Simple HDD Question .... I Think :(

[AstroDav]

This is sort of a continuation of my 1st question concerning a new build which lives over here:

http://forums.anandtech.com/showthread.php?t=2300883

As in that question, I have narrowed my hardware selections down to a few & there I'm stuck. And as in that question, I don't mind at all in being guided to a completely different model .... as long as it's in the same general $$ amount.

The HD's I've settled on are these:

1. WD Black WD2500BEKT
2. WD Black WD3200BEKT
3. WD Scorpio Black WD5000BPKT
4. WD Green WD5000AZRX

The first 3 appear identical, save for capacity. I'm not sure if #1 & #2 are considered "Scorpios" or not, & if not then what difference does that mean? #4 is obviously very different. So here are the points which concern me.

1. The Blacks are 2.5" mobile drives. Do these perform equally well in a desktop as 3.5s?
2. The Green is SATA-III, as compared to SATA-IIs. Is the difference noticeable in general computing?
3. The Green has 64MB cache, as compared to 16MB. Seems significant, but again, will it be noticeable in general computing?
4. I noticed the RPM on the Green isn't listed like the 7200 on the Blacks, but rather is "IntelliPower", which seems to mean it's variable. Anyone know if this works well?

Capacity isn't a big concern. There will be 2 of these, so the least I would have is 500G, which I can work with.
My OS & most-used programs will be on SSD, so these drives will be storage basically. Therefore they don't have to be fast as lightning .... but I won't gripe if they are.
Heat, noise, physical size, etc. is also not extremely important. I use big cases & know how to cool them.

My previous CPU thread was a great help from here, so much so that I didn't even bother posting THIS one any other place. The same type replies are what I'm after. What do you think of my choices? Which one is best? Is there a BETTER choice in same general $$ amount which escaped me?

Have at it. Your words are well received & heeded. A couple more similar threads will follow in near future, probably MOBO, SSD, & possibly memory.

 

[AstroDav]

P.S. Won't be RAID'ing.

P.P.S. Just say you have the same identical drive, save different capacity, as in #1 & #2. Will there be any REAL difference in search/read/write speed? 1st thought says the smaller drive will be slightly faster, but I also feel in modern drives that any difference will be VERY slight .... or maybe not? Also, there are rumours that larger drives cr@p out more often. Any real truth to that?

 

[smitbret]

Greens rotate at 5400rpm, give or take. But they also use less power than your other choices. They are plenty fast for my needs at over 100MB/s. The Black is better for NAS and other services that are more demanding than normal home service. Green probably has the bigger cache because it's a desktop and not a mobile drive. Scorpio means it's a mobile HDD.

I don't see where you would gain anything from going Black. If you were worried about warranty, you should just go Red and save the $$$.

http://www.wdc.com/en/products/internal/desktop/

 

[Cerb]

1. The Blacks are 2.5" mobile drives. Do these perform equally well in a desktop as 3.5s?

No, but you'd be hard-pressed to tell the difference, without running benchmarks.

2. The Green is SATA-III, as compared to SATA-IIs. Is the difference noticeable in general computing?

SATA 3Gbs v. SATA 6Gbps? No. WD Black v. WD Green? Yes.

3. The Green has 64MB cache, as compared to 16MB. Seems significant, but again, will it be noticeable in general computing?

Maybe?

4. I noticed the RPM on the Green isn't listed like the 7200 on the Blacks, but rather is "IntelliPower", which seems to mean it's variable. Anyone know if this works well?

It's marketing. They're 5400 RPM drives.

Capacity isn't a big concern. There will be 2 of these, so the least I would have is 500G, which I can work with.

Why not just get a single 3.5" 500GB or 1TB? A WD Blue 500GB is $60fs, and a 1TB $75fs.

 

[Phynaz]

The Black is better for NAS and other services that are more demanding than normal home service.

A bit off topic, but you would never put a Black in a NAS, it's strictly a desktop drive.*

*Assuming you value your data.

