Discussion of HDD selection philosophies and strategies

In another thread, Purko and I have been discussing (debating?) HDD selection philosophies and strategies.  In order to keep these forums a bit more organized and to put and end to our hijacking of that thread, I figured I would move the discussion to this dedicated thread.  Hopefully others will find our discussion useful or at least interesting (that's assuming that my position has at least a modicum of merit, and I'm not just blowing smoke ). 

 

Purko: I edited our discourse only slightly to make it more readable to someone just jumping in on it.  I don't feel I have quoted you out of context anywhere, but if you feel differently, please call me on it.

 

Here's what has been said thus far:

 

Whenever I buy a new hard drive for my unRAID server, my primary rubric is the bang-for-your-buck ratio, which is essentially how many GBs of space are you getting for every dollar you spend.  Consider this:

 

HDD

Price + Shipping

GB/$ (a.k.a. bang-for-buck ratio)

Western Digital 1TB$97.2710.28Western Digital Caviar Green 1.5TB$104.9914.29Western Digital Caviar Green 2TB$179.9911.11

 

The 1.5 TB drive is the clear winner.  Keep in mind that this is a moving target, a 'sweet spot' in hard drive prices.  Do these calculations everytime you need a new drive, then let that guide you to the new one.  Take this with a grain of salt, since I only compared Western Digital drives here.  I'm a big fan of WD's quality and I generally prefer to buy their drives, however, if some other brand goes on sale then I will deviate.  Choose by the prices and the quality, since unRAID doesn't care much about the drive's speed.

 

That calculation is not very correct, because it does not take into consideration these two things:

-- The cost of the server, together with the limitation of how many disk slots it can have in it;

-- The cost of the unRAID license.

When you take these into consideration, then the 2TB disks start to seem much more attractive.

 

You are correct, my method does break down a bit when used for an initial build for the reasons you mentioned.  It is primarily intended to guide a purchase when upgrading an already built server.  For example, say my 8 disk server is filled to its max capacity, but I'm out of space.  My obvious course of action is to start replacing my smallest drives, currently 500 GBs, with new large drives.  In this scenario, which new drive should I choose?  I would use the above method to determine that the 1.5 TB drive is the right choice.

 

Still, if you feel like going through the trouble, I would be interested to see an algorithm that accounts for all the variables you mentioned.  I'm considering building a new server soon myself, so I would probably even put it to use.

 

There is no practical distinction between building and upgrading. My reasoning still holds.

You'll normally build a 'new' server with just a couple of disks to start, then 'upgrade' by adding more disks, so there's no difference.

Either way you face the question about how many TeraBytes you can stick in that box.

 

Still, if you feel like going through the trouble, I would be interested to see an algorithm that accounts for all the variables you mentioned.

 

There's no special algorithm. You can figure out the total usable size from data disks,

then divide that by the cost of the server plus cost of all disks including parity plus cost of the license.

Only this way you can have a somewhat realistic estimate for gb/dollar.

 

And, mind you, I haven't even mentioned the cost of electricity so far.

For example, I have more than fifty ancient 200GB disks lying around on the shelves in the office...

I could easily build a 20-disk unRAID server with some of those.

Or, I could achieve the same result with just two 2TB modern green drives.

Think about the difference in electricity costs!  (FYI, I also put a Ultra-Low-Voltage CPU in my latest server)

 

Bottom line, your 'formula' doesn't really give you the real picture about the "most bang for your buck".

 

There is no practical distinction between building and upgrading. My reasoning still holds.

You'll normally build a 'new' server with just a couple of disks to start, then 'upgrade' by adding more disks, so there's no difference.

Either way you face the question about how many TeraBytes you can stick in that box.

 

Perhaps we just see this issue from different perspectives, but in my view there is a large practical difference between building and upgrading.  When initially building a server, you method is obviously sound.  However, lets assume my server is at capacity with 8 drives.  These drives range from 500 GB to 1.5 TB.  I run out of space and want to start replacing the smallest drives, the 500 GBs.

 

Given the rate at which I fill up my drives (maybe 250 GB/month on average), I see no reason to upgrade my server until the hardware dies.  I should have no problem just replacing my smallest drives on a regular basis and keeping up with my data storage demands.  Therefore, I throw out the option of building a new server.

