Something like 85% of the population of the US live in cities. "America's a big place" is not a reason they should have service worse than anywhere in Europe.
What if you don't have 8 months time ? How many deaths could have been prevented by allowing the cost of NHS scans to increase to 1200 pounds ? Thousand ? Ten thousand ? A million ? Nobody knows, but it certainly is a positive number.
Perhaps you can elaborate on why you think a higher price would result in the machine being able to perform scans faster ?
Do you also think that more expensive shoes will make you run faster ?
Today, almost every piece of a car is custom-designed, not shared with other models.
Huh ? Component reuse in vehicles is massive, especially in today's world where only a handful of parent companies control the dozens of different vehicle brands. Everything from the basic chassis, through drivetrains, to switches and trim, is extensively reused across multiple vehicles, often when they're even from different "manufacturers".
I don't know where you picked up the idea that modern cars are all once-offs, but it's completely wrong. There's no way vehicles could be as cheap as they are today if that were true.
Where do I legally download XP SP3 with all the updates, if my laptop came with XP SP1, or if I bought XP SP1 around 8 years ago?
You download SP3 from Microsoft, then either apply it straight after installing (1 reboot), or slipstream it before installing (no reboots).
This constant harping on about reboots as if they're the end of the world just highlights how out of touch the group of people here is. Most people *shut down* their computer at least a few times a week, if not daily.
Why not? There is no reason data integrity and availability can't come from a software solution.
Of course there's not. That's ultimately all the big SAN vendors are selling anyway.
Or am I missing something?
Building a properly redundant storage system with multiple controllers, mirrored cache, adequate throughput, etc, is *hard*. It's not just a matter of slapping together a PC, some disk shelves and installing Linux/FreeBSD/OpenSolaris . There are many reasons other than 500% markups on hardware why big storage systems cost as much as they do.
This is why software based raid is the way to go for ultimate performance. The big SAN providers ought to be shaking in their boots when they look at what's possible using software like starwind or open-e with host-based raid controllers and SSDs.
The big SAN vendors understand that data integrity and availability are typically more important to their customers than raw performance, and hence don't feel they have much to fear from DIYers.
1. updates that frequently fail to install: My friend's laptop constantly complains about needing ~44 critical updates, but any attempt to install them results in an instant "44 updates failed to install". On my laptop, luckily, most updates installed, but I always had 2 or 3 that wouldn't.
And neither I nor anyone I know has ever had it happen.
2. Multiple reboots: Install Windows XP (without any service packs) and try updating. Count the number of reboots. Try a similar test on any Linux distro (pick a CD as old as you want), notice how you only need to reboot once to have all of the updates apply.
Having a ca. 2001 distro update itself automatically to a current distro and actually work, would be an amazing feat on its own.
Still needing to update anything else manually:
Pretty much any application I can think of that I use automatically checks for its own updates.
Start up any arbitrarily old Linux distro and run an update. Notice how everything is up to date (note: By default, most distros don't give you the newest version of most software, but this is intentional and can be worked around by using a distro that doesn't suck).
I would be jaw-droppingly astounded if I could update, say, Red Hat 7.0 or RHEL2.1, or even Ubuntu 4.10 to the latest release of each in a single step and have it work - and that's starting from a clean install. For a machine that had actually been used a bit beforehand with some custom-compiled or out of repo packages, and it would be quite a ride.
After using Linux I don't understand how Windows users put up with the Microsoft updates that frequently fail to install, sometimes require multiple reboots and then still needing to update everything else manually.
Because none of those things are particularly accurate ?
Those are great comments, as they relate to a single such system. If you investigate this particular provider, however, you'll see that they are using systems like this as a building-block in a much larger system, with data replicated on more than one such system. All those problems you noted are mostly solved by this, except perhaps the speed. They may have solved that for the most part with redundancy and smart caching as well. As long as they have geographic redundancy, I'd say it's about as safe as you can get. But I haven't done business with them, so by all means investigate further.
I already knew about the provider, and my comments stand. Their boxes, as they exist today, are incredibly vulnerable to power outages. A blip in a power circuit takes out every single "pod" connected to it, with a non-trivial chance of data corruption occurring, given they're not using ZFS and the sheer time involved in replicating and verifying those sorts of data volumes at the poor performance levels they'd have. I'm not convinced that they're keeping enough redundancy around at higher levels to make up for that, especially when such relatively little additional expense would have added such massive improvements.