 

[Phynaz]

Why not just get a single 3.5" 500GB or 1TB? A WD Blue 500GB is $60fs, and a 1TB $75fs.

This. Get the 1TB single platter Blue.

 

[AstroDav]

I always use twin drives with identical data as a form of safety, although I use a back-up utility rather than a RAID array.

While I was true when saying I welcomed alternate suggestions, Blues weren't included because most everything I've read hinted that the Blacks, Greens, & Reds were a higher performance drive, perhaps even a bit higher quality. Am I wrong? I may be. The Reds seem to be for RAIDs & servers.

Also there are reports that smaller drives tend to last better than huge ones, another reason there are no 1Ts+ in the list. I'm not saying that's actually correct, but I like giving my hardware every chance I can find to last as long as possible.

 

[smitbret]

A bit off topic, but you would never put a Black in a NAS, it's strictly a desktop drive.*

*Assuming you value your data.

For a home NAS? Why not? Sure it's overkill and the heat and noise might mean other drives are better suited to it, but what makes this particular drive bad for NAS?

 

[AstroDav]

I'm actually the asker here instead of the answerer, but I'm gonna take a stab at this one. In narrowing down these 4 choices of drives, I probably read a couple hundred or more descriptions, reviews, & site-specs on maybe 50 specific brands/models of drives. As is obvious from my list of 4, much of that time was spent with Western Digital.

I studied their Blues, Blacks, Greens, Reds, Enterprise, & a couple other types. From what I saw, which isn't "wrong" but could be only "half-right", all catagories except the Reds, Enterprise, & possibly the "AV" (Audio-Visual) makes are not designed for the 24/7 constant usage which is common in servers, NAS, etc. The first 3 makes above need to rest every now & then. If not, they'll sooner-or-later spin themselves to death.

At least that was my take on it, as i was at one time considering 2 Reds for a RAID array. But I decided not to go RAID, so I didn't need drives specially built for that environment.

But again, I may be completely wrong on that.

 

[Phynaz]

For a home NAS? Why not? Sure it's overkill and the heat and noise might mean other drives are better suited to it, but what makes this particular drive bad for NAS?

With the assumption that any NAS is going to have some kind of raid configuration. Now that desktop drive hits some kind of error...a bad read, a seek error, whatever. What does it do? It starts retrying and going through its error recover procedures. While it's doing that it isn't responding to the RAID controller.

So the RAID controller doesn't see a response and drops the drive - buh-bye array.

You don't use desktop drives in storage servers - period.

 

[smitbret]

So Blacks, like WD Greens, don't have TLER enabled?

 

[bononos]

So Blacks, like WD Greens, don't have TLER enabled?

Black/blue/green don't have/support tler. Red supports tler.

 

[Cerb]

I always use twin drives with identical data as a form of safety, although I use a back-up utility rather than a RAID array.

That's a fair way to keep redundant data. RAID does not provide the data redundancy that a backup does, but only protection against a detected disk drive error, or complete failure of a drive. Backups protect your data, while RAID decreases downtime. For most individuals and very small businesses, backups alone are generally the better way to go, unless or until you reach a point of wanting a file server or NAS box.

While I was true when saying I welcomed alternate suggestions, Blues weren't included because most everything I've read hinted that the Blacks, Greens, & Reds were a higher performance drive, perhaps even a bit higher quality. Am I wrong? I may be. The Reds seem to be for RAIDs & servers.

Yes.

AVs are made for constant writing, low power, will perform wear-leveling of sectors, and won't try to re-read to recover hard to read sectors. They're made mainly for surveillance DVRs and the like.

Greens are low-power, quiet, and slow. They are also cheaper per GB.

Blacks are WD's fast drives. They are louder, use more power, and offer higher performance.

Blues are like detuned Blacks. Less power, less noise, a bit less performance, but not by much.

REs are basically Blacks or Greens with firmwares tweaked for RAID operation. RE3 and newer get better performance than desktop drives in RAID 5, 6, and 10, and they have a few features so as to not drop out of arrays (you used to be able to change these settings in many desktop drives, prior to the RE series existing). Oh, and a 5-year warranty comes with them.

