMR WD40EFRX. The SMR drive has a much larger cache than the CMR version, 256MB vs 64MB, which perhaps helps account for the win here. In these kinds of shorter burst activity workloads, one can see how SMR may be used as a substitute.
Next, we will move on to the tests focused on the WD40EFAX and NAS RAID arrays.
WD Red SMR v. CMR Part II: The Not So Good
First up is the file copy test. Just a reminder, this test was performed as immediately as possible after completing the drive preparation process.
In the file copy test, the effects of the slower SMR technology starts to show itself a bit. The WD40EFAX turns in performance numbers that are significantly worse than the CMR drives.
Something we noticed is that the test that immediately followed the file copy test was a sequential CrystalDiskMark workload:
As you can see, with a heavy write workload immediately preceding the CDM test, the SMR drive was notably slower. In some ways, this is like timing a runner’s sprint time after running a marathon. One could argue that you may not transfer 125GB files every day, but that is less data than the video production folder for this article’s companion video we linked at the start. Still, this is a good indicator of the drive working through its internal data management processes and impacting performance.
Due to the nature of our last test, it was not performed in rapid succession with the previous two. With that said, all of the tested drives were disconnected as soon as their previous benchmarks were complete, and before plugging them back in for use in our test NAS array. During this time, scrubs were disabled for the pool and resilvering priority was completely disabled.
Unfortunately, while the SMR WD Red performed respectably in the previous benchmarks, the RAIDZ resilver test proved to be another matter entirely. While all three CMR drives comfortably completed the resilver in under 17 hours, the SMR drive took nearly 230 hours to perform an identical task.
The WD40EFAX performed so poorly that we repeated the test on a second disk to rule out user error; the second disk exhibited the same extremely slow resilver speeds. We also tested the SMR drives before and after the CMR drives to ensure that it was not a case of something happening due to the order of testing.
The only positive here is that the resilver did finish, and encountered no errors along the way, but the performance operating in the RAIDZ array was completely unacceptable. That 9 day and almost 14-hour rebuild means that using the WD Red 4TB SMR drive inadvertently in an array would lead to your data being vulnerable for around 9 days longer than the WD Red 4TB CMR drive or Seagate IronWolf. If you use WD Red CMR drives, you had class-leading performance in this test but if you bought a WD Red SMR drive, perhaps not understanding the difference, you would have another 9 days of potentially catastrophic data vulnerability.
According to iXsystems, WD Red SMR drives running firmware revision 82.00A82 can cause the drive to enter a failed state during heavy loads using ZFS. This is the revision of firmware that came on both of our drives. We did not experience this failure mode, and instead only received extremely poor performance. Perhaps that was because we were testing the use of the drive as a replacement rather than building an entire array of SMR drives. In either case, we suggest not using them.
There are other NAS vendors who are staying silent on this issue, even if they utilize ZFS and these WD Red SMR drives. We do want to point out that we likely want to see a more rigorous drive certification process at iXsystems, but also that they at least have done a good job communicating it on their blog. Still, the WD Red line reaches well beyond the iXsystems blog audience and even to local retail channels where these drives may be purchased without the benefit of online research.
The performance results achieved by the WD Red WD40EFAX surprised me; my only personal experience with SMR drives prior to this point was with Seagate’s Archive line. Based on my time with those drives, I was expecting much poorer results. Instead, individually the WD Red SMR drives Are essentially functional. They work aggressively in the background to mitigate their own limitations. The performance of the drive seemed to recover relatively quickly if given even brief periods of inactivity. For single drive installations, the WD40EFAX will likely function without issue.
However, the WD40EFAX is not a consumer desktop-focused drive. Instead, it is a WD Red drive with NAS branding all over it. When that NAS readiness was put to the test the drive performed spectacularly badly. The RAIDZ results were so poor that, in my mind, they overshadow the otherwise decent performance of the drive.
The WD40EFAX is demonstrably a worse drive than the CMR based WD40EFRX, and assuming that you have a choice in your purchase the CMR drive is the superior product. Given the significant performance and capability differential between the CMR WD Red and the SMR model, they should be different brands or lines rather than just product numbers. In online product catalogs keeping the same branding means that it shows as a “newer model” at many retailers. Many will simply purchase the newer model expecting it to be better as previous generations have been. That is not a recipe for success.
Thanks to the public outcry, WD is now properly noting the use of SMR technology in the drives on their online store, and Amazon and Newegg have also followed suit. The potential for confusion is still high though. The general population does not follow drive technology closely. The differences between SMR and CMR are fairly nuanced where regular STH readers may understand, but those regular readers are the same IT professionals that keep up on the latest technology trends in the market. Absent that context, simply putting the word “SMR” in a product listing does not help an uninformed purchaser choose the correct product. Still, it is a good step.
If you are reading this piece, and know someone who uses, or may use WD Red drives in NAS arrays but may not keep track of trends, send them this article, a chart from it, or the video. So long as there is proper disclosure and people are making an informed choice, then SMR is a valid technology. Simply saying SMR is not enough without showing the impact.