Depends what type of RAID. RAID1 is mirrored disks. Your data is written to two (or more) drives.
So you have a very large overhead on your disk use - you get one disk worth of storage, using two disks for it.
However you also get a lot of resilience - you can afford to lose one drive, with no overhead/data loss.
And you get good read performance - as you surmise, if it can be read from two places, then you read from both drives at once, for a faster overall transfer. 5 copies of the drive, means... well, not _quite_ 5x transfer rate, but very nearly. (You'll probably hit problems eleswhere in your system, as the combined drive transfer rate exceeds system bus throughput)
It doesn't help for write performance though, because you still have to write that file to every drive.
The article is talking about RAID 5 though. That doesn't actually write data in multiple locations in teh same sense. It's a bit faster, because a file might have 100k on one disk, 100k on the next, and 100k on the next and so forth - meaning you can read the whole file a bit faster, but it's not a linear speed increase. It can also service multiple file requests faster, as a randomly distributed data request will be using all the drives at ones.
LTO4 will do 800Gb of raw capacity (vendors quote between 2:1 and 3:1 with compression).
It'll also do anything between 30MB/s to 120Mb/sec on a decent drive.
They're at a price that's more per gig than you'd pay for a cheap SATA drive for your home system, but less than you'd pay for the similar amount of high end drives.
I think they're keeping up quite nicely actually, and we still use it extensively. But for home use? No, a cheap SATA drive is 'good enough' as my backup. But then, I don't cry too much if my porn and mp3 archive is partially unrecoverable, but my employer has stuff that it considers significantly more valuable.
Ah that's the thing I love about IT. Starting with someone coming to you with a spec for something fantastic, and watching their dreams crumble into dust as you point out that everything is possible, but THIS thing will cost them.
But you're correct that backups are one of the places where we can pretty reliably get funding. People don't like losing data, especially when there's regulations that require them to have it, or get fined a lot.
But that's actually kind of odd, as there's such a huge mountain of 'stale' data accumulating on our filesystems. I wouldn't be surprised to find I could implement a rolling delete of anything that hasn't been accessed in a year, and no one would ever notice.
Depends if you're comparing the complete data loss, or any data loss figures.
If I take 5 drives, and slap some data on each of them without any form or RAID whatsoever, one failure will cost me some data. But it won't cost me _all_ my data - I'll still have 4 drives with some stuff on them, that's as usable as it was.
If I RAID5 across those 5 drives, I lose a bit of space for doing so, but what I gain is that no one failure will cost me data. It requires two drives to go down, within the time it takes to repair the RAID. That time is... well, quite variable - some systems use hot spares that come online immediately and start rebuilding. Others you'll have to wait a week as you go buy a new disk, swap out the failure and do the rebuild by hand.
So you may be right - you lose more data if a RAID 5 goes down, vs. the JBOD. You're also losing performance on your raid5 when it'd degraded, because it'd having to do on the fly parity calculations. This is entirely dependant on how long the replacement operation takes you though, and the probability of a second failure in the set during that window.
And of course, the relative acceptability of 'partial' data loss - I don't know about you, but it doesn't actually matter to me if I'm doing a restore of 1 drive or 5, my service is still hosed, but there's some environments where it might be OK. Restore 1/5th of your users mailboxes from last night's backup, as the other 4/5ths are working fine and haven't had a service interruption.
Even large storage arrays are using 'really big' disks, which mean long rebuild times and less 'duplication'. I've had a controller fault in one of my arrays which has meant 16 of my 300Gb drives have been offline for 24hours or so. Nothing died because of that, but with a failure rate of 1-2 drives per month, the odds of a double disk raid 5 failure was significantly higher than I liked.
Our replica array was fine, so we weren't faced with data loss, but mean time between failures becomes much more significant when you're talking about multi-hour drive rebuilds.
MMO events are really cool, but do require someone to think about the dynamics involved - there's always going to be someone who crashes it 'because they can' and there's always going to be a whole bunch of people who 'want to be the heros'.