Reds are a file server centric mishmash, with the AV's wear-leveling, the Green's low spin speed, and the RE's TLER and such.

Also there are reports that smaller drives tend to last better than huge ones, another reason there are no 1Ts+ in the list. I'm not saying that's actually correct, but I like giving my hardware every chance I can find to last as long as possible.

True, but now 1TBs are going to be among those smaller drives. Smaller drives tend to last due to using fewer platters, and generally using older technology, that has had most of its kinks worked out. The top end of capacity is pretty much always the least reliable. That said, I'm not sure WD or Seagate has really gotten it all back together since the flood.

 

[sub.mesa]

AV wear levelling? 5400rpm is slow?

I would also argue that a backup is not redundancy. Redundancy is strengthening the error correction capabilities of a single volume containing a single version of data. A backup is an independent copy of your data and by definition different from redundancy. That is why ZFS is more than just RAID; it also provides history in the form of snapshots that can determine the contents of your data in a time in the past. This is something regular RAID or redundancy can not provide.

 

[Cerb]

AV wear levelling?

Indeed: http://wdc.com/en/products/products.aspx?id=150
It is one of the few features that differentiates them from other drives. It's what makes them reliable, with basically no re-reading for error recovery (the latter feature, without wear-leveling, would be recipe for data loss).

5400rpm is slow?

Yes. Most drives are 7200RPM, and there is a significant performance difference between such drives. 5400RPM desktop drives could probably perform faster than they do, but all of them are 5400RPM in part because that is one part of enabling better power efficiency, so it is one of quite a few tweaks that end up reducing performance.

I would also argue that a backup is not redundancy.

It is, though. It's not the end-all be-all, but it is a level of redundancy.

Redundancy is strengthening the error correction capabilities of a single volume containing a single version of data. A backup is an independent copy of your data and by definition different from redundancy.

Redundancy is by definition duplication of resources, or excess resources, beyond what is minimally necessary. Backups are, by definition, a form of redundancy.

That is why ZFS is more than just RAID; it also provides history in the form of snapshots that can determine the contents of your data in a time in the past. This is something regular RAID or redundancy can not provide.

ZFS does not improve error-correction capabilities, though, and with a ZFS RAID, you still want to have a backup. It provides better error-checking, but then leaves you to handle the correction. ZFS, and competition for it, such as BTRFS, should be more prolific, and Windows needs something similar, but it's biggest advantage is giving you warnings that you might not get with other FSes, even with a scrubbing RAID (between the CERN RAID 5 v. 10 tests, and cases of ZFS helping to find firmware and hardware problems, I like additional CRCs, these days!).

Backups, however, can provide exactly that kind of versioning, including added CRCs, depending on software used, without needing a separate computer running FreeBSD. For example, the free Cobian can provide that, though you don't exactly get a nice interface into your old data, like Acronis can offer you.

 

[sub.mesa]

AV drives indeed employ zero recovery time, also known as TLER=0. This is to prevent loss of frames due to the harddrive not being able to keep up with the datarate due to random seeks. Any corruption would show up as a bleep or video artifact, but the harddrive should maintain the datarate at all costs. These drives are not very suitable for generic use, except for special circumstances like video capturing.

But I wouldn't call this wear levelling. What is being levelled? What is the 'wear'? It's not an SSD.

Furthermore you seem to think that 7200rpm drives really are much faster than 5400rpm. There are plenty of examples to show otherwise, meaning that 5400rpm drives sometimes even outperform 7200rpm drives on sequential workloads due to higher data densities (higher capacity platters).

Examples of these are in the previous Samsung F4EG generation which could compete very well with the 500GB platter WD Black at the time, which runs at 7200rpm instead of the 5400rpm for the Samsung. Due to the Samsung employing higher capacity 666GB platters, which meant they got up to 140MB/s sequential speed (outer tracks/start of capacity).