 

In this (I believe rather common) scenario, the server is already bought and paid for, and all SATA slots are created equal (negating any PCI bus bottlenecks).  I already have all the associate hardware, such as expansion cards and hot swap bays.  The only things left to change are the hard drives themselves.

 

In this scenario, I don't see any downside to using my simple calculation to guide you to a new replacement drive.  The 1.5 TB drives are currently in the sweet spot, and in a couple of months the 2 TB drives will most likely be there.  After that it will be the 4 TB, 6 TB, 8 TB, etc (assuming this half-step trend continues).

 

Granted, power consumption does come into play when dealing with older hardware, but in terms of modern hardware, I don't think its too much of a concern.  In most cases we would be talking about a difference of (at most) $5 per month.  It hardly seems worth it (to me, at least) to fret too much over that.  I buy green drives because they are the cheapest; the 'warm fuzzies' I get from having a low power server is just a fringe benefit.

 

Still, if you feel like going through the trouble, I would be interested to see an algorithm that accounts for all the variables you mentioned.

There's no special algorithm. You can figure out the total usable size from data disks,

then divide that by the cost of the server plus cost of all disks including parity plus cost of the license.

Only this way you can have a somewhat realistic estimate for gb/dollar.

 

This looks like an algorithm to me:

 

(total advertised size of disks in GB * 0.928)/(cost of server hardware not including disks + cost of unRAID license + cost of data disks + cost of parity disk + cost of cache disk, if used) = cost per drive slot in GB/$

 

Perhaps we just see this issue from different perspectives, but in my view there is a large practical difference between building and upgrading.

...In this (I believe rather common) scenario, the server is already bought and paid for...

 

If you regard these as different scenarios, you are only deceiving yourself.

 

Granted, power consumption does come into play when dealing with older hardware, but in terms of modern hardware, I don't think its too much of a concern.

 

Power consumption always comes into play.

 

Two of the modern green 1TB drives will still consume about twice the power of one 2TB disk.

 

Four 1.5TB drives will consume about 30% more power than three 2TB drives for the same usable space.

 

Imagine yourself for a moment being assigned the job of building a data center that will host 20,000 hard disks in just one room.

Will power not be a concern for you? 

 

Do you know that the cost of power can greatly surpass the cost of the hardware for its life time?

Perhaps we just see this issue from different perspectives, but in my view there is a large practical difference between building and upgrading.

...In this (I believe rather common) scenario, the server is already bought and paid for...

 

If you regard these as different scenarios, you are only deceiving yourself.

 

I'm really not following you here.  If the server hardware is already bought and paid for and one is not planning on upgrading it, doesn't it cease to be a variable?  I certainly understand using your method when initially building a server, but I don't understand the benefit gained in the scenario of upgrading a fixed-number-of-drives server.  I'm not trying to be belligerent, I really want to understand this.

 

Two of the modern green 1TB drives will still consume about twice the power of one 2TB disk.

 

Four 1.5TB drives will consume about 30% more power than three 2TB drives for the same usable space.

These are valid points, and some I haven't fully considered.  However, despite the percentages, the actual power consumption we are discussing is less than 10 W.  This equates to dollars per month at most.  Given that any drive in any of our servers at this moment will be obsolete within a few years (in terms of capacity vs power consumption), it hardly seems important to compare the power savings of, say, four 1.5 TB WD Green drives versus three 2 TB WD Green drives.  Let's say each drive uses an average of 5 W (that's a rough average of 10 W for spin up and 1 W for idle), the power savings of three drives versus four drives is only 5 W.  That's less than a dollar per month.  At current prices, four 1.5 TB WD Green drives would cost $439.96.  Three 2 TB WD Green drives would cost $509.97.  That's a difference of $70; it would therefore take 70+ months (over 5 years) for the power savings of the 2 TB drives over the 1.5 TB drives to make up for the extra cost of purchasing them today.  As I said before, 2 TB drives will be quite obsolete 5 years from now.  Hence, I don't see how power consumption is really an important factor in this comparison.

