Break MS up into at least two companies, each with full rights to the source code for Windows, patents, copyright, and trademarks, as well as half the resources (both manpower and money). Forbid these new companies from any collaboration or non-public communication. The desire of the new companies to make money will mean each will start working on a version of Windows that they think OEMs will like more than the other former MS company. OEMs will choose based upon which they think computer buyers will want. Both will be able to bundle anything they want and it won't be a problem because it won't be backed up with the monopoly influence since people can always choose to go with the other former MS company. In short order the market will open up, people will have real choices, and innovation will return.
This is one of the sanest suggestions on the topic I've ever seen.
Saying OSX is loosly based on Next is like saying Windows 2000 is loosly based on Windows 98... I'd say that the comparison between 2k and XP would be more apt taking Rhapsody, Blue Box and Yellow Box into account.
Windows 2000 (or XP) to Windows Vista would be the best analogy to NeXT -> OS X. The changes in both comparison of are much the same scope, magnitude and type (and the timeframes are roughly equivalent - although OS X has lagged a bit with the lower-level features, that's understandable since the codebase has not been under as intensive development as NT).
(With that said, there are a lot of low-level features that will only appear for the first time in OS X 10.5, that Windows has already had for many years, in particular with regards to efficient use of multiple CPUs.)
"Hardware comes with a warranty. Software comes with a disclaimer."
This does not change the fact that there are very real and tangible consequences for a fix to proprietry software that causes extensive breakage (lost customers and, more importantly, revenue), whereas the consequences for the same in the cowboy-esque OSS world have little impact outside the developer's ego.
Your software might not come with a warranty, but if enough people stop paying for it, rest assured that the vendor will take notice and act to try and recover those customers and revenue. If I stop using $J_RANDOM_HACKER's bit of open source code - almost always distributed with the attitude of "it's free, you get what you pay for, if there's a problem fix it yourself" - why is he going to care ?
Fixing Microsoft-created holes is the basic reason why anti-virus firms exist; and why they do such roaring business; and also why they are trusted MORE than Microsoft, which makes the underlying crappy OS.
Anti-virus programs don't "fix holes" in the OS, they fix holes in the *user*.
Outlook licenses are part of Office, not Exchange, which includes a couple of licenses for administration. Exchange Server Client Access Licenses (CALs) and Outlook licenses have to be purchased for each and every user.
An Exchange CAL includes an Outlook (or Entourage) license. This has been true since at least 2003 (and probably well before then).
That is certainly a valid stance. Why does Windows STILL use this lame thing called a Registry, other than for allowing dinosaur software that needs it to run?
Why would they ? It's a centralised, transactional, multiuser-safe database with fine-grained user permissions. The concept is perfectly sound and by most technical measures, vastly superior to alternatives like text files and trees of symlinks.
On the Mac, a quick search for recently installed startup items in two folders will give even a semi-knowledgeable user an idea if some recent download put something there. A quick deletion of these startup items will keep the malware from ever starting again.
Uh huh. And how about when malware burrows its way into the various UNIX init scripts and/or attaches itself to executables like/bin/sh ?
There's just as many places to hide in a UNIX system.
Only with Vista did the marketing team get a hold of the version numbers: NT 6.0
Rubbish. Vista is easily worth a major version increment, well in line with previous examples of same (NT 3.x to NT 4.x, NT 4.x to NT 5.x). The changes from XP to Vista are of the same magnitude as the changes from from OS X 10.0 (more like NeXT/OPENSTEP 4, really) to 10.5.
A major version number is supposed to indicate a break with backwards compatibility.
Agreed, but he's also got a point. BMWs (at least the M5) are not built for this kind of stuff. Perhaps a Porsche 911 would've been more appropriate.
Huh ? What *else* is an M5 built for if not a high-speed cross-country road trip ? They're designed for cruising on Autobahns where 100mph is "average".
Speed limits are there for a reason, so stick to them!
Indeed, and that reason typically has nothing to do with safety.
Traffic tickets should be depending on income/wealth instead of being fixed like they are now.
Exceeding the speed limit, in and of itself, should not be an offense (although it should certainly be taken into account in the case of other charges, like dangerous driving).
