Which hard drive is best for my Synology DiskStation NAS?

avatar

Overview
This blog post will explain some of the major detailed differences between various hard disk driveRef: 1 (or hard drive, “HDD”) classes and technologies available on the market, as well as what considerations are needed when selecting a hard drive for storage. I admit this post is a lengthy read – but it contains valuable information about the various factors when it comes to looking for a hard drive to purchase.

Table of Contents


Situation
Here are the most commonly asked questions by our users when considering a Synology DiskStation NAS.

  • What hard drive should I use with my DiskStation?
  • What is the difference between these hard drives?
  • Is one hard drive brand better than the other?

Given all these questions, I’ve compiled a list of common questions that were asked about hard drives and will explore the answers this blog. At the end of the blog, I’ll make suggestions about which hard drives to use for which application.


Attributes of Hard Drives

How does spindle speed affect hard drive performance?
Often asked is whether different spindle speeds affect transfer speeds, to which the answer is not appreciably. This is because spindle speeds mainly affects how fast the hard drive will find given a sector. Once the sector has been found, the reading/writing speed of the data will remain consistent, at least for the sake of this discussion.

  • To satisfy my curiosity about the performance difference between energy-conscious HDDs versus the standard performance HDDs – a while ago, I conducted a performance test between the WD10EACS and the WD1002FBYS. The results of the performance test resulted in a less than 3% difference while using the two drives, and for handling throughput performance testing, less than a 3MB/Sec difference. While it’s a measurable performance difference, it’s not much in everyday environments as fully saturating a DiskStation’s bandwidth capabilities seldom occurs on a daily basis.
  • Where 7200rpm drives will show greater performance over the energy-conscious drives, is in improved IOPS performance. Higher RPMs on drives will lower the reaction time, or “seek time”, of a drive to find a sector – and using the previous theoretical mechanical math formula for calculating IOPS,Ref: 2 we can see that having 7200rpm drives will greatly improve the IOPS performance.
    • WD10EACS, assuming 5400rpm (I assume 5400rpm for Western Digital drives that features variable RPM IntelliPower technology. This reason is that Western Digital originally stated that these drives has spindle speeds between 5400-7200RPM or does not specifically disclose the RPM valueRef: 4, and I’m assuming the worst for my math), has a theoretical mechanical IOPS of 60 IOPS, whereas the WD1002FBYS has 80 IOPS. Having higher spindle drives will affect IOPS performance, useful for database applications (such as MySQL usage), or in Virtual Storage Applications (such as iSCSI, and NFS) as these applications are more IOPS driven, than throughput driven.


Are energy saving hard drives worth it?

  • To compare the difference between a regular hard drive and an energy-conscious drive, let’s pick two Desktop-class drives (DCDs): the WD10EALXRef: 3 which spins at 7200rpm and the WD10EARSRef: 4 which spins at a variable rate is optimized for energy-conscious applications. In my theoretical SMB scenario, the drives are spending a cumulative total of two hours a day conducting read/write access, and idling for six hours.
    • During access, the energy savings for the Green Drive is 1.5 watts. While idling, the energy savings is 2.8 watts.
    • Assuming 8 hours a day, 5 days a week, 52 weeks out of the year, the energy savings is 5.1kWh per year. Assuming that energy costs $0.1176 USD/kWh,Ref: 5 the yearly energy savings when comparing these two drives is $0.60 USD per year, when using the WD10EARS.
  • To compare at Enterprise-class drives (ECDs), I’m going to select the HUA5C3030ALA640Ref: 6 and the HUA723030ALA640,Ref: 7 where the access energy savings is 5.1 watts with the HUA5C3030ALA640.
    • As these drives are intended for Enterprise usage – we can assume 24 hours a day, 7 days a week, 52 weeks out of the year – the energy savings is 44.5kWh a year. 44.5kWh of saved energy equates $5.23 USD saved per year per drive.
    • Using the same SMB scenario applied to Hitachi example – then 24.6 watts would be saved each day when comparing the two. This equates to 6.3kWh saved per year, which saves about $0.75 USD.
  • What this means is that energy savings are very possible when using energy-conscious drives. However, it will be up to the Storage Administrator to evaluate what is more important: Saving time for processing data, or reducing total operating costs.
    • Besides the direct energy savings by the drives, there are also secondary energy savings when cooling systems are considered. Energy-conscious hard drives produce less heat – and therefore, require less cooling, which ultimately lowers operating costs in the long term.
    • The data reflected here primarily applies to large-scale systems where hard drives are operating all day. As these hard drives are operating all day – the energy savings from energy-conscious hard drives are appreciable and can quickly lower total operating costs. On a smaller scale, because there are fewer drives used and less operational time, there are fewer benefits.


