Seriously, the GPL is about as simple as it gets. It has to be written in legalese so it's going to be a bit obfuscated no matter what,...
A quibble. I suspect you meant "incomprehensible" by obfuscated, but I will respond to the word you used: "legalese," as you put it, is infinitely far from "obfuscated." It is very very clear, generally worded quite specifically, using a jargon of its parent language that is more precise than the parent language alone, with specific words with well-understood legal meanings being chosen deliberately and so forth. My opinion, anyway.
You nervous about exceeding 3000 pounds? Anyway, yah, I know you can do this, but admit it: you had to get an engineer to analyze whether or not you could.:-)
Putting new stuff in - a rack is 42U high. Right at the top of that rack, is going to require overhead lifting.
With blades and other new high density things (48 disk trays, anyone), a lift will be required regardless.
But yah, good analysis.
The thing about being a data center designer is that it is multidisciplinary. You have to know a lot about facilities in particular to be anywhere near qualified, and all the computer stuff is there as well on so many levels.
I'm a fan of Carl Sagan, but I do find it kind of amusing that he would easily reject one idea that there is no evidence for (God), but so willingly embrace another idea for which there is no evidence (intelligent extraterrestrial life).
Why, this is no surprise at all. Contact is a roadmap explaining exactly how Sagan himself would have liked to have been contacted by the creator. Who's to say he wouldn't have liked there to be "more to this"? I say he might have liked that, but was properly incredulous.
I've met with these folks personally. The upcharge for the card is, essentially, the custom ASICs that do the 'RAID'. With a little volume, the pricing will be much more in line with other flash.
Anyway, no, they're not a replacement for a SAN. That's just marketing hyperbole.
You'll find them in your SAN soon enough, though. Their real target market is things like embedded devices inside SAN/NAS controllers. Think "journal device" or MRU unit for SAN blocks.
They would also make for a great swap file device. Think "backing store" for a cache unit. Think "WAN accellerators".
I'm just waiting for the price to come down out of the stratosphere, so as to be sufficiently affordable for "enthusiast" use.... I want one.:-)
All the high speed flash devices coming out now are essentially 'sort of' native RAID solutions. They feature multiple lanes, akin to RAID-0 striping. That's how the FusionIO product gets it's 600MB/s write speed. It has 160 lanes.
All the 'latest' SSD's are dominating the hard drive now. FusionIO's special case, utterly devestates the hard drive.
I might be impelled to buy a 2TB SATA drive if, in the drive, there was a few GB of SSD circuitry that itself ran at the full speed of the SATAII bus. You know, a persistent cache
Anyway, I plotted a similar set of intersection curves, coming to about the same conclusions. Your data is much more specific than mine, though, so I was wondering where you got the exact figures?
There has been this trend of lightweight virtualization lately, that people are investigating for the purpose of containerizing single applications. Think "security," if you want to know what they're thinking about.
Anyway, in those environments, it is perhaps not necessary to have much "installed" in the OS area at all. Not even a fully featured OS at all. It just needs to be able to run programs.
What I'm thinking about here is a sort of "XP compatibility mode" checkbox that really works.
A couple of gigs of storage and a small memory footprint are increasingly irrelevant. $25/GB last I checked, for RAM... although of course prices vary. *wink*
There is not particularly much overhead with modern virtualization, no. SPEC CPU benches score at around the 97% level or so. I know, I've run them.
Some forms of I/O are hit by 15-25% (in particular the network), but this matters little to the home user, as the low speeds the average home is wired with are not hit at all.
As for your judgment that doing such things is "bad for the customer," I disagree. Customers discriminate heavily towards some very simple things. One of them is compatibility.
You will of course understand that virtualization and emulation are different technologies. In one technology, most instructions pass through "as-is" to the CPU. In the other, they do not.
So, virtualization will work find for compatibility layers insofar as you do not switch CPU architectures. Performance is generally pretty damn good.
Yah. Windows Vista has been a bit of a learning experience for them. What they discovered is that the popular press, overflowing with security concerns, was not entirely representative of their customer base. Their customer base does want security, but they by no means want their security ahead of compatibility... or even convenience, for that matter.
