SATA vs NL-SAS vs SAS For New Array
-
@BRRABill said:
The server I am looking to put these drives into has 3.5 inch bays, so I could use either 3.5 or 2.5 inch disks.
Is there a preferred one to go with in this scenario? Or does size really not matter, so to speak? Is one preferred?
Sometimes on xByte the 2.5 drive is cheaper than the 3.5 which is why I ask...
2.5 has faster seek because... well it sweeps less disk!
2.5 is generally more $/GB
2.5 generally uses less power
2.5 @7200rpm is a pretty dang fast drive per $/GB3.5 comes in larger sizes
3.5 comes in 15k rpm (2.5 used to come in 10k IIRC but they were rare)
3.5 comes in "hybrid" SSD chunk addedSpecifically for your scenario I'd go 3.5 PURELY because 3.5 to 2.5 adapter thing-a-ma-whatsits can be expensive, rattle and fiddly.
Were I spec'ing out a server I'd consider 2.5 for some scenarios.
-
I priced out all the scenarios I am looking at.
Basically, we only have about 300GB across our physical servers ATM, and I do not see it growing very quickly. Pretty sure I could get away with 900GB. The VMs on this server will be a file server, and a non-Exchange mail server (MDaemon). I've been watching the IOPS on the mail server, and they are pretty low.
Here are the potential options. Do these numbers seem about right? They don't take into consideration any cache. (This server has a PERC 710.)
Be curious to hear which way people would go if this was your server. Go for the most IOPS since storage isn't need? Go with a little extra storage and lower cost?
-
@BRRABill I'd go the 500's in OBR10.... but.......
You could get 3 cheap 960GB SSD for that money and RAID5 them....
Also that seems like a lot for 500gb drives, $75 a pop? shrug
-
@BRRABill said:
I priced out all the scenarios I am looking at.
Basically, we only have about 300GB across our physical servers ATM, and I do not see it growing very quickly. Pretty sure I could get away with 900GB. The VMs on this server will be a file server, and a non-Exchange mail server (MDaemon). I've been watching the IOPS on the mail server, and they are pretty low.
Here are the potential options. Do these numbers seem about right? They don't take into consideration any cache. (This server has a PERC 710.)
Be curious to hear which way people would go if this was your server. Go for the most IOPS since storage isn't need? Go with a little extra storage and lower cost?
Is that cost the total for all the drives that would be used?
If IOPS needs are met, I'd go with less expensive. What would the need to spend more money be?
@MattSpeller said:
@BRRABill I'd go the 500's in OBR10.... but.......
You could get 3 cheap 960GB SSD for that money and RAID5 them....
Also that seems like a lot for 500gb drives, $75 a pop? shrug
Depends, are we talking consumer or enterprise class drives at that $75 each?
-
@travisdh1 said:
Depends, are we talking consumer or enterprise class drives at that $75 each?
He does not mention SAS so I was assuming 7200rpm SATA commodity disposable spinny rust.
-
Sorry, these are all DELL drives, for a DELL server.
After having my stuff crash twice now with non-DELL drives, I will heed the warning and only put in DELL stuff.
-
@MattSpeller said:
Also that seems like a lot for 500gb drives, $75 a pop? shrug
Actually $99 each, from xByte.
I already have 2 of them, came with the server.
-
@travisdh1 said:
Is that cost the total for all the drives that would be used?
Yes, for DELL drives from xByte.
-
@BRRABill said:
@MattSpeller said:
Also that seems like a lot for 500gb drives, $75 a pop? shrug
Actually $99 each, from xByte.
I already have 2 of them, came with the server.
Then your sheet should show the rest cost.. just because you already paid for them does not matter, it's still money spent.
-
@Dashrender said:
Then your sheet should show the rest cost.. just because you already paid for them does not matter, it's still money spent.
Well, there are a couple ways of looking at this.
Considering the drives will sit unusued if I go anything OTHER than 7.2K 500GB, it's actually wasted money if I were to go that way.
Granted, I'm not saying to spend more money to save making a bigger mistake. I am just saying I'm not entirely sure how to cost that out. For MY scenario that is the "additional cost" to implement.
-
In your situation you need two tables - one showing real costs, and one for management that show the what you will be spending to do the project costs. The difference is the things you already have on hand.
-
@MattSpeller said
He does not mention SAS so I was assuming 7200rpm SATA commodity disposable spinny rust.
Yeah the 7.2K drives were SATA.
The 10K and 15K were SAS.
The 7.2K NL-SAS drives would have added at least $200 to the total cost, not including how one would figure out how to figure out my cost including the drives I already own, which are not NL-SAS.