Like I said, I wouldn't trust my data to them, based on their description of their technology, and I think there's going to be many cases of their clients either losing their data completely, or at the very least having it corrupted.
Would you mind elaborating on how you'd go about making it fast and reliable for $1K?
A decent motherboard, disk controllers, and multiple PSUs. That would probably cost less than a grand even without subtracting the cost of the components being replaced - and that's at _retail_ prices.
Not that I think it can't be done, I think the solution as it is IS reliable, and medium speed, but if you think it can be done better, I'd be interested to hear how.
It's not even close to reliable. A single power blip will take out the whole box, and it's reasonably likely that in doing so the entire array will be corrupted. Additionally, the non-redundant PSU and single network connection makes it impossible to do many types of scheduled maintenance without an outage. The performance will be dismal because a) it's bottlenecked by a single 32-bit PCI bus (so, for example, RAID rebuilds will proceed at something in the ballpark of 30MB/sec - and that's assuming you aren't trying to use it at the same time), and b) it's bottlenecked by a single 1Gb ethernet controller (so getting data in and out of it tops out at 100MB/sec even in ideal circumstances).
A decent motherboard will have three x8, or at least x4, PCIe slots, and use 8-port SAS controllers with SAS expanders (keeping the SATA port multipliers would also be acceptable, though some performance is sacrificed), dramatically increasing disk bandwidth (and reducing RAID rebuild times and impact). It would also have two to four 1Gb NICs, doubling to quadrupling how fast data can be served and adding redundancy. Multiple PSUs will mean power blips don't bring down the whole box in a data-threatening way.
If you have _massive_ amounts of redundancy built in at the application level, then the backblaze solution might be acceptable (though not even then IMHO - a power blip on a circuit will take out all the "pods" on that circuit, likely a significant chunk of the data at that site). However, when the cost of making the solution so much better is relatively so trivial, not doing so is a great reason not to do business with them, IMHO. When I looked at this the first time, my conclusion was that it was a great big advertisement about why NOT to do business with them - they're sacrificing data integrity (their fundamental product !) to save relatively insignificant amounts of expense.
SAS interconnects are obviously better in this regard, but the hardware doesn't end up being any cheaper than standard fibre channel RAID, [...]
You're either getting massively ripped-off by your SAS vendor, or insanely good deals from your FC vendor.
IME, SAS infrastructure will cost around 1/2 to 2/3 as much as FC RAID. You do sacrifice some flexibility for that (ie: by having to direct connect hosts rather than having a SAN), but if you don't need that flexibility then the price should be dramatically lower.
I can use any search engine I like any time I like. I can switch search engines easier than can change toothpaste.
You can switch OSes trivially as well. Of course, the impact of doing so might be significant, but so is the impact of changing your search engine, if your new one doesn't deliver pertinent results.
When MS pushed IE, you had little choice (since there was --for all practical purposes-- no valid competitor to the OS) about using IE. They forced it down your throats (Considering you can't uninstall it)...
You didn't have to use IE, and Microsoft did nothing to stop you from installing alternatives. Indeed, until IE was competitive (IE3) it was typically only used to download Navigator, then never loaded again. It wasn't until IE was superior (IE4+) that it really started taking market share away from Navigator.
A "natural monopoly" is a monopoly where a successful company will, with no uncompetitive practices, become a monopoly. It appears that search engines make natural monopolies as well, as the more searches that are done, the more valuable the search engine becomes, and a search engine starting out, with no searches, can't be as valuable as the established one, no matter how much they spend or what they do.
They do. Basically, just substitute "application" for "search".
Microsoft, realizing they don't have a natural monopoly, have exploited their monopoly to push other products to expand their market share in other areas, as well as keep their existing monopoly.
This is a non-sequitur. Microsoft did not have a monopoly on "Operating Systems", they had a monopoly on "Operating systems for x86 PCs".
I'd like to see what kind of justification politicians will come up with to argue that corporations can make suggestions, governments can provide input, but god forbid the people actually have a say in the way this sausage is made.
Easy - the people already had their say when they elected said politicians.
That would be the ones with the permanent 10% unemployment rate?
What countries are you thinking of that have permanent 10% unemployment rates ? Because in this list the only country over 10% that stands out is Ireland - and it's been unusually hard hit - and France, at 10%.
We currently have about a 10% unemployment rate here in the United States, and it's considered a catastrophe.
And given the number fiddling that's used to hide the real unemployment rate, that's hardly surprising.
When I lived in Asia, I worked with a German guy, and he was thrilled to be working outside Germany, because at home he'd been in a 50% tax bracket.