Today, the lower rpm disks no longer have an advantage in terms of higher data densities, since 1000GB platters are the maximum densities seen in the wild thus far, with both 5400rpm and 7200rpm models employing these. However, even in this level playing field the 7200rpm doesn't add that much speed at all; the potential 33% faster sequential performance often isn't utilised effectively. The actual gain is more like 15% faster sequential performance.

Comparing 666GB platters, if 5400rpm does 140MB/s, then 7200rpm should do up to 186MB/s.
Comparing 1000GB platters, if 5400rpm does 158MB/s, then 7200rpm should do up to 210MB/s.

The faster random I/O operations is something only rpm can increase. However, since SSDs simply blow the roof of, any solution demanding IOps (like the system drive) will be handled by a SSD in most cases. So harddrives pretty much are still useful for storage of large amounts of data in sequential workloads.

For this kind of task - mass storage - I would argue that other properties are probably more important: power consumption which equals heat generation but also vibrations and resonances that can be really annoying and dangerous as well to mechanical components.

Redundancy is by definition duplication of resources

A duplicate ('backup') will behave as a separate entity, independent from the original. Redundancy instead, is always tied to the original and always changes together with the original. It is part of the original, and therefore not a duplicate.

ZFS does not improve error-correction capabilities, though, and with a ZFS RAID, you still want to have a backup. It provides better error-checking, but then leaves you to handle the correction.

What would the user have to do? ZFS has no filesystem check. It checks itself. It handles all corruption itself. It is a self-healing filesystem. Any damage it encounters is fixed on the spot. If the damage exceeds the error correction mechanisms that ZFS employs, the file in question is under quarantine and may not be accessed by applications. This happens if you encounter a bad sector on a RAID0 configuration for example.

If the bad sector affects the all-important filesystem metadata, you would be screwed if you used a legacy filesystem. In ZFS, metadata is always replicated even in single disk and RAID0 configurations. Bad sectors cannot damage ZFS itself.

Backups stored on legacy filesystems are still vulnerable to silent corruption. ZFS allows utilising send and receive, causing all checksums to be transferred as well when transferring data from one system to the other. Employing both backups and ZFS with snapshots will grant you a formidable protection against a wide range of dangers; something legacy filesystem cannot provide.

 

[Cerb]

But I wouldn't call this wear levelling. What is being levelled? What is the 'wear'? It's not an SSD.

The writing surface loses strength over time, from reads, or just by bad luck from the factory. In a typical desktop drive, this would cause some pause, and possibly a re-allocation. The medium itself should not wear out, but the bits stored there very much can.

Furthermore you seem to think that 7200rpm drives really are much faster than 5400rpm. There are plenty of examples to show otherwise, meaning that 5400rpm drives sometimes even outperform 7200rpm drives on sequential workloads due to higher data densities (higher capacity platters).

Examples of these are in the previous Samsung F4EG generation which could compete very well with the 500GB platter WD Black at the time, which runs at 7200rpm instead of the 5400rpm for the Samsung. Due to the Samsung employing higher capacity 666GB platters, which meant they got up to 140MB/s sequential speed (outer tracks/start of capacity).

http://www.storagereview.com/samsung_spinpoint_f4eg_review_hd204ui
http://www.storagereview.com/western_digital_caviar_blue_1tb_review_wd10ealx

That's not even the Black, and it's not as fast, either in sequential or random.

For this kind of task - mass storage - I would argue that other properties are probably more important: power consumption which equals heat generation but also vibrations and resonances that can be really annoying and dangerous as well to mechanical components.

Or, you can just leave the OS' power settings alone, and save power by it going to sleep. With a desktop duty cycle, it won't make that much difference, either way, unless you force the HDD to stay on all the time. Just using the latest generation of hardware will save the most electricity use.