 

Imagine yourself for a moment being assigned the job of building a data center that will host 20,000 hard disks in just one room.

Will power not be a concern for you?

First of all, I would love to have that job  

 

In that scenario, yes of course power consumption would be a concern for me.  In that scenario the potential power savings of using 2 TB drives instead of 1 or 1.5 TB drives is immense, probably on the order of hundreds of dollars per month (just a guesstimate).  My point is that on the scale of the 'average' unRAID user (which from my observation seems to be about 10 drives), the differences in power consumption shrink to the point of being negligible.

 

Do you know that the cost of power can greatly surpass the cost of the hardware for its life time?

No, I can't say I did know that.  Intuitively, I want to believe you, but I would have to see the numbers to be truly convinced.  Still, its another interesting point you have brought up, and I thank you for that.  I'll go crunch some numbers now and see what they tell me.

hmm going by my power costs it would take 10 years for a 1tb drive to use its own price in electricity based on 7w 24x7x365 but the point is still valid once you compare 20 1tb drives to 10 2tb drives.

Here's some quick calculations:

 

According to the Western Digital's tech specs, the 2 TB WD Green and the 1.5 TB WD Green each use the same amount of power during read/write, which is 6 W.  Lets assume we are running torrents or something on our hypothetical server that keeps the drives reading and writing 24/7/365, thereby giving us the maximum amount of savings of using three 2 TB drives instead of four 1.5 TB drives.

 

According to this rather cheesy website, the US average electricity rate is 12¢/kWh.  Convert that to the more HDD-appropriate scale of watt-hours (Wh), and it becomes 0.012¢/Wh.

 

A single 6 W HDD running for one hour consumes 6 Wh.

 

1.5 TB drives

4 HDDs x 6 Wh = 24 Wh

 

24 Wh x 24 hrs/day x 365 days/year = 210,240 Wh/year

 

210,240 Wh/year x 0.012¢/Wh = 2522.88 ¢/year or $25.23/year

 

2 TB drives

3 HDDs x 6 Wh = 18 Wh

 

18 Wh x 24 hrs/day x 365 days/year = 157,680 Wh/year

 

157,680 Wh/year x 0.012¢/Wh = 1892.16 ¢/year or $18.92/year

 

That's a difference of $6.31/year.

 

Please check my math.

 

Granted, if we let the drives spin down and don't force them to spin 24/7/365, then those numbers will shrink.  So I guess your server's usage patterns may affect these calculations as well.  Even still, at a rate of saving a maximum of $6.31/year by using three 2 TB drives instead of four 1.5 TB drives, it would take over 11 years for the power savings to outweigh the difference in initial cost.  I think its obvious that both 2 TB and 1.5 TB drives will be quite obsolete 11 years from now; one would be quite lucky if a drive lasted that long in the first place.

I am new to unraid and in the planning stages, so this thread is very interesting to me.  I think I am going to start with a 2TB parity drive and 2 or 3 1.5TB data drives.  My thinking is with the current prices, 1.5TB is the best "bang for the buck" but I want to start out with a larger parity drive so down the road when 2TB becomes more affordable, I can add 2TB drives without changing my parity drive.

• 
   
  
• How hard is it to insert a new larger drive and make it the parity drive?  If this is an easy process, it might make more sense to use only 1.5TB drives now and when the time comes for larger drives, to worry about it then.
   
  
• Even though I won't be running my server 24x7, I never thought that hard about the cost of running the drives.  I currently have 8x250GB drives in a traditional Raid5 array and thought that I might initially re-use these in addition to 2x1.5TB drives when I move to unraid.  It probably makes more sense to retire these 8 drives and replace them with a 1.5TB drive from the start.  These are 4 or 5 year old drives, so they probably use even more than 6Wh each.
   
  
• This is probably only a question that I can answer as to what my time is worth to me, but what do others use as a guideline for excess capacity vs. the value of your time?  Let's say, theoretically that I need 2TB of space now and this will increase by 1TB/year.  Do I start with 2x1.5TB drives now, then in 2 years add more capacity a drive at a time or do I start with 3x1.5TB drives and avoid having to open up my server and add more drives for 1 or 2 additional years?  Benefits of adding excess capacity now are that I save time not having to add drives to my server later and cons are that drives will be larger and cheaper in 2 years than now and I will be wasting electricity powering up excess drives now when I don't need them.
  