She commented to me on how fearful the men in the company seemed in their behavior toward women.
If she was talking about interactions with men within the company and especially during work time I can understand their attitude. The risk of a career-ruining sexual-harassment lawsuit should not be underestimated.
May I ask why? Are you saying any large drives in 4 or 5 disk raid-5 arrays somehow present a larger risk of failure than 4 or 5 10GB drives from years ago?
Because raw read error rates suggest that once drive sizes get over 1TB or so, the risk of a double-disk failure increases dramatically. There is also the factor of the sheer amount of time rebuilding an array like that takes and the significant performance and reliability impact that is suffered while the rebuild is occurring.
Even being generous and saying a RAID5 will rebuild at a constant 100MB/sec you're still looking at ca. *40 hours* to rebuild the array if a disk fails. During that rebuild, even a single sector read error on one of the other drives will likely hose the entire thing. Further, current raw read error rates pretty much guarantee (statistically) an unrecoverable read error about every 12TB or so worth of read data. In the context of rebuilding an array of 4TB drives, that's a risk factor _way_ above my comfort level.
Personally, I don't trust RAID5 _now_ and haven't for a couple of years. IMHO the probability of a second drive failure during a rebuild is already too high. That's why I only use RAID6 or RAID10 (and even RAID10 will start getting iffy soon, unless the expected error rates improve dramatically).
Today, we have a Prime Minister, complete with new Cabinet, who has never been voted into that office by the electorate, and indeed who just backed down from holding an election to get a mandate.
I'll admit my understanding of the UK system is a bit sketchy, but I was under the impression the position of Prime Minister was appointed by the Queen (typically the leader of the majority party) and Cabinet was then appointed by the Prime Minister (presumably under advisement from his senior Party members).
Or, in other words, the electorate has _never_ "voted in" any of the people in those positions, because that's simply not how the system works.
(Anyone who is about to bleat about a party political system where the Labour Party was elected would do well to remember that they were elected after Blair said he would serve a full third term. There is no rational way you can argue that the Brown administration has a mandate based on party politics. And even if they had been elected without that promise, our first-past-the-post system is so broken that you could hardly call it representative.)
Well they _do_ (apparently, I don't follow UK politics especially closely) have a mandate. That's the way the system works - people *don't* vote directly for certain individuals to be in specific positions (like Prime Minister).
The system here in Australia is basically the same, which is why I wonder about people who get so uptight about who is (or might be) Prime Minister. The point is that it doesn't matter because a) the Prime Minister could be (literally) anyone in parliament and b) individually, they don't hold any more power than anyone else there. It is little more than a ceremonial position and, ultimately, the Prime Minister's power comes from those he can influence, not inherently from his position as Prime Minister.
Problem: If you suffer random corruption on a disk, ZFS won't know anything about it. The RAID subsystem will have to try to guess which disk has actually failed, and if the RAID gets it right ZFS is happy. If not, ZFS fails a checksum and is stuck because the RAID layer already decided wrong, and probably overwrote the good data with the incorrectly recovered stripe.
You're not supposed to run ZFS on top of a hardware RAID array. That defeats most of the reasons for using ZFS in the first place.
Get a Clariion AX150. It takes SATA drives and can do RAID rather well...
We looked at the AX150i, but it's still ~3x as much as DIY. That's a big difference when all you really want is a huge slab of disk without high uptime requirements (99% availability is fine for us - and most people, IMHO). Also, I don't think the AX150 does RAID6 yet ?
I just happen to know this product, I'm sure other SAN vendors have cheap SATA arrays too.
Relative to its contemporaries, it's cheap - but compared to DIY it's just not in the same league. I do have to wonder why you'd be looking to to buy one for home when you could build something just as good (better, if anything) for a *lot* less money (at least, I'm assuming you don't need dual, redundant controllers for your home server:). Also, don't forget you won't be able to put off-the-shelf drives into it, and will be stuck buying drives from EMC (or off Ebay) at a ~500% markup.
There'd certainly be a bit of "cool" factor, but personally I'd rather have the extra disk space (or several thousand $$$$s worth of something else).
On top of the fact that you slam 16 drives in the bloody thing ( minimum for this kind of data ), and have half as hot spares to a raid6 array.