Do I need SATA 6Gbit/s for better performance?

  • Often asked is whether SATA 6 Gbit/sRef: 8 mechanical drives will perform better than a SATA 3 Gbit/s mechanical drives, because the interface between the drive and the host controller is wider with SATA 6 Gbit/s. The answer is no. The reason is that mechanical drives can hardly even saturate SATA 3 Gbit/s, which means that having a 6 Gbit/s interface will make no practical difference. If we take 3 Gbit/s, that’s 3,000 Mbps, add in 8b/10b encodingRef: 9 overhead, which results in 2,400 Mbps of available bandwidth, which equates 300 MB/sec! While there is a hard drive on THG’s chartsRef: 10 that almost breaks 210 MB/sec in a controlled lab environment, but there is no mechanical drive that can even come close to fully saturating 300 MB/sec.
  • SATA 6 Gbit/s is best used with a high performance drive, which is mainly the SSD,Ref: 11 where this technology can take advantage of SATA 6 Gbit/s. I would say that even if utilizing this interface, within the DiskStation, in terms of performance, the bottleneck would become the network port of the DiskStation, which is commonly 1GbE.
  • As SATA 6 Gbit/s is backwards compatible with SATA 3 Gbits/s, a hard drive with SATA 6 Gbit/s interface installed within a SATA 3 Gbit/s controller will simply operate at SATA 3 Gbit/s capabilities.
  • Essentially, with today’s SATA hard drives, and given the application of general-purpose data storage, the interface and hard drive is fast enough for today’s applications. Attempting to make the pre-requisite that everything must be SATA 6 Gbit/s in computer hardware, even though no mechanical device can appreciate that bandwidth, or there’s often other limitations in the computing device; is in fact a waste of money and resources.


Do I need NCQ capable hard drives?

  • TCQ or Tag Command Queuing,Ref: 12 allows an OS to send multiple requests to a hard drive and the hard drive can sort out the requests and get them processed. Before TCQ – the OS itself had to arrange the requests and commit them one at a time to the hard drive, hindering the OS’ productivity. SATA’s NCQ technology is much more efficient for handling bulk commands to be processed within a hard drive.
  • NCQ or Native Command Queuing,Ref: 13 allows for bulk commands within a HDD to be internally optimized for the most efficient read/write operation of those commands, in an effort to minimize seek time or rotational latency for a platter to be in position to receive those commands. NCQ is mainly beneficial in environments where there are numerous transactions occurring concurrently, such as Virtual Storage, or Database Environments. Most drives today that I’ve seen are NCQ capable and the DiskStation can take advantage of this technology.


Are SSDs the better choice over hard drives?

  • This was asked to me in a recent discussion I had with resellers – and in my opinion, SSDs are not a superior option. When compared to $100 USD per Terabyte, SSDs are expensive when it costs about $1000 USD per Terabyte. SSDs can still share same firmware concerns as their mechanical brethren.
  • In my view, in the Enterprise market – SSDs are better suited towards data caching or high-speed data access. I wouldn’t recommend using SSDs for handling constant heavy IOPS, given the concern that NAND FlashRef: 34 can only be written a certain amount of time before it fails. In contrast – the hard drive doesn’t have specific limits of how many times data can be written to a specific sector, however, it can still suffer mechanical problems that can cause data loss. Typically, with SSDs, firmware errors are the common problem that can cause data loss. Both technologies must be used properly in a storage ecosystem to tap the best potential out of both technologies.