 

Any guidance and opinions you can offer would be greatly appreciated.

 

thanks,

Murray

• How hard is it to insert a new larger drive and make it the parity drive?  If this is an easy process, it might make more sense to use only 1.5TB drives now and when the time comes for larger drives, to worry about it then.

You are spot on, my advice would be to get all 1.5 TB drives now and then add 2 TB drives as they become cheaper.  2 TB drives should occupy the 'sweet spot' within a few months.  I believe that Purko would not agree with me on this, see the above discussion for his reasons.

 

However, I'm sure both Purko and I would agree that it is very easy to upgrade a parity drive to a larger drive at any point in the future.  All you have to do is install the new drive, start up the server, go to the 'devices' page and unassign your current parity drive, assign your new drive as parity, then go back to the 'main' page and start the array.  unRAID will begin building parity on the new drive.  Once that has completed and you are satisfied that all your data is still safe, you can go back to the 'devices' page and assign your old parity drive into a data or cache drive slot.

 

There is even a method to ensure that you don't lose parity protection during the process, if you are extra-paranoid.  If for some reason the new parity drive fails as you are populating it with the parity info, then you can use the Trust My Array procedure to revert back to your old configuration.

 

• Even though I won't be running my server 24x7, I never thought that hard about the cost of running the drives.  I currently have 8x250GB drives in a traditional Raid5 array and thought that I might initially re-use these in addition to 2x1.5TB drives when I move to unraid.  It probably makes more sense to retire these 8 drives and replace them with a 1.5TB drive from the start.  These are 4 or 5 year old drives, so they probably use even more than 6Wh each.

You can look at the manufacturer's tech specs and determine how much power each of those 250 GB drives consumes, then perform a simple calculation similar to mine in the post above.  My guess is that the difference will still be somewhat negligible, though probably on the order of $20-30/year.  However, keep in mind Purko's point about the associated hardware necessary to run 5 drives instead of just one.  If you have to buy a $50 expansion card or a fancier motherboard to accommodate all the extra drives, then you must factor those costs into the equation as well.  As I have said before, I believe that my method only applies when considering upgrading an already built, fixed-number-of-drives server (though Purko and I continue to disagree on this point).

 

• This is probably only a question that I can answer as to what my time is worth to me, but what do others use as a guideline for excess capacity vs. the value of your time?  Let's say, theoretically that I need 2TB of space now and this will increase by 1TB/year.  Do I start with 2x1.5TB drives now, then in 2 years add more capacity a drive at a time or do I start with 3x1.5TB drives and avoid having to open up my server and add more drives for 1 or 2 additional years?  Benefits of adding excess capacity now are that I save time not having to add drives to my server later and cons are that drives will be larger and cheaper in 2 years than now and I will be wasting electricity powering up excess drives now when I don't need them.

Like you said, that's really a judgment that only you can make.  If you are using external (hot swap) drive bays, then adding or replacing drives is most likely a 5 minute procedure.  If you are using internal drive bays, then it may be a 15 minute procedure.  In my view, I would rather open up my server more often and save money when buying drives (waiting for sales, riding the wave of hard drive 'sweet spots', etc.).

drives will be larger and cheaper in 2 years than now and I will be wasting electricity powering up excess drives now when I don't need them.

 

I would go this route. I have a fairly decent sized server and I'm still on a 1.5tb parity drive, a couple 1.5tb drives and the rest 1tb drives.

 

When I need space, I swap one of the 1tb drives for a 1.5tb drive, retire, reuse somewhere else or sell the 1tb drives.

 

This is one of the reasons I went with the trayless removable sata units.  See my larger server in the pimp your rig thread.

http://lime-technology.com/forum/index.php?topic=2031.msg20486#msg20486

 

I can swap drives like they are floppies.

 

Upgrading parity is a simple process.

Generate parity on the new drive or do the parity swap procedure(I think).

 

In response to another of Purko's points:

 

Do you know that the cost of power can greatly surpass the cost of the hardware for its life time?