-
@BRRABill said
Yeah the 7.2K drives were SATA.
"Enterprise" SATA, I should say, not consumer grade.
FWIW.
-
@BRRABill said in SATA vs NL-SAS vs SAS For New Array:
The server I am looking to put these drives into has 3.5 inch bays, so I could use either 3.5 or 2.5 inch disks.
Is there a preferred one to go with in this scenario? Or does size really not matter, so to speak? Is one preferred?
Sometimes on xByte the 2.5 drive is cheaper than the 3.5 which is why I ask...
If you have 3.5" bays, you'd use those disks. With very rare exception. 3.5" is much cheaper but uses lots of space. But you are already using that space so...
-
@BRRABill said in SATA vs NL-SAS vs SAS For New Array:
Well, there are a couple ways of looking at this.
Considering the drives will sit unusued if I go anything OTHER than 7.2K 500GB, it's actually wasted money if I were to go that way.
No, that's called the sunk cost fallacy and should never be considered. That money is already spent and is part of both equations equally. You own the disks, so the possibility of using the disks for "free" should be considered as an option, but the cost of them is not a factor. If you got them for free, for $300 or for a billion dollars doesn't matter, what you have today are some disks and nothing more. Don't give into the emotions of sunk cost.
-
@scottalanmiller said in SATA vs NL-SAS vs SAS For New Array:
No, that's called the sunk cost fallacy and should never be considered. That money is already spent and is part of both equations equally. You own the disks, so the possibility of using the disks for "free" should be considered as an option, but the cost of them is not a factor. If you got them for free, for $300 or for a billion dollars doesn't matter, what you have today are some disks and nothing more. Don't give into the emotions of sunk cost.
I agree with that, but the discussion was whether to include that cost in my "total cost" column above.
I guess a better question would be ... if someone gave me 4 of those drives for free, would I put $400 for total cost or $800?
If this was a discussion on straight cost vs. IOPS, I think I would use the $800. But since I am trying to figure it out for MY scenarion, I would think the $400.
-
@BRRABill said in SATA vs NL-SAS vs SAS For New Array:
@scottalanmiller said in SATA vs NL-SAS vs SAS For New Array:
No, that's called the sunk cost fallacy and should never be considered. That money is already spent and is part of both equations equally. You own the disks, so the possibility of using the disks for "free" should be considered as an option, but the cost of them is not a factor. If you got them for free, for $300 or for a billion dollars doesn't matter, what you have today are some disks and nothing more. Don't give into the emotions of sunk cost.
I agree with that, but the discussion was whether to include that cost in my "total cost" column above.
I guess a better question would be ... if someone gave me 4 of those drives for free, would I put $400 for total cost or $800?
If this was a discussion on straight cost vs. IOPS, I think I would use the $800. But since I am trying to figure it out for MY scenarion, I would think the $400.
Depends, are you trying to make a business decision or just a post mortem?
-
@scottalanmiller said in SATA vs NL-SAS vs SAS For New Array:
No, that's called the sunk cost fallacy and should never be considered. That money is already spent and is part of both equations equally. You own the disks, so the possibility of using the disks for "free" should be considered as an option, but the cost of them is not a factor. If you got them for free, for $300 or for a billion dollars doesn't matter, what you have today are some disks and nothing more. Don't give into the emotions of sunk cost.
I understand the concept of sunk cost. (Unless the rest of this post demonstrated I do not. Then just forget I said that.)
But if their "free nature" allows the cost of a system utilizing them to be less, wouldn't that make that option, if equal to the other options, more attractive?
AKA, I not making the decision BASED off of having these disks. That would be an incorrect way to move forward. Throwing good money after bad money and all that.
But to me, it seems like the scenario that utilizes these disks (the 8 7.2K SATA solution) makes just as much sense if not more than the others, taking everything into consideration. Though I'll admit to still being totally on the fence with all options.
I haven't even really ruled out just trying a different set of SSDs from xByte.
-
@BRRABill said in SATA vs NL-SAS vs SAS For New Array:
But to me, it seems like the scenario that utilizes these disks (the 8 7.2K SATA solution) makes just as much sense if not more than the others, taking everything into consideration. Though I'll admit to still being totally on the fence with all options.
Probably, but only because part of the equation is owning four disks today that you can use, if you want. That they are paid for or free or whatever, isn't a factor. The disks themselves are part of what you have. The amount that they cost are not part of "what it will cost."
-
@scottalanmiller said
The amount that they cost are not part of "what it will cost."
That was my original theory on the matter.