What’s the difference between Enterprise-class and Desktop-class Hard Drives?

Another frequently asked question is “What is the difference between the DCD and ECD, and where would ECDs would be most beneficial?” From reading the various datasheets out there, I have surmised that the following factors are most importance between the two different classes. These factors will show where ECDs are superior and worth considering for usage.


Spindle Shaft Design difference between ECD and DCD

  • On ECDs, typically, the Spindle Shaft is mounted at the top and bottom of the drive, to stabilize the entire spindle when spinning at high RPMs. This method of securing the Spindle Shaft can be found on Western Digital’s StableTracRef: 14, 32 or Seagate’s TCA.Ref: 15 With the Spindle Shaft mounted in this fashion, the platters themselves remain flat when viewed laterally. This in turns, keeps the platter “relatively” in place when viewed from the read/write heads which allows for a constant and consistent flow of information. When compared to their DCD counterparts, because the Spindle Shaft is secured at both ends, the hard drive itself doesn’t have to waste as much time to recalculate the relative position of the platters as the drive spins or re-request the data. Hard drives without this technology may wobble much like the children’s toy topRef: 16, which typically results in performance degradation.


Reliability Design between ECD and DCD

  • Power-On-Hours (POH)
    • Typically with ECDs, these drives are designed and rated to operate 24 hours a day, for every single day in the year,Ref: 7,14,17 which equates 8760 hours out of each year. ECDs are often manufactured with more advanced techniques or superior internals, which allow them to operate for greater lengths of time.
    • With DCDs, they are not intended to operate for that long – Seagate documented the POH on one of their DCDs to be 2400 hours a year.Ref: 18 I would believe that other drives will have similar tolerances – with the exception of Hitachi’s 7K4000 Deskstar drive, given that it has the same POHRef: 19 life as their Ultrastar.
    • Exceeding this limit will impose greater wear on the drive, and can result in premature drive failure.

  • Annualized Failure Rate (AFR) / Mean Time Between Failures (MTBF)
    • AFRRef: 20 and MTBFRef: 21 are statistics which are related to each other – more so with AFR being related to MTBF. These numbers are used by manufactures to state the reliability of the hard drive, often the lower the AFR, the lower chance of failure. Typically, ECDs feature a lower AFR than DCDs. Note that there is no manufacturer of any hard drive that can absolutely guarantee zero down time. I would like to cite Google’s hard drive studyRef 22 conducted in 2007.
      • Within their study, they found in general, that the AFR for drives within their first year is 1.7%, and well over 8.6% for drives older than three years. Its meaning reminds me of an old adage: “There are two types of hard drive owners, those who have experienced a drive failure, and those who will experience a drive failure.” I am a member of the former group.


  • Unrecoverable Bit Errors (UBE)

    • UBE Rates how many errors will be encountered per read or write operation on the drive, depending on how the manufacture rates it. It can be interpreted as a measurement intended design of how reliable a drive platter is. Typically with DCDs, the UBE rating is typically 1 x 10E14, while ECDs have a UBE rating of 1 x 10E15.
    • What this means is that ECDs are ten times less likely to encounter a read or write error, given their greater standards in manufacturing technology.

  • ECDs are manufactured with superior internal components and process when compared to DCDs, so they are less likely to have mechanical failure. However, this does not make ECDs invulnerable to failure.