No, I can't say I did know that.  Intuitively, I want to believe you, but I would have to see the numbers to be truly convinced.  Still, its another interesting point you have brought up, and I thank you for that.  I'll go crunch some numbers now and see what they tell me.

 

Based on the calculations above, I don't believe this is currently true, at least in terms of hard drives.  In the future when improved manufacturing techniques allow hard drives to drop drastically in price, then I can see that happening.  However, at current prices, that simply does not seem to be true.

Do you know that the cost of power can greatly surpass the cost of the hardware for its life time?

 

No, I can't say I did know that.  Intuitively, I want to believe you, but I would have to see the numbers to be truly convinced.  Still, its another interesting point you have brought up, and I thank you for that.  I'll go crunch some numbers now and see what they tell me.

 

Based on the calculations above, I don't believe this is currently true, at least in terms of hard drives.  In the future when improved manufacturing techniques allow hard drives to drop drastically in price, then I can see that happening.  However, at current prices, that simply does not seem to be true.

 

There are assumptions here.

The cost of hard drives are always going down and/or the drives are getting denser.

 

The cost of power is constant. This is one situation we cannot count on. The cost of raw fuel can dictate how much our power costs are. This number can go lower or higher.

There have been times when the power company has to raise the prices, However, when fuel costs do drop later on, they do not seem to lower the prices.

 

You can almost always count on the rising cost of fuel and the rising cost of power.

How does that get figured into the equation?

Also factor in the rising cost of fuel may increase the cost of manufacturing.

Perhaps we just see this issue from different perspectives, but in my view there is a large practical difference between building and upgrading.

...In this (I believe rather common) scenario, the server is already bought and paid for...

 

If you regard these as different scenarios, you are only deceiving yourself.

 

I'm really not following you here.  If the server hardware is already bought and paid for and one is not planning on upgrading it, doesn't it cease to be a variable?

 

 

No, it doesn't.  It is still your investment.  It is still your money that could be spent elsewhere.

 

And, it will still be your money when you will need capacity beyond what you can put in that server.

 

Just the same, there's no practical difference between a house you've paid in full for, and a house on which you have mortgage payments.

The only intersting number about it is your equity.

 

 

The cost of fuel is not constant.

 

True, and as you said, we cannot predict the rate of change (increase).  Perhaps it will perfectly balance out any savings gleaned from consolidating server capacity onto fewer drives.  We simply have no way of knowing.

 

And, it will still be your money when you will need capacity beyond what you can put in that server.

When one has to consider upgrading a server past its maximum drive capacity, i.e. building a new server, then yes, I agree with you.  However, as I've emphasized before, I'm proposing a scenario (which I believe to be common within this community) in which there is no reason to build a new server when you can just upgrade the drives on a periodic basis.

 

My server currently holds 8 drives: 1 parity, 6 data, and 1 cache.  Assume my case is filled to the brim and cannot hold any more drives (it can actually hold 2 more, but only with extra drive brackets and other hardware).  I'll stick with my original estimate of adding 250 GB of new data to my server every month.  That means that I'll need an extra 1.5 TB of storage capacity every 6 months.  Given that every upgrade to my server necessitates that I sacrifice/retire my smallest drive, then I will essentially need to upgrade my server by 2 TB every 6 months until all my 500 GB drives are gone, then 3 TB every 6 months until all my 1 TB drives are gone, and so on.  Therefore to keep up with this trend, I'll plan on buying a new drive every 3 months or so.

 

Today I run out of space and need to order a new hard drive.  My server's current capacity is 5.5 TB, spread out across: 1 x 1.5 TB, 3 x 1 TB, and 2 x 500 GB.  Clearly I'm going to replace one of the 500 GB drives with whatever new drive I decide to buy.  Using my method, I decided to buy a new 1.5 TB WD Green.  My server's total capacity increases by 1 TB (since I'm removing 500 GB and adding 1.5 TB) up to 6.5 TB.