That doesn't mean, however, you can't have an event. It just means you need to think 'what would a player do' as you're writing the script.
That's why I like EVE a lot - advancement is real time based, so you're not actually forced into the choice between 'advancement' and 'everything else'. Whether you're roleplaying, or grinding, your advancement rate doesn't change.
Actually in the places where universal healthcare exists, it's actually a positive thing that could could datamine and screen patients with risk factors. Treating heart disease is both easier and cheaper when it's done early, for example. Of course, that presumes you're in a country that doesn't consider basic quality of life details such as healthcare a profit source.
How much universal healthcare would you get, for the 3 trillion the war in iraq cost?
If you use the UK NHS as a model, with it's £105bn budget it'd probably about $900bn/year (at least, assuming you got approximately the same price per person).
So it's only about 3 years worth of 'free' healthcare for every man, woman and child in the US. But the thing is, when you start looking at health as a long term proposition, then things like 'health support' campaigns start to become very justifiable in terms of cost benefit ratio - it's worth educating and supporting people before their health hits a 'crisis' point.
Mediocre? What on earth gives you that impression? The _average_ cost of treating a person, is significantly less than the healthcare premiums you're paying already for your insurance. That's how your insurance company makes it's money, it's not rocket science. Doubly so when you bear in mind that universal healthcare is tax payer funded, so actually the overheads of the healthcare system are absorbed by more people. Making it fundamentally cheaper.
Oh, and there's no middle man, needing to make a profit for the shareholders.
I pay an MMO sub. I find the cost-benefit ratio to be quite good. It's a lot less than a 'brand new' game, and generally gives me more hours of entertainment. It doesn't stop me buying the 'latest' RTS (I like RTS games a lot) but it does mean since playing, I've dropped from buying a new game every month or so, to doing so once or twice a year, and only when it's _really_ good. The sums add up for me. (Although, admittedly somewhat less so now I've started running multiple MMO accounts)
You don't see what's fun with online gaming? Hmm, well, I'd turn that around. I don't see what's wrong with online gaming. Certainly there are some games that don't 'share' particularly well, but there's also a lot that do. Like, every RTS or FPS out there, benefits from having a real intelligence behind it, rather than a 'random AI' which at best is 'well scripted' and at worst is 'just moronic'.
In my experience, big stores do it just as much as little do, if you ask them. OK, so your local supermarket doesn't have much margin to play with, but one of your larger white goods retailers typically does - they build in various assumptions into the price, and have a larger margin which makes some room to negotiate, especially if you find out whether they're on commission, and when their commission 'cycles' are. (e.g. if it's on a monthly basis, it's not uncommon to find they'll be prepared to give you a particularly good price because it'll boost their commission at the end of the month).
They'll also typically build in card fees, delivery and maybe credit terms in some cases, which if you don't want them, you can probably push them to waive.
User expectation has also changed. We've got service level agreements that specify delivery times on emails being... erm. Something low, like under 5 minutes.
However much I try and explain what the point of an asynchronous protocol is, it's hard to get that through to a salesman who expects his emails to arrive instantly.
I hate getting an advert for piracy on a DVD I've bought legitimately. They're almost always 'no-skip', and are berating me for... having bought their product legitimately.
I find that annoying, and I'd be inclined to pirate stuff just to get rid of it.
What amazes me, is that you've just declared that you can't be bothered to virus/malware check, or firewall a system that can act as a gateway to your home network. Trust is also an issue, and however much you say you 'don't trust it', I'd lay odds that _most_ home users don't feel inclined to implement a DMZ rather than taking some more sensible security precautions.
Or are you saying you don't trust it, but still allow it free reign on your internal network?
OK, well if you belive some of the process methodologies out there, the trick is to present to your managers what the 'saving' is by doing your thing.
You do this by looking at what's _actually_ happening as part of the handling process - how does an out of band call get handled normally? What is the overhead for playing 'hunt the doc' vs. being able to pipe up in an IRC channel 'erm. What's... and who should deal with it'?
Productivity is massively increased by wiki-style adaptive documentation, and by passive streamed communications in the form of IM/IRC, because they fit in the gaps in a 'normal day'.