Sector recovery behavior difference between ECD and DCD

  • Often with ECDs, manufacturers will advocate another important technology for RAID environments: that these hard drives can recover from defective sectorsRef: 23 faster – thus, maintaining communication and availability to the host bus adapter (HBA).Ref: 33 The most public of these technologies is Western Digital’s Time-Limited Error Recovery (TLER).Ref 24 There’s also Seagate’s Error Recovery Control (ERC),Ref: 25 Samsung and Hitachi use Command Completion Time Limit (CCTL). All of these technologies achieve essentially the same thing, which is to limit how long the hard drive attempts to remap a defective sector, when it encounters one.
  • If a DCD in a RAID environment encounters a defective sector, then the drive will spend some time trying to repair that sector, more time than an ECD. During this process, the hard drive can become unresponsive to the HBA, which results in performance degradation. Eventually, the drive will be ejected from the array, as it has stopped communicating for too long, and the HBA believes the hard drive is defective. This will unnecessarily require that the drive be reintegrated into the array which results in the process of rebuilding/resynchronizing the volume to ensure the data is consistent.
  • Using an ECD drive will avoid this problem of the drive being ejected from the array due to defective sectors, and lost data from a defective sector can be rebuilt using the redundant mirrors or parities of data on the other drives. Using ECDs will maintain high-accessibility of the volume in large multi-drive Enterprise environments by reducing the interruptions caused by defective sectors.
  • Of note, since DSM 2.2Ref: 26, released in September 2009, Synology introduced their own sector recovery subroutine operating in the background, called Dynamic Bad Sector Recovery, to “further enhance the system reliability” with hard drives. This function basically operates in conjunction with DCDs or ECDs, to help maintain availability of a volume when encountering defective sectors on hard drives.


Warranty differences between ECD and DCD

  • Most ECDs, given their higher grade of internals, typically come with a 5 year warrantyRef: 7, 14 from the manufacturer, whereas DCDs can come with 2 or 3 yearRef: 4 warranty.


Which hard drive brand is better?

  • With regards to which brand is better, in my view, there is no absolute best brand. Based on my experience of using hard drives since 1991 – I’ve experienced hard drive failures from Fujitsu, Hitachi GST, IBM, Maxtor, Quantum, Samsung, Seagate, Toshiba, and Western Digital. In my opinion, as a hard drive is man-made mechanical device, there are bound to be failures. The fact is hard drives are not an infallible piece of technology; the best way to avoid headaches with hard drives is to maintain a good backup strategy.
  • I’ve experienced hard drive failure within two minutes of opening a fresh drive (that came in standard 20-drive pack, no less), where the drive resulted in the Click of DeathRef: 27 – and then I have drives that are over 10 years old and are still operating just fine today within my AV and computer equipment at my house.
  • What this means is that there really is no best drive manufacturer – it’s not a matter of if a drive will fail, it’s a matter of when a drive will fail.


Tips for selecting a hard drive for use with the Synology DiskStation NAS

Which drive to purchase for the Synology DiskStation NAS?

  • Assuming that we’re not talking about capacity requirements, it’s always best to use a drive that Synology has listed as compatible, as referenced on the hard drive compatibility list.Ref 29 Drives listed here have been acquired by Synology and gone through Synology’s testing procedure, which tests for over 70 different parameters – some of which include
    • How does the drive behave when it is pulled out of the array?
    • How does the drive behave with SMART?
    • How does the drive handle defective sectors?
    • How does the drive behave when at various ambient temperatures?
    • How does the drive handle multiple numerous reboots?
    • …and other parameters.


Applications for ECDs is where reliability and availability are preferred over affordability

  • One application for ECDs is serving the needs of many users at once – where data inaccessibility due to drive failure will affect many users at once. Where many users’ inaccessibility of data stored on a hard drive can result in thousands of dollars lost per second due to loss productivity, sales and or both.
  • ECDs are also intended for handling consistently high IO/s concurrently – such as hosting storage for a database server, or Virtual Storage environment
  • ECDs are also used where data needs to be accessible at any given time of the day, 24 hours out of the day, throughout the entire year.


Applications for DCDs is where affordability is preferred

  • DCD applications include serving the needs of a small number of users (or even a single user) – who occasionally need to access the drive. Where this type environment can tolerate down time of a few days as they are not fully dependent on that drive for storage.
  • Where the drive will spend most of the time in drive hibernation
  • As a backup storage
    • As an example, I have DCDs at my house for my backup DiskStation. This particular system is only active two hours a day, just to run the backups. The rest of the time, it is in drive hibernation. This is a perfect example of where to apply DCDs.


Should I use the latest and greatest sized hard drive or use a drive that has been around on the market?