 

In 3 month's time, I run out of space again.  My data drive configuration is now: 2 x 1.5 TB, 3 x 1 TB, and 1 x 500 GB.  I expect that by that time the 2 TB drives will occupy the 'sweet spot', so I buy a 2 TB WD Green.  The 2 TB drive must be my new parity drive, since it is the largest, and my current 1.5 TB parity drive becomes another data drive.  Therefore, my data drive configuration is now 3 x 1.5 TB, and 3 x 1 TB.  All my 500 GB drives have been replaced.  My server's capacity is now 7.5 TB.

 

In 6 month's time, I again run out of space.  My data drive configuration is 3 x 1.5 TB, and 3 x 1 TB.  Perhaps 2 TB drives still occupy the sweet spot, or perhaps there's a 3 or 4 TB drive out that claims the title.  Either way it wouldn't really matter, since only 1 TB will be added to my server's data capacity (since the new drive will be either my new parity drive or it will be equal in size to my current parity drive, and I must retire a 1 TB drive to free up a slot).  My new data drive configuration is 1 x 2 TB, 3 x 1.5 TB, and 2 x 1 TB for a total capacity of 8.5 TB.

 

This trend continues on and on until my basic hardware (mobo, ram, cpu, etc.) eventually dies and I have to build a completely new server.  As you can see, my server's capacity grows by an average of 1 TB every 3 months, leaving a 250 GB buffer to allow for some variation in my 250 GB of new data per month projection.

 

Many of you may protest that it is silly to assume that my need for new storage capacity will remain at a constant 250 GB/month on to infinity.  This is certainly true.  However, I expect that the ever-increasing size of available and cheap hard drives will continue to dwarf my ever-increasing storage needs, as they currently do.

 

This analysis works for my situation, but I fully understand that it may not work for yours.  If your server supports a different number of drives, or if your monthly increase of new data on your server is drastically different from mine, then this may not work for you.  Some people may in fact have to build a new server every year to keep up with their storage needs.  I don't.

Just the same, there's no practical difference between a house you've paid in full for, and a house on which you have mortgage payments.

The only intersting number about it is your equity.

Except that a house will "usually" appreciate in value over time, and in the future a larger house cannot "usually" be purchased for less money.    To purchase a larger house than you currently need (as long as you can afford it) is not "usually" a mistake.  You'll grow into it... or accumulate enough "stuff" to fill it (just like our disks  ).    Now, for the past few years you might have gotten caught up in the drop in values of houses, but in general, over enough time, it will be an investment that will appreciate in value.

 

A disk purchased today will depreciate in value over time,  don't believe me, I've got a number of "huge" 2GB drives you might be interested in...  If you don't think they are worth much now, I'll be glad to sell them for 50% of their original cost, and you can wait for them to be valuable once more.    Regardless, a similar sized disk purchased in the future will "usually" cost less.

 

In my opinion, to max out an unRAID server before you need the space is a mistake.  Track the sweet-spot in GB/$, purchase drives as you need or can catch an occasional super-duper-sale price.  unRAID is too flexible to do it otherwise. 

 

Joe L.

Whenever I buy a new hard drive for my unRAID server, my primary rubric is the bang-for-your-buck ratio, which is essentially how many GBs of space are you getting for every dollar you spend.  Consider this:

 

HDD

Price + Shipping

GB/$ (a.k.a. bang-for-buck ratio)

Western Digital 1TB$97.2710.28Western Digital Caviar Green 1.5TB$104.9914.29Western Digital Caviar Green 2TB$179.9911.11

 

The 1.5 TB drive is the clear winner.

 

I clicked on your links, and here's what I saw:

Western Digital Caviar Green 1.5TB -- $109.99

Western Digital Caviar Green 2TB -- $169.99

 

I guess when you set out to make a point, you can pick any numbers.

BTW, the latest four WD 2TB disks I bought were $139.99 shipped.

 

See, if there is One Truth, and you know that Truth, would they be making any disks other than 1.5TB?

Do you think people are idiots?

 

 

Just the same, there's no practical difference between a house you've paid in full for, and a house on which you have mortgage payments.

The only intersting number about it is your equity.

Except that a house will "usually" appreciate in value over time, and in the future a larger house cannot "usually" be purchased for less money.