Holy shit, dude. There's responsible redundancy, then there's paranoia, then there's overkill, then, far off in the distance, there's having half a shelf dedicated to hot spares.
One hot spare per shelf is heaps. Consider a 7*750G RAID6 that suffers a disk failure. An array rebuild will take ca. 20 hours (assuming it's not offlined during the rebuild). Even a cheap SAN is going to come with a 24x7, 4hr response support contract (or, at worst, next day). That is to say, your failed drive will be replaced ca. 16 hours before your array has even finished recovering from the failure (or, worst case, around the same time as it is finishing). Just what is more than a single hot spare going to buy you ? Heck, if you've got staff on-site 24/7 (or equivalent), you don't even need a hot spare - just a cold spare and decent system monitoring.
If you're going to throw away half your raw space, the best way is to use RAID10 or RAID60 and at least get some performance benefits as well.
Sure, you *might* get away with a consumer level raid5/6 solution. But when things come crashing down, and they very likely would, you want something a bit more serious on the back end.
It's not especially hard to build a centralised storage solution with COTS parts that get up to about 90% of the features and performance of enterprise kit at well under half the cost - and despite what EMC and co. will try to tell you, that covers the requirements of the vast majority of environments. *Especially* when you start moving into archival/backup/second-tier systems with much lower reliability and uptime requirements (eg: can get away with single controllers), the cost difference becomes an order of magnitude or more.
Being in (roughly) the same industry and situation, I can sympathise. Our setup for study archiving is a front end "head" unit that receives data to a local 4-disk RAID10 (via a hardware - for transparency - 3ware card). It is connected to a number of back end "disk servers"[0] via GbE and iSCSI to present their disk space to the front end as block devices which are then stiched together using LVM.
Studies are archived daily, with an automated script simply carving an appropriately-sized LV out of the VG, formatting it, copying the data to it, indexing the study metadata, etc. We have about 15TB online (12.5TB allocated) and archive ca. 5000-6000 studies/140-160GB every day (compressed with JPEG lossless).
HOWEVER, our requirements are probably quite different - we keep archived images more for our own convenience, than because of any legal requirements, and only aim for about a 90 day rolling window (ie: the last 90 days worth of studies). Further, the front end unit has enough local disk to go for a solid 3-4 days without needing to archive off (and there is a backup machine if it fails), so while we've never had significant downtime on that particular aspect of our workflow, theoretically most of the physical machines could die for a couple of days with little more than a minor inconvenience.
While I think this is a sound (and dirt cheap - for what the whole thing is worth you couldn't even buy a single 16*500G drive array for our DS4800) solution, if your requirements are more strict - especially from a legal perspective - I would be extremely careful about recommending what is effectively a DIY solution. In particular, if your client ends up on court, they'll be a lot happier if they can go "we have a legally compliant storage system for our data, as certified by $LARGE_VENDOR" rather than "$CONSULTANT designed and built a storage system for us with COTS parts, but it hasn't been certified as compliant for storing patient data". There's also the greater uptime and redundancy of enterprise storage solutions (the part you're actually paying the 10x as much for)[1].
A simple SATA RAID controller interfaced with 4 such drives can give me 12TB of cheap, fast, storage. At 1TB per year, should be good enough for my needs. H/w vendors currently recommend expensive SAN boxes; which I don't like... no useful value for the application at hand.
Using RAID5 with 4TB drives would be insanity. Heck, using RAID5 *at all* with SATA ~160GB+ drives is crazy, IMHO, even for just my home server (16*250G RAID6) - let alone anything business-critical. Maybe if you restrict yourself to small drive counts (6-7), have a hot spare and do regular disk scrubs, it's worth taking a chance on - but I personally wouldn't do it.
[0] DIY jobbies, since no major vendor sells a box that holds so many SATA drives. We stuff 16 drives into a box, RAID6 them all together with Linux's software RAID, then export the RAID arrays with iSCSI Enterprise target. Each disk server has a pair of GbE links for bandwidth and redundancy and cat saturate both of them simultaneously from the RAID array. There used to be 5 servers, but 3 old IDE-based ones were recently replace with a single machine (holding triple the space). We also keep no less than two spare drives for each machine on-site and enough other spare parts (motherboard, CPU, etc) to completely rebuild one from scratch if necessary.