There's currently no 50% tax rate in Germany (and I don't believe there has been any time recently). The current top bracket is 42%, which kicks in at a quarter-million Euros per annum.
Chances are he wouldn't be so thrilled if he lost his job and ended up out on the street in $SOME_ASIAN_COUNTRY - though he would have the safe option of simply returning to Germany (and would likely be deported if he list his job), and would thus be insulated from any real concern.
Stop and let that number roll around in your mind for a while. The government taking 50% of your salary off the top. Half of your money that you never even get to see. Kind of makes you wonder why you should bother working.
Because 50% of several hundred thousand is still more than enough to live an extremely comfortable lifestyle. Unimaginably more comfortable that you would be if you were relying on welfare payments (ie: not bothering to work). To say nothing of the services and peace of mind that comes from living in a society with a decent social safety net, where being poor and out of work isn't a life-threatening situation, and where healthcare is something you can count on being there whenever you want it, rather than hoping it's there when you really need it.
And no, this guy wasn't rich. He was a sysadmin, like me, and was earning a middle-class salary.
However, if you argue principle and ideals, recreational drug use is bad.
By what measure ? Driving a car, owning a gun, a poor diet and not enough exercise (to pick some fairly common habits in the US) are far, far more likely to ruin your life than "recreational drug use".
Slashdot Mirror
The United States is BIG.
Something like 85% of the population of the US live in cities. "America's a big place" is not a reason they should have service worse than anywhere in Europe.
What if you don't have 8 months time ? How many deaths could have been prevented by allowing the cost of NHS scans to increase to 1200 pounds ? Thousand ? Ten thousand ? A million ? Nobody knows, but it certainly is a positive number.
Perhaps you can elaborate on why you think a higher price would result in the machine being able to perform scans faster ?
Do you also think that more expensive shoes will make you run faster ?
In societies with socialized healthcare, some government bureaucrat is the person who decides how much to spend on my dying wife.
You really shouldn't talk about things you know nothing about.
Today, almost every piece of a car is custom-designed, not shared with other models.
Huh ? Component reuse in vehicles is massive, especially in today's world where only a handful of parent companies control the dozens of different vehicle brands. Everything from the basic chassis, through drivetrains, to switches and trim, is extensively reused across multiple vehicles, often when they're even from different "manufacturers".
For example, look at how many other vehicles are built off the same platform as the Volkswagen Golf: http://en.wikipedia.org/wiki/Volkswagen_Group_A_platform#A5_.28PQ35.29.
I don't know where you picked up the idea that modern cars are all once-offs, but it's completely wrong. There's no way vehicles could be as cheap as they are today if that were true.
Where do I legally download XP SP3 with all the updates, if my laptop came with XP SP1, or if I bought XP SP1 around 8 years ago?
You download SP3 from Microsoft, then either apply it straight after installing (1 reboot), or slipstream it before installing (no reboots).
This constant harping on about reboots as if they're the end of the world just highlights how out of touch the group of people here is. Most people *shut down* their computer at least a few times a week, if not daily.
Why not? There is no reason data integrity and availability can't come from a software solution.
Of course there's not. That's ultimately all the big SAN vendors are selling anyway.
Or am I missing something?
Building a properly redundant storage system with multiple controllers, mirrored cache, adequate throughput, etc, is *hard*. It's not just a matter of slapping together a PC, some disk shelves and installing Linux/FreeBSD/OpenSolaris . There are many reasons other than 500% markups on hardware why big storage systems cost as much as they do.
This is why software based raid is the way to go for ultimate performance. The big SAN providers ought to be shaking in their boots when they look at what's possible using software like starwind or open-e with host-based raid controllers and SSDs.
The big SAN vendors understand that data integrity and availability are typically more important to their customers than raw performance, and hence don't feel they have much to fear from DIYers.
I wish Windows 7 ran fine on 5 year old hardware. Maybe the OS runs, but once you start running anything on top of the OS it is TERRIBLE.
It does run fine on 5 year old hardware. Heck, with a video card upgrade, it runs quite usably on *10* year old hardware.
Then, you don't work with enough Windows computers.
"Frequently" suggests to me that I should have seen it happen to at least one of the ~20 computers I have personally owned over the last 15 years.
If it doesn't happen to even 5% of machines, it's not "frequent".