 

Well, Joe, I was trying to make a point for a different reason. Apparently Rajahal comes from a culture where you consider what you fully own as somewhat 'free'. I've been there. And it takes some time to discover that such view is seriously misleading when you try to make business calculations.

 

 

In my opinion, to max out an unRAID server before you need the space is a mistake.

 

I fully agree with that. That's why we build a server with just a few disks in it,

and then add more disks as needed.

 

Rajahal is trying to make some distinction between when you own such a server, and when you are about to build such a server.

 

 

Apparently Rajahal comes from a culture where you consider what you fully own as somewhat 'free'. I've been there. And it takes some time to discover that such view is seriously misleading when you try to make business calculations.

 

I see your point, and I do not think the server is 'free' at any point.  I see my unRAID server as a monthly expense, akin to paying for online storage.  The monthly expense of my server is determined by:

 

The sum of....

A) The initial cost of the server (mobo, RAM, CPU, etc.)

B) The cost of the disks (parity, data, cache)

C) The amount of electricity it consumes

divided by...

D) The length of time (in # of months) that the server lasts

 

Clearly this doesn't work perfectly, since I made an inordinately large 'down payment' on the server (for initial cost), and I tend to buy new disks every few months, not one per month.

 

I only maintain that once the initial cost of the server is out of the way, all that is left to consider is the cost of upgrading disks.  I also maintain that the cost of electricity to run the server is more or less negligible, so I don't worry about it.

Whenever I buy a new hard drive for my unRAID server, my primary rubric is the bang-for-your-buck ratio, which is essentially how many GBs of space are you getting for every dollar you spend.  Consider this:

 

HDD

Price + Shipping

GB/$ (a.k.a. bang-for-buck ratio)

Western Digital 1TB$97.2710.28Western Digital Caviar Green 1.5TB$104.9914.29Western Digital Caviar Green 2TB$179.9911.11

 

The 1.5 TB drive is the clear winner.

 

I clicked on your links, and here's what I saw:

Western Digital Caviar Green 1.5TB -- $109.99

Western Digital Caviar Green 2TB -- $169.99

 

I guess when you set out to make a point, you can pick any numbers.

BTW, the latest four WD 2TB disks I bought were $139.99 shipped.

 

See, if there is One Truth, and you know that Truth, would they be making any disks other than 1.5TB?

Do you think people are idiots?

 

Yes, yes, prices change.  We all know this.  I originally made that table for someone who said they wanted to make a purchase that day, therefore I used the current prices (the 1.5 TB drives happened to be on sale that day).  The purpose of that table was to demonstrate my method is a sort of visual way.  You'll notice that in my calculations on the previous page, I used today's current prices (neither of which are sale prices).

• Even though I won't be running my server 24x7, I never thought that hard about the cost of running the drives.  I currently have 8x250GB drives in a traditional Raid5 array and thought that I might initially re-use these in addition to 2x1.5TB drives when I move to unraid.  It probably makes more sense to retire these 8 drives and replace them with a 1.5TB drive from the start.  These are 4 or 5 year old drives, so they probably use even more than 6Wh each.

You can look at the manufacturer's tech specs and determine how much power each of those 250 GB drives consumes, then perform a simple calculation similar to mine in the post above.  My guess is that the difference will still be somewhat negligible, though probably on the order of $20-30/year.  However, keep in mind Purko's point about the associated hardware necessary to run 5 drives instead of just one.  If you have to buy a $50 expansion card or a fancier motherboard to accommodate all the extra drives, then you must factor those costs into the equation as well.  As I have said before, I believe that my method only applies when considering upgrading an already built, fixed-number-of-drives server (though Purko and I continue to disagree on this point).

I am hoping that I have all the hardware.  My MB has 4 SATA ports and the initial testing that I have done indicate that my current raid card (BC4852) may be able to be used as additional SATA ports for unraid.  So 12 disks, should last me forever as I continue to add/upgrade to larger disks.

 

Thanks for all the advice.  I think I will probably start out with 3 1.5TB drives (1 parity and 2 data).  I am assuming that it would be a complete waste of money to use a 1.5TB disk for a cache disk and that I should re-use one of my current 250GB drives.  I will never write 250GB to the cache drive during a day.