But ideally you'll pick on one current efficiency, highlight how much time it takes, and how your thing will reduce that. As time = money, it's easy to make 5 minutes per day, x however many employees enough to justify your time.
Perhaps the documentation revision process? What happens when 'something changes' or someone notices a typo/error in a document - how long does it take to review, and how much is 'lost' by it being inaccurate whilst it is reviewed. Also, how often does it actually change at the moment, and how often _would_ it if it was less of a pain. These are things that a Wiki makes handy.
A more esoteric example is that of morale, and how that can have such a pivotal effect, but you'll have a hard time getting solid numbers on that. So it's best kept as a general handwavy 'secondary objective', and just keep with the thing you have got some guesstimate numbers for.
I share a house with 6 other geeks. We have an IRC server, and we basically all log in to a particular channel, most of the day.
This allows us to indicate general status, and leave 'deferred feedback' communications with other house members, because they have a channel history. If one of my housemates is out, then he'll still see me suggsting going to our favourite noodle bar tomorrow, or that there's a bill pending for electricity, and I can't afford it this month.
We basically have a day long 'meeting' in the sense that we're all looking at IRC when it suits us, but aren't getting interrupted when we're otherwise occupied. It's a very effective form of communication, because with anything larger than a trivial number, a communication synchronisation takes more time and effort.
That there is. But as long as you understand how it should be done, then you're able to... well, start with the basics, and work up to idioms far quicker than learning from scratch. Whilst the exact mode of array implementation in a given language isn't necessarily something you know, you'll know that what you want _is_ an array, and thus you might need a few algorithms to go with it.
I personally hate VB, because every time I've heard someone declare it 'easier' I've gone on to see them pull out a wodge of utter bodge coding, which is basically unmaintainable. Because they didn't understand what a program was _supposed_ to be like, and VB let them do it. I'm currently working on one bit, that's a really ugly botch of 'goto' style code, but with a few prototyped procedures in there, just to throw me off the scent. It looks a lot like 5 different sets of copy-pasta code, and it's... well, I'm still itching for the time to do a full rewrite, and have almost justified it based upon the amount of extra lead time any form of modification to this ugly hack job is taking.
Erm. Which I think's agreeing with what you said:).
Someone who knows how to program, knows that http://mindprod.com/jgloss/unmain.html is a comedy site, even if they don't know the specifics of a given language.
Hmm, slipstreaming trucks? Does the concept of 'stopping distance' and 'full road visibility' come into it? The closer you are the vehicle in front (and you have to be fairly close to slipstream) the less you can see of the road in front, and the less time you have to react to an emergent condition in the vehicle you're doing that to.
And even a radio, creates a draw on the battery, which in turn means your alternator loads your battery more. Nothing like the load aircon, or even headlights create though I guess.
Mostly though, the best way of saving fuel is driving a car that isn't a fuel hog. Sadly I think that'll not happen due to cultural reasons - far too many people around the world consider SUVs, Pickup trucks and 'performance' engines to be a vital expression of their manliness. I also am under the impression that this is even more so in the US than in Europe, as there's just different economic pressures on vehicle size and efficiency.
Slashdot Mirror
So you have a very large overhead on your disk use - you get one disk worth of storage, using two disks for it.
However you also get a lot of resilience - you can afford to lose one drive, with no overhead/data loss.
And you get good read performance - as you surmise, if it can be read from two places, then you read from both drives at once, for a faster overall transfer. 5 copies of the drive, means ... well, not _quite_ 5x transfer rate, but very nearly. (You'll probably hit problems eleswhere in your system, as the combined drive transfer rate exceeds system bus throughput)
It doesn't help for write performance though, because you still have to write that file to every drive.
The article is talking about RAID 5 though. That doesn't actually write data in multiple locations in teh same sense. It's a bit faster, because a file might have 100k on one disk, 100k on the next, and 100k on the next and so forth - meaning you can read the whole file a bit faster, but it's not a linear speed increase. It can also service multiple file requests faster, as a randomly distributed data request will be using all the drives at ones.