  • Personally, I don’t use the latest sized drive – as I consider it to be bleeding edge technology.Ref: 29 Usually with any bleeding edge technology, whether it’s a drive, program, computer software, or even a smart phone – there may be some issues that cannot be tested or found within a controlled lab environment. Because of this unknown factor, I usually lean towards older drive technology. I value reliability and affordability within reason. With that being said, as today’s largest size drive is 4TB, I would be more comfortable using a 3TB or even 2TB drive, as 2TB technology has been on the market for a couple of years.
  • Another advantage to using older drives is that they are typically more affordable than newer capacities. When more storage is needed – assuming that the DiskStation has a redundant RAID volume, the storage within the DiskStation can be expanded by adding more drives, or swapping out older capacity drives with larger ones.


Using Volume Complete Consistency Check

  • When building a volume with in the DiskStation, please always use complete consistency check. While complete consistency check does require more time up front to build the volume, the disks are being stressed by the DSM to ensure that they do not have defective sectors. I would rather take more time up front to setup my volume correctly in the beginning rather than to deal with no-availability or volume errors down the road.


Use a UPS

  • When using a DiskStation, I always a recommended the use of a UPSRef: 30 to reduce the chance of data loss. As today’s hard drives typically come with 64MB of cache, using say a DS1512+, that would be 320MB of data using a 5-drive configuration. If the DS1512+ were to experience an unsafe shutdown, that would mean 320MB of cached data would be lost from the hard drives. Along with 320MB of data loss from the drives, 1GB of data will be lost from the DiskStation’s RAM. That much data lost can result in data corruption on the DiskStation. A UPS is rather affordable, some that reach as low $50 USD. It’s a small investment to avoid wasting time with handling restoration from backups, or paying several thousands of dollars per hour for data recovery services.


Summary
Given all that has been discussed, I hope that I’ve answered all of the common questions of what is there to know about hard drive technology. In general, when I choose a drive, I usually refer to the Synology HDD Compatibility list, identify the capacity that I need, and select a drive that meets my budget and those capabilities I want it to have. Again, by being conservative with drive selection, I would use a drive size or model that has been on the market for a while, as the market itself has proven that drive reliable. Even after I chose the drives for my project, I shouldn’t become complacent about making backups.Ref 31


References
1. Wikipedia: Hard Disk Drive
2. Synology Blog: Synology XS Series with 100,000 IOPS and 1000MB/Sec
3. Western Digital Caviar Blue Datasheet
4. Western Digital Caviar Green Datasheet
5. United States Energy Information Administration
6. Hitachi GST Ultrastar 5K3000 Datasheet
7. Hitachi GST Ultrastar 7K3000 Datasheet
8. Wikipedia: Serial ATA
9. Wikipedia: 8b/10b Encoding
10. Tom’s Hardware Guide: Charts, benchmarks HDD Charts 2012…
11. Wikipedia: Solid-State Drive
12. Wikipedia: Tagged Command Queuing
13. Wikipedia: Native Command Queuing
14. Western Digital RE4 Datasheet
15. Seagate Constellation ES.2 Product Overview
16. Wikipedia: Top
17. Seagate Constellation ES.2 Datasheet
18. Seagate Barracuda Datasheet
19. Hitachi GST Deskstar 7K4000 Datasheet
20. Wikipedia: Annualized Failure Rate
21. Wikipedia: Mean Time Between Failures
22. Failure Trends in a Large Disk Population
23. Wikipedia: Bad Sector
24. Western Digital: Difference between Desktop edition and RAID (Enterprise) edition drives
25. Seagate: What is Error Recovery Control?
26. Synology Inc: DiskStation DS508 Release Notes
27. Wikipedia: Click of Death
28. Synology Inc: What hard drives does Synology Product support?
29. Wikipedia: Bleeding Edge Technology
30. Wikipedia: Uninterruptible power supply
31. Synology Wiki: What is a Backup?
32. Western Digital: StableTrac
33. Wikipedia: Host adapter
34. Wikipedia: Flash memory

Trademarks
Throughout this article, trademarked names and or descriptions are used. Rather than put a trademark symbol in every occurrence of a trademarked name, I state I am using the names only in an editorial fashion, and to the benefit of the trademark owner with no intention of infringement of the trademark.