Vista's mistakes are understandable from a certain point of view.
Really, they should take a major hint from apple. Go ahead and make major transitions, but use virtualization to bridge the gap. Under no circumstances break compatibility.
I buy redundant active-passive HA storage controllers, all storage attached to the system and networked to the storage controllers, with fibers, SFPs, RACKED, and with IP addresses inserted for me at the factory and turnkeyed for us, for under $2/GB... per usable gigabyte, that is, after the file system and RAID-6 overhead is removed... from one of the leading and most well known vendors in the storage community today.
That's for Tier-II. For Tier-1, it's in the $4.50 range. I am presuming that the HP solution was a SATA one, and hence my eye rolling at their fantabulous $2/GB figure.
Did I not tell you I get to see prices (as in quotes for multi-petabyte systems) from all the significant magic quadrant vendors? I did say that, didn't I?
I'll give HP a bit of credit, at least, for doing with PolyServe something that traditional NAS vendors can't yet do, and therefore perhaps some credit on pricing.
But $15/GB? Salesman, please step away from the crack pipe.
Nah, what they mean when they refer to "Web 2.0" hardware is the same as Web 1.0... distributed stateless servers. Commodity replaceable parts, with software architectures designed to "run anywhere, we don't care where".
With that approach, the big SMP systems are now the detritus of technologies past. Everything is cluster this, cluster that, basically. Web 2.0 doesn't change this any, but it makes for a nice buzz word.
But never mind that. The article presents HP as having a storage solution at a "significant fraction" of the cost of their competitors. It goes on to state that "Current enterprise storage costs are around $15 per gigabyte".
Um. Hello?!? I get to see the no kidding per usable gigabyte prices of every significant vendor in the Gartner magic quadrants... even the most expensive vendors, top of the line, king of the kings, are no where near $15 per gig! WTF!?
If an HP salesman told me something like that, I might have to show him the door. "Lies do not become us."
And if they are targeting $2/GB, they really need to be looking at the competition. Seriously. Many vendors will be sub $1 before the end of the year.
It's more a phenomenon of the founders not really anticipating properly that the people would lay down and give it all up, although of course they did have some inkling. What to do, what to do?
The Google and Carnegie Mellon studies can both be found online. I think these are drives that "hard fail," and need to be RMA'd, but if you're really interested, take a look. They were both presented at one of the Usenix FAST conferences. 06, I think.
Let me help you out, however. How many drives exactly, do you think Seagate tests for 1,200,000 hours before going to market with a drive series? Exactly zero.
The difference between Seagate's data and the Google and Carnegie Mellon data, I would say is that the latter teams really are running large batches of drives and seeing no-kidding when they fail.
The drive makers use some other methodology. There being only 8760 hours in a year, dontcha know.
BTW, I think the Carnegie Study said that SAS and SATA drives fail the same, BTW. And Google said something similar.
...my point was they were gerrymandering their MTBF the same way as HDD makers...
I personally have no idea how they are gerrymandering their MTBF, hence my comment. The advantage of this particular statement of mine is that it has a 100% chance of being correct. *wink*
"One cannot know whether or not the SSD makers "lie the same" as the disk makers."
Sure you can, they're Human of course they lie.
I have no doubt, but we cannot know if they lie the same. What I mean is that just because hard drives seem to fail 15X more often than manufacturers claim, according to some third party study, doesn't mean the SSD drives will fail 15X more often than the SSD makers claims.
These are different industries, and different circles of practice and so forth.
This isn't a professional and academic circle. Information is available at the touch of a button. Anyway, if you read Spatial's response, there was a funny in it. Silver platter, get it?
Slashdot Mirror
Seriously, the GPL is about as simple as it gets. It has to be written in legalese so it's going to be a bit obfuscated no matter what,...
A quibble. I suspect you meant "incomprehensible" by obfuscated, but I will respond to the word you used: "legalese," as you put it, is infinitely far from "obfuscated." It is very very clear, generally worded quite specifically, using a jargon of its parent language that is more precise than the parent language alone, with specific words with well-understood legal meanings being chosen deliberately and so forth. My opinion, anyway.