 

Murray

 

Please evaluate my reasoning in this post.  I fear it was skipped over because I took too long in writing it.

I think I will probably start out with 3 1.5TB drives (1 parity and 2 data).  I am assuming that it would be a complete waste of money to use a 1.5TB disk for a cache disk and that I should re-use one of my current 250GB drives.  I will never write 250GB to the cache drive during a day.

That sounds quite reasonable to me.  I also would never use a 1.5 TB drive for a cache drive in the foreseeable future, as appealing as the 'warm spare' concept may be (I explain it in response to Q5 in this post).

Whenever I buy a new hard drive for my unRAID server, my primary rubric is the bang-for-your-buck ratio, which is essentially how many GBs of space are you getting for every dollar you spend.  Consider this:

 

HDD

Price + Shipping

GB/$ (a.k.a. bang-for-buck ratio)

Western Digital 1TB$97.2710.28Western Digital Caviar Green 1.5TB$104.9914.29Western Digital Caviar Green 2TB$179.9911.11

 

The 1.5 TB drive is the clear winner.

 

I clicked on your links, and here's what I saw:

Western Digital Caviar Green 1.5TB -- $109.99

Western Digital Caviar Green 2TB -- $169.99

 

I guess when you set out to make a point, you can pick any numbers.

BTW, the latest four WD 2TB disks I bought were $139.99 shipped.

 

See, if there is One Truth, and you know that Truth, would they be making any disks other than 1.5TB?

Do you think people are idiots?

 

Yes, yes, prices change.  We all know this.

 

So, basically, you have invented the division of a number (capacity of a disk) by a number (price of that disk).

Do you really think that warrants this much discussion?

 

 

So, basically, you have invented the division of a number (capacity of a disk) by a number (price of that disk).

Do you really think that warrants this much discussion?

 

In short, yes, I believe that this issue warrants this much discussion because it is an issue with which all unRAID users are confronted at some point or another.  There are countless threads on these forums in which members ask, essentially, 'should I buy this drive or that drive?'.

 

I haven't invented anything.  I advised a new-to-unRAID member on a way to determine which hard drive currently occupies the 'bang-for-your-buck' sweet spot.  You protested that this simple calculation isn't very accurate.  I conceded that you are correct in this member's case, since he was building a new server from scratch.  I then proposed that my method is appropriate when upgrading an already-built server.  You contested this as well, and I'm still trying to understand why.  I view this thread as a learning opportunity both for myself and others who are new to unRAID.  I consider your contributions to this discussion to be interesting and informative, and I have enjoyed the manner in which you have lead me to critically analyze my assumptions.

 

If you view this thread as fruitless, then you are welcome to bow out of the discussion, as you have already brought a plethora of advice and experience to the table.  Perhaps other members of this community will care to continue the discussion with me, or perhaps it will end here.  Either way, I feel that I have learned a new method (yours) for evaluating the value of a server's precious drive slots (your 'non-algorithm' from the previous thread).  For this I am grateful.

 

If you feel like continuing this, I would appreciate your feedback on my lengthy post on the previous page.  I feel that my reasoning is sound and my assumptions valid, but you have been successful at challenging my reasoning and assumptions in the past.  I'm here to learn, not to preach.

Well, Rajahal, what can I say?  You are right.

 

It's just that when you started this thread with such a massive post of reposts, it felt like we are blowing this topic out of proportion.

But if you are so interested in the topic, then yes, by all means, that's what the forum is for.

 

Anyway, you'll probably be interested of thinking about 'sweet spot' numbers in this scenario.

 

Purko

 

-----

P.S.:  BTW, I'm glad you kept your composure.   

 

 

It was/is an interesting exchange of thoughts and positions.

it made me think about strategy and then I realize we were assuming certain constants.

I knew there was a reason for retiring my smaller drives, I just could not remember it. 

 

Seriously, since adding unRAID to my network and the increase in power costs, my out of pocket expense is higher.  So as time goes on, I'm taking drives out of other machines and putting them in unRAID with larger hard drives and network shares rather then local drives.

 

Plus I'm switching out processors to mobile or LV processors where I can.