[1] This particular system is one of the few times I have chosen to go DIY over off-the-shelf, simply because the cost savings were so massive and the uptime/reliability requirements were relatively low (only 99% uptime). One thing I did discover researching this, however, was the dearth of low-end storage solutions in the Australian market (hence my decision to DIY).
So you're saying that a modern bloated OS is a better platform for running the new bloated application software, and that throwing a quad-core processor and several GB of RAM at the solution proves this?
No, I'm saying that a modern "bloated" OS takes better advantage of current hardware. Windows Vista will make _vastly_ better use of a quad-core, 4G RAM machine that NT4 ever could. As will Linux 2.6 (vs older versions).
No little bullying insults? It's all there above and a bit is quoted here:
No. Although when you choose to deceptively quote out of context, you can certainly make it look that way.
Try to talk about the subject matter instead of using methods to bully those with less confidence in their abilities or mistaken repect - it is uncivilised.
I am trying to talk about the subject matter. You, on the other hand, repeatedly ignore attempts to do so, preferring to concentrate on a deconstruction of how I choose to make my points.
The 1993 bit could throw the young off track a bit and make them question what they know - was that another little trick?
That depends on whether or not you consider a piece of factual information a "trick" or not.
There's a whole lot in earlier posts about your view of myself and very little about the subject matter.
In fact, my comments about attitude are explicitly conditional on whether or not certain mindsets are held. I'm not calling people with the opinions you have expressed fools for the hell of it, as an insult, I'm calling them fools for holding a biased opinion unsupported by fact or reasoned argument.
You have won this little game you are playing and I have already stated that I do not believe this old simplistic argumant that has shown no sign of occuring over more than a decade - I came here for something other than this little game that is a bit above kindergarden level.
Apparently you're a hypocrite as well. You attack me for not "talking about the subject matter" yet continually evade any attempts to engage upon it. Please note that simply repeating different variations of "you're wrong" is not "talking about the subject matter".
No little bullying insults? It's all there above and a bit is quoted here:
No. Although when you choose to deceptively quote out of context, you can certainly make it look that way.
Try to talk about the subject matter instead of using methods to bully those with less confidence in their abilities or mistaken repect - it is uncivilised.
I am trying to talk about the subject matter. You, on the other hand, repeatedly ignore attempts to do so, preferring to concentrate on a deconstruction of how I choose to make my points. That is to say, essentially nothing more than ad hominem "arguments".
The 1993 bit could throw the young off track a bit and make them question what they know - was that another little trick?
Indeed. Damn those facts, they trickses us all the time.
There's a whole lot in earlier posts about your view of myself and very little about the subject matter.
In fact, my comments about attitude are explicitly conditional on whether or not certain mindsets are held. I'm not calling people with the opinions you have expressed fools for the hell of it, as an insult, I'm calling them fools for holding a biased opinion unsupported by fact or reasoned argument.
You have won this little game you are playing and I have already stated that I do not believe this old simplistic argumant that has shown no sign of occuring over more than a decade - I came here for something other than this little game that is a bit above kindergarden level.
Apparently you're a hypocrite as well. You attack me for not "talking about the subject matter" yet continually evade any attempts to engage upon it. Continually repeating different variations of "I think you're wrong" is not "talking about the subject matter", it's sticking your fingers in your ears and seeing who can yell louder.
This will depend on which folder you're dragging to. The Start Menu displayed is combined from the user's personal Start Menu and the "All Users" Start Menu. Obviously if you target an "All Users" folder you will need to elevate privileges, since your normal account doesn't have write permissions to it.
This is *no different* to any of the other platforms. If you try to write to a location you don't have write permissions to, you will be prompted to elevate privileges.
Slashdot Mirror
However, it's trivially easy to uninstall Safari if you're inclined to do so. You drag and drop the .app file into the trash.
It's trivially simple to delete iexplore.exe as well.
