Our 6TB Hard Drive Face-Off Revisited

By | April 30th, 2015

6TB Hard Drive Comparison
A few months back we published our initial evaluation of 6TB drives, comparing Seagate (model ST6000DX000) and Western Digital (WD60EFRX). Both drives performed well in our evaluation of 45 drives each, but we wanted to see if the results held up as we scaled the quantity of drives. We now have 450 Seagate and 495 WD drives deployed and as you’ll see, things are a bit different this time around.

Reviewing Our Initial 6TB Hard Drive Comparison

For our initial comparison of the Western Digital and Seagate 6TB drives we installed the drives from each vendor in their own v3.0 Storage Pod. Each pod was identically configured, except for the drives being different. During the evaluation period we measured the amount of data each Storage Pod loaded each day. Our results were somewhat unexpected as the Western Digital drives loaded data faster even though they were spinning slower, 5900 rpm for Western Digital versus 7200 rpm for the Seagate drives.

Our Updated 6TB Hard Drive Comparison

For our expanded comparison, let’s first look at the current stats for each drive model as of April 27, 2015:

Comparing Seagate and Western Digital 6TB drives

The Operational Environment

As noted, the initial evaluation was done with v3.0 Storage Pods. Subsequently we switched over to v4.5 Storage Pods and the new drives from both vendors were deployed into the newer pods. One of the primary differences between the two different Storage Pod versions is the use of SATA 6Gb/s (SATA III) in the v4.5 pods versus SATA 3Gb/s (SATA II) in the v3.0 pods.

The Backblaze cloud backup service currently receives about 140,000 GB of data from our customers each day. Data arrives in similarly sized encrypted chunks. On any given day, multiple Storage Pods accept the data as it arrives. The data is passed to a Storage Pod, if that Pod is busy, the data is passed to the next pod in line. Over the course of a day this results in all available Storage Pods being given the same opportunity to accept data at the same rate.

We tracked and recorded the Storage Pods filled with the 6TB Seagate and Western Digital drives. We started with the day they went online and continued each day until the pods were 80% full. Once they reach this point, the pods are designed to reduce the amount of data they accept.

Western Digital 6TB Drives

Let’s start with Western Digital. Below we compare the Western Digital 6TB drives from the different Storage Pods models (v3.0 and v4.5). While a specific Storage Pod ID is referenced, it is representative of all similarly configured Storage Pods.

Western Digital 6TB Drives

A quick look would seem to indicate that the Western Digital drives accepted less data per day in the newer version Storage Pods. The following table sharpens that observation.

blog_compare_WD_pods

Seagate 6TB Drives

Now we’ll compare the Seagate 6TB drives in the different Storage Pods models. As noted previously, when a specific Storage Pod ID is shown, it is representative of all similarly configured Storage Pods.

Seagate 6TB Drive Comparison

For the Seagate drives, when installed in a v4.5 Storage Pod, they load substantially more data on average each day then when installed in a v3.0 pod. The following table summarizes the difference:

Comparing Seagate 6TB Drives

Comparing Seagate to Western Digital

Let’s compare the Seagate and WD 6TB drives in the v4.5 Storage Pod:

Seagate vs Western Digital

In the v4.5 Pod, the Seagate drives consistently finished in 32-34 days while the Western Digital systems took much longer. This was over the same observation period in the same environment. Breaking this down into a table we see:

Comparing Seaagte vs WD 6TB drives

Given these observations, the Seagate drives in v4.5 Storage Pod compare favorably to the WD drives in the v3.0 pod.

What Changed

This time the Seagate 6TB drives loaded data faster than the WD 6TB drives. This result was observed across multiple Storage Pods and held true even when we normalized the data to account for variances in the amount of data received during the observation periods as well as the total number of Storage Pods available for data storage on any given day.

We believe the primary reason for the difference is the use of SATA 6 Gb/s in the v4.5 Storage Pods versus the use of SATA 3 Gb/s in the v3.0 pods. More precisely, our use of SATA 6 Gb/s let us take advantage of the faster spindle speed for the Seagate drive (7200 versus 5900 rpm). This allowed the Seagate drives to accept more data on average over the observation period.

As always, we’re reporting what we’ve observed on our environment. If you or someone you know works on hard drive internals we’d love to hear your theories. Please feel free to join the conversation.

What About Western Digital?

The results do not mean Western Digital is shut out. Both the Seagate and WD drives work in our environment and both drives store data in a reasonable amount of time. Besides, once a Storage Pod is full (or nearly full), we just need the hard drives inside to keep spinning, waiting for the day when a customer needs to recover their data. It’s not exciting, but in this case that’s a good thing.

 

Andy Klein

Andy Klein

Director of Product Marketing at Backblaze
Andy has 20+ years experience in technology marketing. He has shared his expertise in computer security and data backup at the Federal Trade Commission, Rootstech, RSA and over 100 other events. His current passion is to get everyone to back up their data before it's too late.
Andy Klein

Latest posts by Andy Klein (see all)

Category:  Cloud Storage
  • Louis E.

    If you run a cloud backup service,why are you using Western Digital’s Red drives rather than their datacenter-grade Gold drives?