1. updates that frequently fail to install: My friend's laptop constantly complains about needing ~44 critical updates, but any attempt to install them results in an instant "44 updates failed to install". On my laptop, luckily, most updates installed, but I always had 2 or 3 that wouldn't.
And neither I nor anyone I know has ever had it happen.
2. Multiple reboots: Install Windows XP (without any service packs) and try updating. Count the number of reboots. Try a similar test on any Linux distro (pick a CD as old as you want), notice how you only need to reboot once to have all of the updates apply.
Having a ca. 2001 distro update itself automatically to a current distro and actually work, would be an amazing feat on its own.
Still needing to update anything else manually:
Pretty much any application I can think of that I use automatically checks for its own updates.
Start up any arbitrarily old Linux distro and run an update. Notice how everything is up to date (note: By default, most distros don't give you the newest version of most software, but this is intentional and can be worked around by using a distro that doesn't suck).
I would be jaw-droppingly astounded if I could update, say, Red Hat 7.0 or RHEL2.1, or even Ubuntu 4.10 to the latest release of each in a single step and have it work - and that's starting from a clean install. For a machine that had actually been used a bit beforehand with some custom-compiled or out of repo packages, and it would be quite a ride.
What is this "shut down" of which you speak?
That's what normal people do with their computers every night.
After using Linux I don't understand how Windows users put up with the Microsoft updates that frequently fail to install, sometimes require multiple reboots and then still needing to update everything else manually.
Because none of those things are particularly accurate ?
Those are great comments, as they relate to a single such system. If you investigate this particular provider, however, you'll see that they are using systems like this as a building-block in a much larger system, with data replicated on more than one such system. All those problems you noted are mostly solved by this, except perhaps the speed. They may have solved that for the most part with redundancy and smart caching as well. As long as they have geographic redundancy, I'd say it's about as safe as you can get. But I haven't done business with them, so by all means investigate further.
I already knew about the provider, and my comments stand. Their boxes, as they exist today, are incredibly vulnerable to power outages. A blip in a power circuit takes out every single "pod" connected to it, with a non-trivial chance of data corruption occurring, given they're not using ZFS and the sheer time involved in replicating and verifying those sorts of data volumes at the poor performance levels they'd have. I'm not convinced that they're keeping enough redundancy around at higher levels to make up for that, especially when such relatively little additional expense would have added such massive improvements.
Like I said, I wouldn't trust my data to them, based on their description of their technology, and I think there's going to be many cases of their clients either losing their data completely, or at the very least having it corrupted.
Would you mind elaborating on how you'd go about making it fast and reliable for $1K?
A decent motherboard, disk controllers, and multiple PSUs. That would probably cost less than a grand even without subtracting the cost of the components being replaced - and that's at _retail_ prices.
Not that I think it can't be done, I think the solution as it is IS reliable, and medium speed, but if you think it can be done better, I'd be interested to hear how.
It's not even close to reliable. A single power blip will take out the whole box, and it's reasonably likely that in doing so the entire array will be corrupted. Additionally, the non-redundant PSU and single network connection makes it impossible to do many types of scheduled maintenance without an outage. The performance will be dismal because a) it's bottlenecked by a single 32-bit PCI bus (so, for example, RAID rebuilds will proceed at something in the ballpark of 30MB/sec - and that's assuming you aren't trying to use it at the same time), and b) it's bottlenecked by a single 1Gb ethernet controller (so getting data in and out of it tops out at 100MB/sec even in ideal circumstances).
A decent motherboard will have three x8, or at least x4, PCIe slots, and use 8-port SAS controllers with SAS expanders (keeping the SATA port multipliers would also be acceptable, though some performance is sacrificed), dramatically increasing disk bandwidth (and reducing RAID rebuild times and impact). It would also have two to four 1Gb NICs, doubling to quadrupling how fast data can be served and adding redundancy. Multiple PSUs will mean power blips don't bring down the whole box in a data-threatening way.
If you have _massive_ amounts of redundancy built in at the application level, then the backblaze solution might be acceptable (though not even then IMHO - a power blip on a circuit will take out all the "pods" on that circuit, likely a significant chunk of the data at that site). However, when the cost of making the solution so much better is relatively so trivial, not doing so is a great reason not to do business with them, IMHO. When I looked at this the first time, my conclusion was that it was a great big advertisement about why NOT to do business with them - they're sacrificing data integrity (their fundamental product !) to save relatively insignificant amounts of expense.
SAS interconnects are obviously better in this regard, but the hardware doesn't end up being any cheaper than standard fibre channel RAID, [...]