On Slashdot, you don't need to make excuses for the size of your porn archive. We understand.
It'll also do anything between 30MB/s to 120Mb/sec on a decent drive.
They're at a price that's more per gig than you'd pay for a cheap SATA drive for your home system, but less than you'd pay for the similar amount of high end drives.
I think they're keeping up quite nicely actually, and we still use it extensively. But for home use? No, a cheap SATA drive is 'good enough' as my backup. But then, I don't cry too much if my porn and mp3 archive is partially unrecoverable, but my employer has stuff that it considers significantly more valuable.
But you're correct that backups are one of the places where we can pretty reliably get funding. People don't like losing data, especially when there's regulations that require them to have it, or get fined a lot.
But that's actually kind of odd, as there's such a huge mountain of 'stale' data accumulating on our filesystems. I wouldn't be surprised to find I could implement a rolling delete of anything that hasn't been accessed in a year, and no one would ever notice.
If I take 5 drives, and slap some data on each of them without any form or RAID whatsoever, one failure will cost me some data. But it won't cost me _all_ my data - I'll still have 4 drives with some stuff on them, that's as usable as it was.
If I RAID5 across those 5 drives, I lose a bit of space for doing so, but what I gain is that no one failure will cost me data. It requires two drives to go down, within the time it takes to repair the RAID. That time is ... well, quite variable - some systems use hot spares that come online immediately and start rebuilding. Others you'll have to wait a week as you go buy a new disk, swap out the failure and do the rebuild by hand.
So you may be right - you lose more data if a RAID 5 goes down, vs. the JBOD. You're also losing performance on your raid5 when it'd degraded, because it'd having to do on the fly parity calculations. This is entirely dependant on how long the replacement operation takes you though, and the probability of a second failure in the set during that window.
And of course, the relative acceptability of 'partial' data loss - I don't know about you, but it doesn't actually matter to me if I'm doing a restore of 1 drive or 5, my service is still hosed, but there's some environments where it might be OK. Restore 1/5th of your users mailboxes from last night's backup, as the other 4/5ths are working fine and haven't had a service interruption.
Even large storage arrays are using 'really big' disks, which mean long rebuild times and less 'duplication'. I've had a controller fault in one of my arrays which has meant 16 of my 300Gb drives have been offline for 24hours or so. Nothing died because of that, but with a failure rate of 1-2 drives per month, the odds of a double disk raid 5 failure was significantly higher than I liked.
Our replica array was fine, so we weren't faced with data loss, but mean time between failures becomes much more significant when you're talking about multi-hour drive rebuilds.
That doesn't mean, however, you can't have an event. It just means you need to think 'what would a player do' as you're writing the script.
That's why I like EVE a lot - advancement is real time based, so you're not actually forced into the choice between 'advancement' and 'everything else'. Whether you're roleplaying, or grinding, your advancement rate doesn't change.
Actually in the places where universal healthcare exists, it's actually a positive thing that could could datamine and screen patients with risk factors. Treating heart disease is both easier and cheaper when it's done early, for example. Of course, that presumes you're in a country that doesn't consider basic quality of life details such as healthcare a profit source.
And then they might have to move to a country that didn't have such regressive understanding of what healthcare actually means.
If you use the UK NHS as a model, with it's £105bn budget it'd probably about $900bn/year (at least, assuming you got approximately the same price per person).
So it's only about 3 years worth of 'free' healthcare for every man, woman and child in the US. But the thing is, when you start looking at health as a long term proposition, then things like 'health support' campaigns start to become very justifiable in terms of cost benefit ratio - it's worth educating and supporting people before their health hits a 'crisis' point.
Oh, and there's no middle man, needing to make a profit for the shareholders.
Which is ok, given that most of the rest of the world have figured out that state sponsored healthcare is a much better idea.
