once again uncommon outside of the MS Windows environment.
This isn't at all true. Proprietary or one-off formats are extremely common in my experience, *especially* on older non-windows platforms. There were no standard file formats for most data, so vendors invented their own. We have 20 years worth of financials copied from an old AS400 server that is impossible to decipher. The original application and vendor are long gone.
Once you move beyond text and numbers, there are basically no standards at all for older systems. We've had a lot of trouble with old CAD files from the 1980s. The de-facto AutoCAD DWG format had dozens of versions and still isn't open, and there was no credible alternative CAD format for about two decades.
Unless data was intentionally converted to some open or self-documenting format for archival purposes way back when, it is likely in an application specific format with little or no documentation. It doesn't matter at all if the originating system was mainframe, mini, UNIX, Apple II, or other pre-Windows system. Programmers and vendors have always assumed their applications would exist forever.
Umm, no. It's not textual data we're worried about. Sure, anything in ASCII and maybe even HTML might be readable. But what about image formats? Video? MySQL database files? CVS or Subversion databases? Random Wiki formats?
It can be difficult to read "popular" and open file formats from 10 years ago, much less 100. If they were created by an open source program, you might be able to port it forward, but what if there is no C compiler on your 2050 platform?
Maintaining an active archive with format conversion every N years is a lot less risky, and probably a lot more cost effective.
Far fewer than all the other brands, I'd imagine. We've tried just about everything over the last 15 years, and always come back to Seagate for reliability.
Of course, we don't buy server drives from anybody but our SAN vendor these days. "Rolling your own" storage just isn't possible anymore when you need high-end SAN featuers. Crack open one of their hot-swap carriers, though, and it says Seagate on the mechanism. I imagine our SAN vendor, which sells millions of drives, has taken a hard look at reliability numbers. They still choose Seagate (the cost of an RMA is very high for any vendor).
...for $20 instead of getting 20 minutes of long distance...
There, fixed that for ya.
For all the poeple who bitch about the "massive" profits telcos and ISPs "get for free", please read a balance sheet once in a while. Telco margins are razor-thin compared to most industries, and for example long distance prices have dropped by more than 10x in the last few years. It's a totally commoditized market, especiialy in big cities. It's actually tough to consistently make money as a telco or even a non-incumbent ISP.
I never used the 9x series much; we went from 3.1 to NT 3.51 back in the mid 1990s.
But 2000->XP was completely seamless on almost every machine; we deployed that upgrade via scripts. The only two XP->Vista upgrades I've done required only an update of AV software, and the DVD player software on my laptop. No messing with drivers at all.
Granted, we now usually do re-images for most end user machines, but since 2000 the upgrade process in Windows on the Workstation and Server products has been impressively robust IMHO.
A workaround to this is to make a manifest of all the rpm's installed prior to upgrading to a new release, then making a manifest of the new rpm's, then doing a diff and installing the missing rpm's.
You're saying the above is better than having a Windows machine slow down once a week during a virus scan? You are one seriously brainwashed fanboi, my brother.
they are making immense profits off those boxes at taxpayer expense.
There, fixed that for ya. It amazes my how many people think the "government" has all this money and is some separate entity. Like, it's okay to sue the local township for $1M because you wore high heels in January and slipped on the ice. After all, you're only screwing "the government" out of $1M. In the end, you're truly screwing your friends and neighbors.
Three different times Apple's jumped platform and each time they had less backwards compatibility problems than XP to Vista has
There are something like 0.0001% of the applications, 3rd party hardware, and APIs to code for and QA on the Mac platform. You're comparing plugging a drain with building the Hoover Dam.
In any case, Apple sucks at this sort of thing, too. The 68k to PPC transition was abosultely horrible for about 12-18 months. I was IT manager at a newspaper then, and or shiny new PowerMacs had to be restarted almost hourly for a few months.
Similar issues with OS X came out. Almost none of the required utilites for publishing work (Suitcase, scanner drivers, some Adobe stuff) worked on OS X at all, even in emulation mode. We waited more than a year before upgrading.
That's how Sweden, Japan and South Korea did it. They gave a shovel and a hundred yards of fiber to everybody.
First, that's not how they did it. Do some research on the subject and you'll understand why those schemes could never work in the USA (our population is far less urban, and we actually value individual property rights). Sweden only has 30% broadband penetration, so I cannot figure out why it is held up as some sort of shining example by the Slashdot crowd.
Don't be an idiot. It doesn't matter who does it as long as it's done. Companies, government, as long as joe schmoe gets his cheap fast fiber.