C//
You nervous about exceeding 3000 pounds? Anyway, yah, I know you can do this, but admit it: you had to get an engineer to analyze whether or not you could. :-)
C//
Putting new stuff in - a rack is 42U high. Right at the top of that rack, is going to require overhead lifting.
With blades and other new high density things (48 disk trays, anyone), a lift will be required regardless.
But yah, good analysis.
The thing about being a data center designer is that it is multidisciplinary. You have to know a lot about facilities in particular to be anywhere near qualified, and all the computer stuff is there as well on so many levels.
Blade weight, blade smeight. Try those 48 drive ultra high density storage trays (like thumpers). A full rack pushin' toward two ton rack weights!
I am the raised floor nazi: no more raised floor for you!
C//
I'm a fan of Carl Sagan, but I do find it kind of amusing that he would easily reject one idea that there is no evidence for (God), but so willingly embrace another idea for which there is no evidence (intelligent extraterrestrial life).
Why, this is no surprise at all. Contact is a roadmap explaining exactly how Sagan himself would have liked to have been contacted by the creator. Who's to say he wouldn't have liked there to be "more to this"? I say he might have liked that, but was properly incredulous.
C//
The fusion IO system stripes 160 wide, the way I understand it...
C//
I've met with these folks personally. The upcharge for the card is, essentially, the custom ASICs that do the 'RAID'. With a little volume, the pricing will be much more in line with other flash.
:-)
Anyway, no, they're not a replacement for a SAN. That's just marketing hyperbole.
You'll find them in your SAN soon enough, though. Their real target market is things like embedded devices inside SAN/NAS controllers. Think "journal device" or MRU unit for SAN blocks.
They would also make for a great swap file device. Think "backing store" for a cache unit. Think "WAN accellerators".
I'm just waiting for the price to come down out of the stratosphere, so as to be sufficiently affordable for "enthusiast" use.... I want one.
C//
All the high speed flash devices coming out now are essentially 'sort of' native RAID solutions. They feature multiple lanes, akin to RAID-0 striping. That's how the FusionIO product gets it's 600MB/s write speed. It has 160 lanes.
All the 'latest' SSD's are dominating the hard drive now. FusionIO's special case, utterly devestates the hard drive.
C//
Well. You're right about that.
TOH:
I might be impelled to buy a 2TB SATA drive if, in the drive, there was a few GB of SSD circuitry that itself ran at the full speed of the SATAII bus. You know, a persistent cache
Anyway, I plotted a similar set of intersection curves, coming to about the same conclusions. Your data is much more specific than mine, though, so I was wondering where you got the exact figures?
C//
There has been this trend of lightweight virtualization lately, that people are investigating for the purpose of containerizing single applications. Think "security," if you want to know what they're thinking about.
Anyway, in those environments, it is perhaps not necessary to have much "installed" in the OS area at all. Not even a fully featured OS at all. It just needs to be able to run programs.
What I'm thinking about here is a sort of "XP compatibility mode" checkbox that really works.
A couple of gigs of storage and a small memory footprint are increasingly irrelevant. $25/GB last I checked, for RAM... although of course prices vary. *wink*
C//
There is not particularly much overhead with modern virtualization, no. SPEC CPU benches score at around the 97% level or so. I know, I've run them.
Some forms of I/O are hit by 15-25% (in particular the network), but this matters little to the home user, as the low speeds the average home is wired with are not hit at all.
As for your judgment that doing such things is "bad for the customer," I disagree. Customers discriminate heavily towards some very simple things. One of them is compatibility.
Another one is cost.
C//
You will of course understand that virtualization and emulation are different technologies. In one technology, most instructions pass through "as-is" to the CPU. In the other, they do not.
So, virtualization will work find for compatibility layers insofar as you do not switch CPU architectures. Performance is generally pretty damn good.
C//
Yah. Windows Vista has been a bit of a learning experience for them. What they discovered is that the popular press, overflowing with security concerns, was not entirely representative of their customer base. Their customer base does want security, but they by no means want their security ahead of compatibility... or even convenience, for that matter.
Vista's mistakes are understandable from a certain point of view.
Really, they should take a major hint from apple. Go ahead and make major transitions, but use virtualization to bridge the gap. Under no circumstances break compatibility.