Break MS up into at least two companies, each with full rights to the source code for Windows, patents, copyright, and trademarks, as well as half the resources (both manpower and money). Forbid these new companies from any collaboration or non-public communication. The desire of the new companies to make money will mean each will start working on a version of Windows that they think OEMs will like more than the other former MS company. OEMs will choose based upon which they think computer buyers will want. Both will be able to bundle anything they want and it won't be a problem because it won't be backed up with the monopoly influence since people can always choose to go with the other former MS company. In short order the market will open up, people will have real choices, and innovation will return.
This is one of the sanest suggestions on the topic I've ever seen.
Everything (shell, graphics system, window manager, etc.) in the kernel. (This is Windows as it is now.)
No, it's not.
Saying OSX is loosly based on Next is like saying Windows 2000 is loosly based on Windows 98... I'd say that the comparison between 2k and XP would be more apt taking Rhapsody, Blue Box and Yellow Box into account.
Windows 2000 (or XP) to Windows Vista would be the best analogy to NeXT -> OS X. The changes in both comparison of are much the same scope, magnitude and type (and the timeframes are roughly equivalent - although OS X has lagged a bit with the lower-level features, that's understandable since the codebase has not been under as intensive development as NT).
(With that said, there are a lot of low-level features that will only appear for the first time in OS X 10.5, that Windows has already had for many years, in particular with regards to efficient use of multiple CPUs.)
OSX was loosly based on NeXT.
Indeed. In the same way Windows Vista is "loosely" based on Windows 2000.
It's kernel is Darwin which is based on NetBSD.
Darwin is "based on" Mach, with a bunch of code welded in from the various BSD projects (mostly FreeBSD).
"Hardware comes with a warranty. Software comes with a disclaimer."
This does not change the fact that there are very real and tangible consequences for a fix to proprietry software that causes extensive breakage (lost customers and, more importantly, revenue), whereas the consequences for the same in the cowboy-esque OSS world have little impact outside the developer's ego.
Your software might not come with a warranty, but if enough people stop paying for it, rest assured that the vendor will take notice and act to try and recover those customers and revenue. If I stop using $J_RANDOM_HACKER's bit of open source code - almost always distributed with the attitude of "it's free, you get what you pay for, if there's a problem fix it yourself" - why is he going to care ?
Fixing Microsoft-created holes is the basic reason why anti-virus firms exist; and why they do such roaring business; and also why they are trusted MORE than Microsoft, which makes the underlying crappy OS.
Anti-virus programs don't "fix holes" in the OS, they fix holes in the *user*.
Outlook licenses are part of Office, not Exchange, which includes a couple of licenses for administration. Exchange Server Client Access Licenses (CALs) and Outlook licenses have to be purchased for each and every user.
An Exchange CAL includes an Outlook (or Entourage) license. This has been true since at least 2003 (and probably well before then).
That is certainly a valid stance. Why does Windows STILL use this lame thing called a Registry, other than for allowing dinosaur software that needs it to run?
Why would they ? It's a centralised, transactional, multiuser-safe database with fine-grained user permissions. The concept is perfectly sound and by most technical measures, vastly superior to alternatives like text files and trees of symlinks.
On the Mac, a quick search for recently installed startup items in two folders will give even a semi-knowledgeable user an idea if some recent download put something there. A quick deletion of these startup items will keep the malware from ever starting again.
Uh huh. And how about when malware burrows its way into the various UNIX init scripts and/or attaches itself to executables like /bin/sh ?
There's just as many places to hide in a UNIX system.
Only with Vista did the marketing team get a hold of the version numbers: NT 6.0
Rubbish. Vista is easily worth a major version increment, well in line with previous examples of same (NT 3.x to NT 4.x, NT 4.x to NT 5.x). The changes from XP to Vista are of the same magnitude as the changes from from OS X 10.0 (more like NeXT/OPENSTEP 4, really) to 10.5.
A major version number is supposed to indicate a break with backwards compatibility.
More rubbish. By whose definition ?
I wouldn't exactly call this 'bashing'. More of a jab. With six version of Vista, MSFT pretty much walked into that punchline.
When (/if) Apple has a sufficiently large market to add more levels of price discrimiation than they currently have, rest assured that they will.
Agreed, but he's also got a point. BMWs (at least the M5) are not built for this kind of stuff. Perhaps a Porsche 911 would've been more appropriate.