A duplicate ('backup') will behave as a separate entity, independent from the original. Redundancy instead, is always tied to the original and always changes together with the original. It is part of the original, and therefore not a duplicate.

http://www.merriam-webster.com/dictionary/redundant
"3: serving as a duplicate for preventing failure of an entire system (as a spacecraft) upon failure of a single component"

There is not a distinct definition of redundancy used for IT. We use that one, which even includes the keyword, "duplicate." RAID is a form of redundancy. Parity is form of redundancy. Backup is a form of redundancy. It is a term with a wide scope, and not one for splitting hairs. Any feature which provides useful excess, to prevent or recover from, a malfunction, is a an implementation of a feature that adds redundancy. Since cheap and plentiful consistently wins out over expensive and reliable, the more redundancy that can be added, the better.

What would the user have to do?

Use a dedicated file server of some ilk, paying handsomely for an easy one to set up; instead of using software on the computer they already have, to do it simpler and easier. It has its place, but a home user that has no *n*x or general sysadmin ambitions probably isn't one of them.

Regardless of FS, the user would need to locate a good older copy of their file(s)--IE, make use of a backup.

ZFS has no filesystem check. It checks itself. It handles all corruption itself.

If by handling, you mean, gives an error, then yes. But, your data is still gone, so that's a very disingenuous statement.

It is a self-healing filesystem. Any damage it encounters is fixed on the spot. If the damage exceeds the error correction mechanisms that ZFS employs, the file in question is under quarantine and may not be accessed by applications.

But now your data is toast. The file system is good, though, hooray! You will have still needed backups, and still will have needed to verify the backups before the data loss. ZFS employs no error-correction mechanisms, at the current time. If a data-located error occurs, you will be told it happened with more certainty than other FSes, but ZFS will not, and currently can not, do anything to fix it. You will need RAID for online automatic attempts at recovery, and backups for any further attempts. ZFS will not alleviate that need in any way (if doing RAID, however, it can provide more robust RAID than standard FS-independent RAID).

I'm not alone in wishing it had some parity for single-drive configurations, so I wouldn't be surprised if either it, or BTRFS, got some, in the future. But, today, there aren't any. Being able to reserve n/2^M blocks for simple parity would be fantastic, IMO, for occasional small errors. Basic XOR or R-S stuff would be just fine, to me. For now, any such implementation needs to be done at a file level, with 3rd-party software.

This happens if you encounter a bad sector on a RAID0 configuration for example.

That happens in any configuration of any *n*x or Windows, today, regardless of FS or disk configuration.

If the bad sector affects the all-important filesystem metadata, you would be screwed if you used a legacy filesystem. In ZFS, metadata is always replicated even in single disk and RAID0 configurations. Bad sectors cannot damage ZFS itself.

If the bad sector affects your all-important research notes, you will be just as screwed if you use ZFS. It does in no way alleviate the need for RAID and/or backups.

Backups stored on legacy filesystems are still vulnerable to silent corruption.

Backups stored with checksum or parity files, or in archives, have that taken care of. Several commercial backup programs will do that, along with versioning, and often provide a UI that non-techies can manage to utilize, to get to that older data.

ZFS allows utilising send and receive, causing all checksums to be transferred as well when transferring data from one system to the other. Employing both backups and ZFS with snapshots will grant you a formidable protection against a wide range of dangers; something legacy filesystem cannot provide.

You have two words wrong: "protection against," should be, "detection for." RAID and/or backups and/or parity are what can provide protection. ZFS alone provides enhanced detection only.

If the OP is backing things up, and occasionally making sure the backups can be verified internally, and restored to usable files, (s)he is in pretty good shape. Off-site (what if your house burns down, or your computer stolen?) and hard-copy (such as quality DVD or BD recordable media) backup would be the next logical step, not a file server with ZFS.

If you're already going to get a nice NAS box, or go geek out and build a file server, then get one running FreeBSD under the hood, with ZFS RAID support, and do it right. But, that's one Hell of a jump from an extra drive or two, a massive learning curve for most users, and it will provide little or nothing that a run of the mill backup program cant do, if you don't have other uses for a file server.