Yes, it really does matter how it's done. If it's not done economically, it will generate crippling debt. And why is cheap, fast fiber such a priority? If you ask most Americans, there are quite a few items higher up on the TODO list. If you want it, you pay for it. Don't be so cavalier with other people's money.
$1500 per household for deployment x 100 million US residences = ~150 BILLION DOLLARS in captial expenditure.
Pre-emtive "insightful" comment from pinko college kid living of Daddy's money: "That's way less than we've spent in Iraq..."
...which only illustrates this: If we ask the governement to build out this network, it will cost about $1 Trillion with all the overage, delays, and waste. I'd rather take over another oil-rich country for that sort of investment.
I have a shovel you can borrow. Call me when you're done.
Oh, you're not going to do it yourself? Then who is going to install and maintain all that fiber? Surely you don't want the government to do it? The government (local, state, or federal) can't pave fucking roads without massive waste and lots of outright fraud, and you want them to run a high-tech infrastructure?
Maybe we can have convicted hackers and meth dealers build and maintain it as part of their community service.
Actually, the Windows Kernel is much less monolithic than Linux, and very well designed and stable. The Windows kernel is not the problem. It's the fact that you can only get the Windows Kernel with "everything else that is windows" on top.
Microsoft is exploring what the market really wants. The new "server core" role in Windows Server 2008 is a step in this direction. No GUI, minimal footprint. And some crippling limitations, some of which are techinical, and some of which were created by MSFT marketing.
We have already seen a huge shift away from laptops as the mobile computer towards dedicated devices like the Blackberry and smartphone.
Those mobile devices have serious compromises that make them all but unusuable for anything but email very simple browsing. Anything that requires a lot of screen real estate or a lot of data entry and futzing (almost most business applications) just worn't work on a small form factor device.
Just as significantly, pervasive wireless bandwidth just isn't there. I have the fastest "mobile broadband" available in Chicago, and it still is absolutely painful to visit some sites or work remotely via VPN.
The desktop is being replaced by the laptop, but I seriously doubt the laptop is going anywhere soon. Market share reports indicate the laptop is where its at.
Anyway...I've never seen a dangerous pothole on any road with a speed limit of >55. They're eliminated as soon as they happen. Obviously, I can't speak for everywhere...but this has been my experience traveling in Florida, Georgia, Tennessee, Illinois, and Indiana...the primary states I go through in a car.
You weren't in Chicago mich this winter, were yyou? The roads are in terrible shape. One news channel watched a 8x10 ft pothole on I94 go unattended for two weeks, and counted over 200 cars with severe tire damage or worse from said pothole.
Conversely, the private I90 Skyway into Indiana was as smooth as a baby's butt all winter. Of course it costs $3 just to cross the skyway bridge, plus mroe tolls on the Indiana side.
To paraphrase PJ O'Rouke: "To call something 'public' is to define it as dirty, insufficient, and hazardous."
Dear Canada, Give us the oil, or we will stop all trade with your country. Your eceonomy is entirely dependent upon us ugly Americans, while we sould live without you for quite a while. It would hurt, but not destroy us. You would be bankrupt withing months. Even your professional hockey teams require American dollars to operate!
So you're doing regular full database backups as well as mailbox-level backups? That's fine, I suppose, so long as you have the backup window and tape/disk media space to spare.
I still suggest testing a full restore of Exchange regularly, as well as mailbox-level test restores.
Then you don't have the hard drive space to completely restore your backups either!
When you restore from a mailbox-level backup, single-instance-stoage (SiS) is broken. Every message you restore is inserted into the Exchange database separately. At my site, our SIS ratio (which can be seen using perfmon) is 5.6, meaning each email message is sent to 4.6 people on average (plus one copy in sent items), but only stored once.
Assuming your site's ratio is similar, if you had to restore all of your mailboxes using a mailbox-level backup, you would need 5.6 times as much disk space as you currently use for Exchgange databases!
Have you ever done a full test restore? If not, you're in for some rude surprises.
Disk space is CHEAP. Data loss and down time are not. We use a deleted items retention period of 180 days, and it only increases our Exchange database sizes by about 30%
Considering that 20 percent of the U.S. corn crop was converted into 5 billion gallons of ethanol in 2006, (and that amount replaced only 1 percent of U.S. oil consumption).
Source? Almost all gasoline is actaully 10% ethanol these days. Since gasoline accounts for 60% of oil consumption, wouldn't it stand to reason that ethanol replaces about 6% of our oil consumption at this point?