Huh ? What *else* is an M5 built for if not a high-speed cross-country road trip ? They're designed for cruising on Autobahns where 100mph is "average".
Speed limits are there for a reason, so stick to them!
Indeed, and that reason typically has nothing to do with safety.
Traffic tickets should be depending on income/wealth instead of being fixed like they are now.
Exceeding the speed limit, in and of itself, should not be an offense (although it should certainly be taken into account in the case of other charges, like dangerous driving).
She commented to me on how fearful the men in the company seemed in their behavior toward women.
If she was talking about interactions with men within the company and especially during work time I can understand their attitude. The risk of a career-ruining sexual-harassment lawsuit should not be underestimated.
May I ask why? Are you saying any large drives in 4 or 5 disk raid-5 arrays somehow present a larger risk of failure than 4 or 5 10GB drives from years ago?
Because raw read error rates suggest that once drive sizes get over 1TB or so, the risk of a double-disk failure increases dramatically. There is also the factor of the sheer amount of time rebuilding an array like that takes and the significant performance and reliability impact that is suffered while the rebuild is occurring.
Even being generous and saying a RAID5 will rebuild at a constant 100MB/sec you're still looking at ca. *40 hours* to rebuild the array if a disk fails. During that rebuild, even a single sector read error on one of the other drives will likely hose the entire thing. Further, current raw read error rates pretty much guarantee (statistically) an unrecoverable read error about every 12TB or so worth of read data. In the context of rebuilding an array of 4TB drives, that's a risk factor _way_ above my comfort level.
Personally, I don't trust RAID5 _now_ and haven't for a couple of years. IMHO the probability of a second drive failure during a rebuild is already too high. That's why I only use RAID6 or RAID10 (and even RAID10 will start getting iffy soon, unless the expected error rates improve dramatically).
Today, we have a Prime Minister, complete with new Cabinet, who has never been voted into that office by the electorate, and indeed who just backed down from holding an election to get a mandate.
I'll admit my understanding of the UK system is a bit sketchy, but I was under the impression the position of Prime Minister was appointed by the Queen (typically the leader of the majority party) and Cabinet was then appointed by the Prime Minister (presumably under advisement from his senior Party members).
Or, in other words, the electorate has _never_ "voted in" any of the people in those positions, because that's simply not how the system works.
(Anyone who is about to bleat about a party political system where the Labour Party was elected would do well to remember that they were elected after Blair said he would serve a full third term. There is no rational way you can argue that the Brown administration has a mandate based on party politics. And even if they had been elected without that promise, our first-past-the-post system is so broken that you could hardly call it representative.)
Well they _do_ (apparently, I don't follow UK politics especially closely) have a mandate. That's the way the system works - people *don't* vote directly for certain individuals to be in specific positions (like Prime Minister).
The system here in Australia is basically the same, which is why I wonder about people who get so uptight about who is (or might be) Prime Minister. The point is that it doesn't matter because a) the Prime Minister could be (literally) anyone in parliament and b) individually, they don't hold any more power than anyone else there. It is little more than a ceremonial position and, ultimately, the Prime Minister's power comes from those he can influence, not inherently from his position as Prime Minister.
Problem: If you suffer random corruption on a disk, ZFS won't know anything about it. The RAID subsystem will have to try to guess which disk has actually failed, and if the RAID gets it right ZFS is happy. If not, ZFS fails a checksum and is stuck because the RAID layer already decided wrong, and probably overwrote the good data with the incorrectly recovered stripe.
You're not supposed to run ZFS on top of a hardware RAID array. That defeats most of the reasons for using ZFS in the first place.
Get a Clariion AX150. It takes SATA drives and can do RAID rather well...
We looked at the AX150i, but it's still ~3x as much as DIY. That's a big difference when all you really want is a huge slab of disk without high uptime requirements (99% availability is fine for us - and most people, IMHO). Also, I don't think the AX150 does RAID6 yet ?
I just happen to know this product, I'm sure other SAN vendors have cheap SATA arrays too.