You're either getting massively ripped-off by your SAS vendor, or insanely good deals from your FC vendor.
IME, SAS infrastructure will cost around 1/2 to 2/3 as much as FC RAID. You do sacrifice some flexibility for that (ie: by having to direct connect hosts rather than having a SAN), but if you don't need that flexibility then the price should be dramatically lower.
Here's a compelling petabyte online RAID system for cheap:
Cheap, slow and unreliable.
The good news is you'd probably only need to spend another $500-$1k to make it fast and reliable (but less cheap).
That's one thing everyone seems to be forgetting.
How could it be forgotten when it gets brough up every single time SSDs are discussed ?
And, of course, every single time the same point is made that wear-levelling solved the problem years ago.
Holding a monopoly position isn't bad at all. USING that monopoly position to impact the market is what's bad.
It's pretty much impossible to hold a monopoly without "abusing" it.
I can use any search engine I like any time I like. I can switch search engines easier than can change toothpaste.
You can switch OSes trivially as well. Of course, the impact of doing so might be significant, but so is the impact of changing your search engine, if your new one doesn't deliver pertinent results.
When MS pushed IE, you had little choice (since there was --for all practical purposes-- no valid competitor to the OS) about using IE. They forced it down your throats (Considering you can't uninstall it)...
You didn't have to use IE, and Microsoft did nothing to stop you from installing alternatives. Indeed, until IE was competitive (IE3) it was typically only used to download Navigator, then never loaded again. It wasn't until IE was superior (IE4+) that it really started taking market share away from Navigator.
However, OSs aren't that way.
By this logic:
A "natural monopoly" is a monopoly where a successful company will, with no uncompetitive practices, become a monopoly. It appears that search engines make natural monopolies as well, as the more searches that are done, the more valuable the search engine becomes, and a search engine starting out, with no searches, can't be as valuable as the established one, no matter how much they spend or what they do.
They do. Basically, just substitute "application" for "search".
Microsoft, realizing they don't have a natural monopoly, have exploited their monopoly to push other products to expand their market share in other areas, as well as keep their existing monopoly.
This is a non-sequitur. Microsoft did not have a monopoly on "Operating Systems", they had a monopoly on "Operating systems for x86 PCs".
I'd like to see what kind of justification politicians will come up with to argue that corporations can make suggestions, governments can provide input, but god forbid the people actually have a say in the way this sausage is made.
Easy - the people already had their say when they elected said politicians.
That would be the ones with the permanent 10% unemployment rate?
What countries are you thinking of that have permanent 10% unemployment rates ? Because in this list the only country over 10% that stands out is Ireland - and it's been unusually hard hit - and France, at 10%.
We currently have about a 10% unemployment rate here in the United States, and it's considered a catastrophe.
And given the number fiddling that's used to hide the real unemployment rate, that's hardly surprising.
When I lived in Asia, I worked with a German guy, and he was thrilled to be working outside Germany, because at home he'd been in a 50% tax bracket.
There's currently no 50% tax rate in Germany (and I don't believe there has been any time recently). The current top bracket is 42%, which kicks in at a quarter-million Euros per annum.
Chances are he wouldn't be so thrilled if he lost his job and ended up out on the street in $SOME_ASIAN_COUNTRY - though he would have the safe option of simply returning to Germany (and would likely be deported if he list his job), and would thus be insulated from any real concern.
Stop and let that number roll around in your mind for a while. The government taking 50% of your salary off the top. Half of your money that you never even get to see. Kind of makes you wonder why you should bother working.
Because 50% of several hundred thousand is still more than enough to live an extremely comfortable lifestyle. Unimaginably more comfortable that you would be if you were relying on welfare payments (ie: not bothering to work). To say nothing of the services and peace of mind that comes from living in a society with a decent social safety net, where being poor and out of work isn't a life-threatening situation, and where healthcare is something you can count on being there whenever you want it, rather than hoping it's there when you really need it.
And no, this guy wasn't rich. He was a sysadmin, like me, and was earning a middle-class salary.
Then he was lying.
However, if you argue principle and ideals, recreational drug use is bad.
By what measure ? Driving a car, owning a gun, a poor diet and not enough exercise (to pick some fairly common habits in the US) are far, far more likely to ruin your life than "recreational drug use".
If you want prosperity, low taxes are an essential component. That's so basic I can't believe there are people who don't understand it.
They're probably confused by all those high-taxing countries that are quite prosperous - like, say, a fairly large chunk of the civilised world.