Finally, after processing corn for Ethanol, a great deal of high-protien livestock feed remains. The sugars from the corn get converted to ethanol, and the "everything else" is still used as livestock feed.
It's really a lot more complicated than you make it sound. Corn-based Ethanol will not solve our transportation energy needs, but it isn't all bad.
In a properly managed Exchange system, there's almost no reason to want mailbox-level backup. Deleted items and deleted mailbox retention cover almost all of the situations where you would want a mailbox-level backup, without requiring any type of restore or 3rd party software. In the very infrequent case of needing really old data not covered by deleted items retention, you can restore a whole database using a recovery storage group and move items indiviudally from there.
Yes, many tools do mailbox-level backups, but they are really inefficient, and quite brittle and buggy. Have you tested a restore this way?
Also, there is no way to do reliable incremental backups with per-mailbox tools. There is no Microsoft API for individual mailbox backup, so these tools impersonate an Outlook client (using MAPI) to suck down messages. Finally, a real exchange backup does a check of all database pages for corruption during the backup. You don't get that very important feature with mailbox-levelbackups.
You don't want to backup or restore individual mailboxes. That's what the deleted items/mailbox retention feature and recovery storage groups are for. Using those features, you don't need to restore individual mailboxes unless you need to roll back by many months.
Backing up individual mailboxes is horribly inefficient because it breaks single-instance storage, and requires an index seek for each item being backed up. We did a test once, and our 60 GB exchange databases exploded to >200 GB with a mailbox-level backup, and took 6 times longer, hammering the crap out of the disk and CPU in the Exchange box in the process.
Think of it this way... would you back up each row in a SQL database table separately with individual queries? No. But that's what you're asking Exchange to do with mailbox-level backups.
What a bunch of incompetents. 12 day old backups for one, and a complete inability to open task manager to discover which process was doing all the disk writes.
If it was exchange itself, a mail loop was probably the cause. Older Exchange versions didn't totally prevent users from creating ping-pong forwarding rules with certain external mail systems. Again, a few minutes with simple tools like perfmon would have diagnosed the issue.
Give the same "admins" a Linux box and the same amount of training they had for Windows/Exchange, and the damage would inevitably be just as bad.
Management is the key in any IT system - we run Exchange and other Windows-based stuff with four-nines reliability. It is possible, and not even that hard with good controls and processes. The same can be said of just about any modern OS, and even most application stacks (although there is some truly horrible, unworkable crap out there in the vertical markets).
Slashdot Mirror
This isn't at all true. Proprietary or one-off formats are extremely common in my experience, *especially* on older non-windows platforms. There were no standard file formats for most data, so vendors invented their own. We have 20 years worth of financials copied from an old AS400 server that is impossible to decipher. The original application and vendor are long gone.
Once you move beyond text and numbers, there are basically no standards at all for older systems. We've had a lot of trouble with old CAD files from the 1980s. The de-facto AutoCAD DWG format had dozens of versions and still isn't open, and there was no credible alternative CAD format for about two decades.
Unless data was intentionally converted to some open or self-documenting format for archival purposes way back when, it is likely in an application specific format with little or no documentation. It doesn't matter at all if the originating system was mainframe, mini, UNIX, Apple II, or other pre-Windows system. Programmers and vendors have always assumed their applications would exist forever.
Speed of light in fiber = 2x108 m/s. So a 1000 km fiber link running at 10 Gb/s can "store" only 5 MB of in-flight data using 8/10 encoding.
So you would need about 20 Million km of 10 Gb/s fiber backbone just to "store" 1 TB. Not terribly practical, I'm afraid.
Uless you're counting on the storage of router queues. But you have to remember that routers drop packets randomly in the face of congestion.
Umm, no. It's not textual data we're worried about. Sure, anything in ASCII and maybe even HTML might be readable. But what about image formats? Video? MySQL database files? CVS or Subversion databases? Random Wiki formats?
It can be difficult to read "popular" and open file formats from 10 years ago, much less 100. If they were created by an open source program, you might be able to port it forward, but what if there is no C compiler on your 2050 platform?
Maintaining an active archive with format conversion every N years is a lot less risky, and probably a lot more cost effective.
Far fewer than all the other brands, I'd imagine. We've tried just about everything over the last 15 years, and always come back to Seagate for reliability.
Of course, we don't buy server drives from anybody but our SAN vendor these days. "Rolling your own" storage just isn't possible anymore when you need high-end SAN featuers. Crack open one of their hot-swap carriers, though, and it says Seagate on the mechanism. I imagine our SAN vendor, which sells millions of drives, has taken a hard look at reliability numbers. They still choose Seagate (the cost of an RMA is very high for any vendor).