  • Keith L

    You know what is ironic, I have significantly higher failure rates with Seagate. I have had 3x6TB seagate drives, ordered from different warehouses so they are different batches, all fail within the first 2 weeks, only 1 out of the 4 survived. All died in the same matter, power up, clatter clatter, power off. Read head failure. I have had dozens of their 3TB drives go too. (all 7200RPM models) Yet I have 3x1TB WD Black drives going for 10 years and counting, no read errors, nothing. I have 8x6TB Black drives, all error free…and faster than the Seagate drives. This article proves nothing but the fact that they coincidentally had a few extra bad WD drives. Granted the drives I am testing are not either companies “green” or RAID drives, but performance drives. Perhaps Seagate is better at low RPM drives and WD is better at high RPM drives? But it should be noted that if you are trying to not have drive failures, you should be buying their enterprise drives, not their consumer level RAID drives, as this test is using. Also, the most reliable drives ATM are the Hitachi Helium filled drives, but who the heck wants to spend $700 vs $350 on the price difference. At that price point buy 2, run redundant and pray they both live.

  • Thomas Moloian

    to andy klein, my name is thomas from lake forest, ca and i’ve been a computer technician since 1992 and certified since 2006, yes i waited until i was fully ready to take my tests which i passed with around 80%+, both hardware and software and i will tell you that i’ve tried all makes of hard drives and know for sure that seagate is the rolls royce of hard drives and western digital sucks big time and i will gladly send you my most recet defective 1tb black caviar wd hard drive if you want proof.

    all i did was try to upgrade from windows 7 home premium to windows 10 compliments of microsoft and when the purple upgrade screen popup up i clicked upgrade and continue and everything went black and the drive stalled. i then put a seagate backup drive i had, 320 gb and it booted perfectly and so i tell you that i even spoke to western digital corporate which is located in irvine, ca., 15 min north of my residence and they told me to send in copy of receipt and they might be able to update my warranty for i was sold a refurbished drive instead of a new one and the vendor is now out of business for unscrupulous business practices via amazon and amazon has removed them since.

    bottom line: western digital suck big time and i will never own one again and i’ve already told this to western digital in irvine, and they laughed at me so to speak so now i have the last laugh.

    NO MORE WESTERN DIGITAL FOR ME! PERIOD!

    • MC_bot

      Weird, I’ve never seen a WD Black go that wasn’t DOA and only one Green which was after five years. That’s about 2% failure rate on my drives. OTOH I’ve seen 7 Seagate failures all in less time and I’ve only ever had about 15 of them so close to 50% fail rate.

      Both have good Enterprise stuff, both have terrible cheap stuff but in the mid range WD is better.

    • dwasifar karalahishipoor

      This is one of the most logic-challenged posts on the topic I’ve ever read. So much so that I started asking myself, “Is this guy a troll?” I’m looking at his Disqus history and it’s hard to say. He’s either a deep-cover troll who never breaks character, or a deeply disturbed individual.

  • frank

    What is the power-difference between a pod with 5900 and 7200rpm drives? Have you any data? By this amount of drives you have, i think that would cut your electricity-bill down.

  • William Sprakel

    I can’t buy the ST6000DX000 anywhere in Europe, any hints?

    • Andy Klein

      Try looking for STBD6000100 which is the retail package containing the ST6000DX000 drive.

    • Thomas Moloian

      william sprakel, my name is thomas and happen to do some research about computer components in europe and only a handful sell them, your model i coudn’t find except for amazon, tiger direct and a hand full of other domestic online retailers, sorry.

      computers are my business

  • Jack Hughes

    I wouldn’t have thought the rotation speed would have much effect on just sequential write, as the heads aren’t having to move around all over the disk. I’m too lazy to look up the seq. write specs for the two drives though.

    I also take issue with the statement:

    “This allowed the Seagate drives to accept more data on average over the observation period.”

    That’s only true for the V4.5 pods, not across all the combinations. The Seagate in the V4.5 accepted the data in the same time as the WD in V3 pods. (33 days, 5.12TB/day).

    The interesting question here is why the WD drives were *slower* in your V4.5 pods, than in V3.

  • Alexander Wood

    I’m having trouble understanding how the speed of the drives or SATA affects the speed of the storage pod loading data. I can saturate a gig connection with a single mediocre hard drive. You’ve got 45 hard drives and 2 gig connections. Even discounting for RAID overhead due to parity bits and all that, shouldn’t your bottleneck by the gig connections by an order of magnitude or two? Even if you were running 1.5gbps SATA connections?

  • ac

    I love reading the posts about disk statistics on this blog. Have you guys ever considered testing horizontal versus vertical orientation of disks to see if there any differences failure rates over time? I don’t think that’s ever been done before. It’s always been said that disks will operate fine in a vertical orientation, as you do with your pods, but I don’t feel like that has ever really been adequately tested.

  • Discpad

    Everything you need to know is in the 5th line of the first table:

    Annual Failure Rate — Seagate 2.80% — Western Digital 5.9%

    In the long run, given that you need the same number of new pods — About four v4.5 pods/day, it doesn’t make any real difference how fast you fill them; however given the cost of replacing a failed drive and the lost speed as the controller rebuilds the RAID array, even if the speed figures between the Seagate & WD drives were reversed (i.e. the Seagate were slower), the overall operating cost would still be lower.

    Dan Schwartz
    Editor, The Hearing Blog
    http://www.TheHearingBlog.com

    • Ulaganath krishnasamy

      I believe it more of optimization of WD sata 5900 to run efficiently . later Seagate technology take advantage of 7200rpm effective at 6Gbps. More over I was curious slower the better or faster the better. as it directly linked to how frequently you need additional storage needs to be deployed and HDD failure replacement. we also need to know how fast backbone backblaze connectivity. Does the band with is sufficient for new threaded uploads.

      Next blog it would be great on Network related as too see how fast and band with being used on daily basis before threaded upload version than old single threaded app. Also how frequently download requested with flash HDD request fulfilled across globe

    • Thomas Moloian

      thank you dan schwartz for this for it proves my point exactly, no equipment is perfect but w/d sucks big time.