Relative to its contemporaries, it's cheap - but compared to DIY it's just not in the same league. I do have to wonder why you'd be looking to to buy one for home when you could build something just as good (better, if anything) for a *lot* less money (at least, I'm assuming you don't need dual, redundant controllers for your home server :). Also, don't forget you won't be able to put off-the-shelf drives into it, and will be stuck buying drives from EMC (or off Ebay) at a ~500% markup.
There'd certainly be a bit of "cool" factor, but personally I'd rather have the extra disk space (or several thousand $$$$s worth of something else).
On top of the fact that you slam 16 drives in the bloody thing ( minimum for this kind of data ), and have half as hot spares to a raid6 array.
Holy shit, dude. There's responsible redundancy, then there's paranoia, then there's overkill, then, far off in the distance, there's having half a shelf dedicated to hot spares.
One hot spare per shelf is heaps. Consider a 7*750G RAID6 that suffers a disk failure. An array rebuild will take ca. 20 hours (assuming it's not offlined during the rebuild). Even a cheap SAN is going to come with a 24x7, 4hr response support contract (or, at worst, next day). That is to say, your failed drive will be replaced ca. 16 hours before your array has even finished recovering from the failure (or, worst case, around the same time as it is finishing). Just what is more than a single hot spare going to buy you ? Heck, if you've got staff on-site 24/7 (or equivalent), you don't even need a hot spare - just a cold spare and decent system monitoring.
If you're going to throw away half your raw space, the best way is to use RAID10 or RAID60 and at least get some performance benefits as well.
Sure, you *might* get away with a consumer level raid5/6 solution. But when things come crashing down, and they very likely would, you want something a bit more serious on the back end.
It's not especially hard to build a centralised storage solution with COTS parts that get up to about 90% of the features and performance of enterprise kit at well under half the cost - and despite what EMC and co. will try to tell you, that covers the requirements of the vast majority of environments. *Especially* when you start moving into archival/backup/second-tier systems with much lower reliability and uptime requirements (eg: can get away with single controllers), the cost difference becomes an order of magnitude or more.
Studies are archived daily, with an automated script simply carving an appropriately-sized LV out of the VG, formatting it, copying the data to it, indexing the study metadata, etc. We have about 15TB online (12.5TB allocated) and archive ca. 5000-6000 studies/140-160GB every day (compressed with JPEG lossless).
HOWEVER, our requirements are probably quite different - we keep archived images more for our own convenience, than because of any legal requirements, and only aim for about a 90 day rolling window (ie: the last 90 days worth of studies). Further, the front end unit has enough local disk to go for a solid 3-4 days without needing to archive off (and there is a backup machine if it fails), so while we've never had significant downtime on that particular aspect of our workflow, theoretically most of the physical machines could die for a couple of days with little more than a minor inconvenience.
While I think this is a sound (and dirt cheap - for what the whole thing is worth you couldn't even buy a single 16*500G drive array for our DS4800) solution, if your requirements are more strict - especially from a legal perspective - I would be extremely careful about recommending what is effectively a DIY solution. In particular, if your client ends up on court, they'll be a lot happier if they can go "we have a legally compliant storage system for our data, as certified by $LARGE_VENDOR" rather than "$CONSULTANT designed and built a storage system for us with COTS parts, but it hasn't been certified as compliant for storing patient data". There's also the greater uptime and redundancy of enterprise storage solutions (the part you're actually paying the 10x as much for)[1].
A simple SATA RAID controller interfaced with 4 such drives can give me 12TB of cheap, fast, storage. At 1TB per year, should be good enough for my needs. H/w vendors currently recommend expensive SAN boxes; which I don't like... no useful value for the application at hand.
Using RAID5 with 4TB drives would be insanity. Heck, using RAID5 *at all* with SATA ~160GB+ drives is crazy, IMHO, even for just my home server (16*250G RAID6) - let alone anything business-critical. Maybe if you restrict yourself to small drive counts (6-7), have a hot spare and do regular disk scrubs, it's worth taking a chance on - but I personally wouldn't do it.