C//
After buying one of these things, it will be more like a lesson in liquidity.
But does floor space really matter when you can cram 24-30kw of servers into the rack?
:-(
I cringe to have to tell you this, but yes.
I buy redundant active-passive HA storage controllers, all storage attached to the system and networked to the storage controllers, with fibers, SFPs, RACKED, and with IP addresses inserted for me at the factory and turnkeyed for us, for under $2/GB... per usable gigabyte, that is, after the file system and RAID-6 overhead is removed... from one of the leading and most well known vendors in the storage community today.
That's for Tier-II. For Tier-1, it's in the $4.50 range. I am presuming that the HP solution was a SATA one, and hence my eye rolling at their fantabulous $2/GB figure.
Did I not tell you I get to see prices (as in quotes for multi-petabyte systems) from all the significant magic quadrant vendors? I did say that, didn't I?
I'll give HP a bit of credit, at least, for doing with PolyServe something that traditional NAS vendors can't yet do, and therefore perhaps some credit on pricing.
But $15/GB? Salesman, please step away from the crack pipe.
C//
Nah, what they mean when they refer to "Web 2.0" hardware is the same as Web 1.0... distributed stateless servers. Commodity replaceable parts, with software architectures designed to "run anywhere, we don't care where".
With that approach, the big SMP systems are now the detritus of technologies past. Everything is cluster this, cluster that, basically. Web 2.0 doesn't change this any, but it makes for a nice buzz word.
But never mind that. The article presents HP as having a storage solution at a "significant fraction" of the cost of their competitors. It goes on to state that "Current enterprise storage costs are around $15 per gigabyte".
Um. Hello?!? I get to see the no kidding per usable gigabyte prices of every significant vendor in the Gartner magic quadrants... even the most expensive vendors, top of the line, king of the kings, are no where near $15 per gig! WTF!?
If an HP salesman told me something like that, I might have to show him the door. "Lies do not become us."
And if they are targeting $2/GB, they really need to be looking at the competition. Seriously. Many vendors will be sub $1 before the end of the year.
C//
Anyway, for single cell flash, I would think that at your volume you'd be obsoleting the drive before it wore out. For multi-cell, I'm not so sure.
If you're working in the financial services industry, shouldn't cost not be that much of a factor?
C//
It's more a phenomenon of the founders not really anticipating properly that the people would lay down and give it all up, although of course they did have some inkling. What to do, what to do?
C//
You can get SSD's with better performance than mechanical drives right now. For an extreme case, go check out http://www.fusionio.com/
The Google and Carnegie Mellon studies can both be found online. I think these are drives that "hard fail," and need to be RMA'd, but if you're really interested, take a look. They were both presented at one of the Usenix FAST conferences. 06, I think.
Let me help you out, however. How many drives exactly, do you think Seagate tests for 1,200,000 hours before going to market with a drive series? Exactly zero.
The difference between Seagate's data and the Google and Carnegie Mellon data, I would say is that the latter teams really are running large batches of drives and seeing no-kidding when they fail.
The drive makers use some other methodology. There being only 8760 hours in a year, dontcha know.
BTW, I think the Carnegie Study said that SAS and SATA drives fail the same, BTW. And Google said something similar.
C//
...my point was they were gerrymandering their MTBF the same way as HDD makers...
I personally have no idea how they are gerrymandering their MTBF, hence my comment. The advantage of this particular statement of mine is that it has a 100% chance of being correct. *wink*
C//
"One cannot know whether or not the SSD makers "lie the same" as the disk makers."
Sure you can, they're Human of course they lie.
I have no doubt, but we cannot know if they lie the same. What I mean is that just because hard drives seem to fail 15X more often than manufacturers claim, according to some third party study, doesn't mean the SSD drives will fail 15X more often than the SSD makers claims.
These are different industries, and different circles of practice and so forth.
C//
http://www.fusionio.com./ These products can be ordered now, although it will be more than two months for delivery (they is intense demand).
This isn't a professional and academic circle. Information is available at the touch of a button. Anyway, if you read Spatial's response, there was a funny in it. Silver platter, get it?
C//