There, fixed that for ya.
For all the poeple who bitch about the "massive" profits telcos and ISPs "get for free", please read a balance sheet once in a while. Telco margins are razor-thin compared to most industries, and for example long distance prices have dropped by more than 10x in the last few years. It's a totally commoditized market, especiialy in big cities. It's actually tough to consistently make money as a telco or even a non-incumbent ISP.
I never used the 9x series much; we went from 3.1 to NT 3.51 back in the mid 1990s.
But 2000->XP was completely seamless on almost every machine; we deployed that upgrade via scripts. The only two XP->Vista upgrades I've done required only an update of AV software, and the DVD player software on my laptop. No messing with drivers at all.
Granted, we now usually do re-images for most end user machines, but since 2000 the upgrade process in Windows on the Workstation and Server products has been impressively robust IMHO.
Except you don't have to do reinstall anyting on Windows, either. Maybe AV software if there is an incompatibility, but not much else.
You're saying the above is better than having a Windows machine slow down once a week during a virus scan? You are one seriously brainwashed fanboi, my brother.
There, fixed that for ya. It amazes my how many people think the "government" has all this money and is some separate entity. Like, it's okay to sue the local township for $1M because you wore high heels in January and slipped on the ice. After all, you're only screwing "the government" out of $1M. In the end, you're truly screwing your friends and neighbors.
Breakthrough Overland Optical Network Delivering Optimum Gigabit Grade Local Ethernet.
Call it BOONDOGGLE for short.
There are something like 0.0001% of the applications, 3rd party hardware, and APIs to code for and QA on the Mac platform. You're comparing plugging a drain with building the Hoover Dam.
In any case, Apple sucks at this sort of thing, too. The 68k to PPC transition was abosultely horrible for about 12-18 months. I was IT manager at a newspaper then, and or shiny new PowerMacs had to be restarted almost hourly for a few months.
Similar issues with OS X came out. Almost none of the required utilites for publishing work (Suitcase, scanner drivers, some Adobe stuff) worked on OS X at all, even in emulation mode. We waited more than a year before upgrading.
First, that's not how they did it. Do some research on the subject and you'll understand why those schemes could never work in the USA (our population is far less urban, and we actually value individual property rights). Sweden only has 30% broadband penetration, so I cannot figure out why it is held up as some sort of shining example by the Slashdot crowd.
Yes, it really does matter how it's done. If it's not done economically, it will generate crippling debt. And why is cheap, fast fiber such a priority? If you ask most Americans, there are quite a few items higher up on the TODO list. If you want it, you pay for it. Don't be so cavalier with other people's money.
$1500 per household for deployment x 100 million US residences = ~150 BILLION DOLLARS in captial expenditure.
Pre-emtive "insightful" comment from pinko college kid living of Daddy's money: "That's way less than we've spent in Iraq..."
...which only illustrates this: If we ask the governement to build out this network, it will cost about $1 Trillion with all the overage, delays, and waste. I'd rather take over another oil-rich country for that sort of investment.
I have a shovel you can borrow. Call me when you're done.
Oh, you're not going to do it yourself? Then who is going to install and maintain all that fiber? Surely you don't want the government to do it? The government (local, state, or federal) can't pave fucking roads without massive waste and lots of outright fraud, and you want them to run a high-tech infrastructure?
Maybe we can have convicted hackers and meth dealers build and maintain it as part of their community service.
Actually, the Windows Kernel is much less monolithic than Linux, and very well designed and stable. The Windows kernel is not the problem. It's the fact that you can only get the Windows Kernel with "everything else that is windows" on top.
Microsoft is exploring what the market really wants. The new "server core" role in Windows Server 2008 is a step in this direction. No GUI, minimal footprint. And some crippling limitations, some of which are techinical, and some of which were created by MSFT marketing.
Those mobile devices have serious compromises that make them all but unusuable for anything but email very simple browsing. Anything that requires a lot of screen real estate or a lot of data entry and futzing (almost most business applications) just worn't work on a small form factor device.
Just as significantly, pervasive wireless bandwidth just isn't there. I have the fastest "mobile broadband" available in Chicago, and it still is absolutely painful to visit some sites or work remotely via VPN.
The desktop is being replaced by the laptop, but I seriously doubt the laptop is going anywhere soon. Market share reports indicate the laptop is where its at.