[0] DIY jobbies, since no major vendor sells a box that holds so many SATA drives. We stuff 16 drives into a box, RAID6 them all together with Linux's software RAID, then export the RAID arrays with iSCSI Enterprise target. Each disk server has a pair of GbE links for bandwidth and redundancy and cat saturate both of them simultaneously from the RAID array. There used to be 5 servers, but 3 old IDE-based ones were recently replace with a single machine (holding triple the space). We also keep no less than two spare drives for each machine on-site and enough other spare parts (motherboard, CPU, etc) to completely rebuild one from scratch if necessary.
[1] This particular system is one of the few times I have chosen to go DIY over off-the-shelf, simply because the cost savings were so massive and the uptime/reliability requirements were relatively low (only 99% uptime). One thing I did discover researching this, however, was the dearth of low-end storage solutions in the Australian market (hence my decision to DIY).
So you're saying that a modern bloated OS is a better platform for running the new bloated application software, and that throwing a quad-core processor and several GB of RAM at the solution proves this?
No, I'm saying that a modern "bloated" OS takes better advantage of current hardware. Windows Vista will make _vastly_ better use of a quad-core, 4G RAM machine that NT4 ever could. As will Linux 2.6 (vs older versions).
No little bullying insults? It's all there above and a bit is quoted here:
No. Although when you choose to deceptively quote out of context, you can certainly make it look that way.
Try to talk about the subject matter instead of using methods to bully those with less confidence in their abilities or mistaken repect - it is uncivilised.
I am trying to talk about the subject matter. You, on the other hand, repeatedly ignore attempts to do so, preferring to concentrate on a deconstruction of how I choose to make my points.
The 1993 bit could throw the young off track a bit and make them question what they know - was that another little trick?
That depends on whether or not you consider a piece of factual information a "trick" or not.
There's a whole lot in earlier posts about your view of myself and very little about the subject matter.
In fact, my comments about attitude are explicitly conditional on whether or not certain mindsets are held. I'm not calling people with the opinions you have expressed fools for the hell of it, as an insult, I'm calling them fools for holding a biased opinion unsupported by fact or reasoned argument.
You have won this little game you are playing and I have already stated that I do not believe this old simplistic argumant that has shown no sign of occuring over more than a decade - I came here for something other than this little game that is a bit above kindergarden level.
Apparently you're a hypocrite as well. You attack me for not "talking about the subject matter" yet continually evade any attempts to engage upon it. Please note that simply repeating different variations of "you're wrong" is not "talking about the subject matter".
No little bullying insults? It's all there above and a bit is quoted here:
No. Although when you choose to deceptively quote out of context, you can certainly make it look that way.
Try to talk about the subject matter instead of using methods to bully those with less confidence in their abilities or mistaken repect - it is uncivilised.
I am trying to talk about the subject matter. You, on the other hand, repeatedly ignore attempts to do so, preferring to concentrate on a deconstruction of how I choose to make my points. That is to say, essentially nothing more than ad hominem "arguments".
The 1993 bit could throw the young off track a bit and make them question what they know - was that another little trick?
Indeed. Damn those facts, they trickses us all the time.
There's a whole lot in earlier posts about your view of myself and very little about the subject matter.
In fact, my comments about attitude are explicitly conditional on whether or not certain mindsets are held. I'm not calling people with the opinions you have expressed fools for the hell of it, as an insult, I'm calling them fools for holding a biased opinion unsupported by fact or reasoned argument.
You have won this little game you are playing and I have already stated that I do not believe this old simplistic argumant that has shown no sign of occuring over more than a decade - I came here for something other than this little game that is a bit above kindergarden level.
Apparently you're a hypocrite as well. You attack me for not "talking about the subject matter" yet continually evade any attempts to engage upon it. Continually repeating different variations of "I think you're wrong" is not "talking about the subject matter", it's sticking your fingers in your ears and seeing who can yell louder.
Right. Yeah. I must be either lying or I "b0rked" my install. It's not possible that Vista is simply a worthless piece of shit.
Please detail how to reproduce the problem on a stock Vista install.
Count em, 3 dialogs.
This will depend on which folder you're dragging to. The Start Menu displayed is combined from the user's personal Start Menu and the "All Users" Start Menu. Obviously if you target an "All Users" folder you will need to elevate privileges, since your normal account doesn't have write permissions to it.
This is *no different* to any of the other platforms. If you try to write to a location you don't have write permissions to, you will be prompted to elevate privileges.