I pay an MMO sub. I find the cost-benefit ratio to be quite good. It's a lot less than a 'brand new' game, and generally gives me more hours of entertainment. It doesn't stop me buying the 'latest' RTS (I like RTS games a lot) but it does mean since playing, I've dropped from buying a new game every month or so, to doing so once or twice a year, and only when it's _really_ good. The sums add up for me. (Although, admittedly somewhat less so now I've started running multiple MMO accounts)
You don't see what's fun with online gaming? Hmm, well, I'd turn that around. I don't see what's wrong with online gaming. Certainly there are some games that don't 'share' particularly well, but there's also a lot that do. Like, every RTS or FPS out there, benefits from having a real intelligence behind it, rather than a 'random AI' which at best is 'well scripted' and at worst is 'just moronic'.
They'll also typically build in card fees, delivery and maybe credit terms in some cases, which if you don't want them, you can probably push them to waive.
Compared to most of the student digs I've seen, that's definitely a step up
However much I try and explain what the point of an asynchronous protocol is, it's hard to get that through to a salesman who expects his emails to arrive instantly.
I find that annoying, and I'd be inclined to pirate stuff just to get rid of it.
Or are you saying you don't trust it, but still allow it free reign on your internal network?
You do this by looking at what's _actually_ happening as part of the handling process - how does an out of band call get handled normally? What is the overhead for playing 'hunt the doc' vs. being able to pipe up in an IRC channel 'erm. What's ... and who should deal with it'?
Productivity is massively increased by wiki-style adaptive documentation, and by passive streamed communications in the form of IM/IRC, because they fit in the gaps in a 'normal day'.
But ideally you'll pick on one current efficiency, highlight how much time it takes, and how your thing will reduce that. As time = money, it's easy to make 5 minutes per day, x however many employees enough to justify your time.
Perhaps the documentation revision process? What happens when 'something changes' or someone notices a typo/error in a document - how long does it take to review, and how much is 'lost' by it being inaccurate whilst it is reviewed. Also, how often does it actually change at the moment, and how often _would_ it if it was less of a pain. These are things that a Wiki makes handy.
A more esoteric example is that of morale, and how that can have such a pivotal effect, but you'll have a hard time getting solid numbers on that. So it's best kept as a general handwavy 'secondary objective', and just keep with the thing you have got some guesstimate numbers for.
This allows us to indicate general status, and leave 'deferred feedback' communications with other house members, because they have a channel history. If one of my housemates is out, then he'll still see me suggsting going to our favourite noodle bar tomorrow, or that there's a bill pending for electricity, and I can't afford it this month.
We basically have a day long 'meeting' in the sense that we're all looking at IRC when it suits us, but aren't getting interrupted when we're otherwise occupied. It's a very effective form of communication, because with anything larger than a trivial number, a communication synchronisation takes more time and effort.
I personally hate VB, because every time I've heard someone declare it 'easier' I've gone on to see them pull out a wodge of utter bodge coding, which is basically unmaintainable. Because they didn't understand what a program was _supposed_ to be like, and VB let them do it. I'm currently working on one bit, that's a really ugly botch of 'goto' style code, but with a few prototyped procedures in there, just to throw me off the scent. It looks a lot like 5 different sets of copy-pasta code, and it's ... well, I'm still itching for the time to do a full rewrite, and have almost justified it based upon the amount of extra lead time any form of modification to this ugly hack job is taking.
Erm. Which I think's agreeing with what you said :).
Someone who knows how to program, knows that http://mindprod.com/jgloss/unmain.html is a comedy site, even if they don't know the specifics of a given language.
Or rather, $7.86/US gallon. (I still don't get why US Pints and Gallons are different sizes to the 'standard imperial' measurement)
And even a radio, creates a draw on the battery, which in turn means your alternator loads your battery more. Nothing like the load aircon, or even headlights create though I guess.
Mostly though, the best way of saving fuel is driving a car that isn't a fuel hog. Sadly I think that'll not happen due to cultural reasons - far too many people around the world consider SUVs, Pickup trucks and 'performance' engines to be a vital expression of their manliness. I also am under the impression that this is even more so in the US than in Europe, as there's just different economic pressures on vehicle size and efficiency.