You weren't in Chicago mich this winter, were yyou? The roads are in terrible shape. One news channel watched a 8x10 ft pothole on I94 go unattended for two weeks, and counted over 200 cars with severe tire damage or worse from said pothole.
Conversely, the private I90 Skyway into Indiana was as smooth as a baby's butt all winter. Of course it costs $3 just to cross the skyway bridge, plus mroe tolls on the Indiana side.
To paraphrase PJ O'Rouke: "To call something 'public' is to define it as dirty, insufficient, and hazardous."
Dear Canada,
Give us the oil, or we will stop all trade with your country. Your eceonomy is entirely dependent upon us ugly Americans, while we sould live without you for quite a while. It would hurt, but not destroy us. You would be bankrupt withing months.
Even your professional hockey teams require American dollars to operate!
The yellow sticker on every gas pump I've seen in the last 10 years. It reads "contains 10% Ethanol by volume".
So you're doing regular full database backups as well as mailbox-level backups? That's fine, I suppose, so long as you have the backup window and tape/disk media space to spare.
I still suggest testing a full restore of Exchange regularly, as well as mailbox-level test restores.
Then you don't have the hard drive space to completely restore your backups either!
When you restore from a mailbox-level backup, single-instance-stoage (SiS) is broken. Every message you restore is inserted into the Exchange database separately. At my site, our SIS ratio (which can be seen using perfmon) is 5.6, meaning each email message is sent to 4.6 people on average (plus one copy in sent items), but only stored once.
Assuming your site's ratio is similar, if you had to restore all of your mailboxes using a mailbox-level backup, you would need 5.6 times as much disk space as you currently use for Exchgange databases!
Have you ever done a full test restore? If not, you're in for some rude surprises.
Disk space is CHEAP. Data loss and down time are not. We use a deleted items retention period of 180 days, and it only increases our Exchange database sizes by about 30%
Source? Almost all gasoline is actaully 10% ethanol these days. Since gasoline accounts for 60% of oil consumption, wouldn't it stand to reason that ethanol replaces about 6% of our oil consumption at this point?
Finally, after processing corn for Ethanol, a great deal of high-protien livestock feed remains. The sugars from the corn get converted to ethanol, and the "everything else" is still used as livestock feed.
It's really a lot more complicated than you make it sound. Corn-based Ethanol will not solve our transportation energy needs, but it isn't all bad.
In a properly managed Exchange system, there's almost no reason to want mailbox-level backup. Deleted items and deleted mailbox retention cover almost all of the situations where you would want a mailbox-level backup, without requiring any type of restore or 3rd party software. In the very infrequent case of needing really old data not covered by deleted items retention, you can restore a whole database using a recovery storage group and move items indiviudally from there.
Yes, many tools do mailbox-level backups, but they are really inefficient, and quite brittle and buggy. Have you tested a restore this way?
Also, there is no way to do reliable incremental backups with per-mailbox tools. There is no Microsoft API for individual mailbox backup, so these tools impersonate an Outlook client (using MAPI) to suck down messages. Finally, a real exchange backup does a check of all database pages for corruption during the backup. You don't get that very important feature with mailbox-levelbackups.
See this link for some more details.
You don't want to backup or restore individual mailboxes. That's what the deleted items/mailbox retention feature and recovery storage groups are for. Using those features, you don't need to restore individual mailboxes unless you need to roll back by many months.
Backing up individual mailboxes is horribly inefficient because it breaks single-instance storage, and requires an index seek for each item being backed up. We did a test once, and our 60 GB exchange databases exploded to >200 GB with a mailbox-level backup, and took 6 times longer, hammering the crap out of the disk and CPU in the Exchange box in the process.
Think of it this way... would you back up each row in a SQL database table separately with individual queries? No. But that's what you're asking Exchange to do with mailbox-level backups.
What a bunch of incompetents. 12 day old backups for one, and a complete inability to open task manager to discover which process was doing all the disk writes.
If it was exchange itself, a mail loop was probably the cause. Older Exchange versions didn't totally prevent users from creating ping-pong forwarding rules with certain external mail systems. Again, a few minutes with simple tools like perfmon would have diagnosed the issue.
Give the same "admins" a Linux box and the same amount of training they had for Windows/Exchange, and the damage would inevitably be just as bad.
Management is the key in any IT system - we run Exchange and other Windows-based stuff with four-nines reliability. It is possible, and not even that hard with good controls and processes. The same can be said of just about any modern OS, and even most application stacks (although there is some truly horrible, unworkable crap out there in the vertical markets).