I could see Google using it to protect their account signup/login/new service signup processes, not their search function.
What's more valuable is other people using reCaptcha technology. Google can now benefit from their use, by using the service to assist their book scanning/OCR efforts.
3 arrays? Why? Do it in software, do RAID 5. How likely is it that you'll have a bad sector, or a miswrite, that hits both the stripes and the parity?.
It doesn't matter. Just one of them has to have been miswritten. With RAID5, you have no way of telling which one is correct and which one is wrong.
Parity = STRIPE D1 (XOR) STRIPE D2
You can customize your RAID implementation to read all 3 disks... it will find the bit where the bit in the parity stripe is not equal to STRIPE1_BITS (XOR) STRIPE2_BITS. It can know that the bit read is inconsistent, there must be the error.... the problem is RAID5 won't know which disk has the error..
How do you know which stripe is right and which stripe is wrong?
The answer is, you don't know unless you have double parity (RAID6), which would further impact performance, and reduce the amount of storage you get per $$$.
Using software RAID5/6 in such a scenario is even worse than RAID1 in a 3-way mirror config, due to the RAID write hole it introduces, you lose terribly, if, say the power fails. Using PCIe controllers with hardware RAID would massively increase the cost of the box.
It's not likely the same bit error occurs, but RAID5 cannot correct these errors for you at all, or RAID6 cannot correct these errors for you (once a single drive has failed). When using consumer drives, the probability of an eventual drive failure in your RAID5/6 array is high, and chances are it happens before you need your data. Different drives can have different bits in error. And when one of those drives fails (before you can detect any error), the redundancy is no longer there to even try. When you rebuild the array, the bits will still be in error.
That's why it's called silent data corruption
RAID5/6 implementations do not read bits from all 3/4 stripes when a read is requested, that would be slow, even slower than it is due to the XOR computation against 2 stripes; they distribute their reads.
And look how massively the Backblaze unit has oversubscribed SATA channels with all these multipliers, software RAID rebuild performance will be quite reduced VS more common server configurations.
To be able to verify in software, you'd have to break the RAID abstraction and somehow read bits from each underlying disk, perform both XORs, compare them, alert if they are different.
And somehow pull those bits from the other box that doesn't detect an error.
Well, Linux software RAID doesn't offer a way to do this... and it is a layering violation, also, kernel developers are not likely to add a feature such as that, as a result.
Or use ZFS. It'll checksum everything, so yes, it'll know which array has the right value.
Yes, a very good answer to all silent data corruption issues, unfortunately, their hardware uses port multipliers, which don't exist in a non-development version of OpenSolaris yet. Maybe in 2010....
Deduplication can help you reduce the size requirements on the backup server.
If buying new capacity, you should probably think about buying a backup server that can be expanded to have more capacity than your existing server, depending on current server usage.
Plan for a few years down the road, when it becomes necessary to expand capacity of the main server, backup more servers... or more likely: store multiple old versions of files that changed over time.
Normally.. if you have a 500 mb video file, and someone made some edits to it and re-saved.
There are now going to be two 'files' in the backup repository for a time: the old version and the new version with the edits (twice the space usage)
So storage requirements on the backup server can actually be much more than storage requirements on the server being backed up.
If it's static, you could spend a few hundred $$ and buy 100 Blu-Ray Recordable disks; burn the 12 TB in sets of 50, and vault them at an off-site location.
get a small hosted account somewhere that you can continuously rsync the 30gb or so between backup sets.
When enough new video has been added to fill a BD-R, burn two copies of it, move the copies on the rsync server to a 'to be deleted directory', vault the BD-Rs off site, and then after that, delete the files off the rsync server.
The hard drives are desktop class, not designed for 24x7 operation. Not designed for massive write traffic that server backups generates.
Latent defects on disks are a real concern.
You write your data to a disk, but there's a bad sector, or miswrite, and when you go back later (perhaps when you need the backup), there are errors on the data you are reading from the disk.
Moreover, you have no way of detecting it, or deciding which array has recorded the "right value" for that bit...
That is, unless every bit has been copied to 3 arrays.
And every time you read data, you compare all 3.
(Or that you have two copies and a checksum)
Well, the complexity of this redundancy reduces the reliability overall, and it has a cost.
Maybe, but you will need at least 3 of them and some pretty smart software to have reasonable reliability. The cheapness of backblaze's pods comes from cutting a lot of corners in the hardware.
If you look at the design of Backblaze, this is not server grade equipment: they have two power supplies, but they are not redundant, and these are desktop power supplies not designed to be operated 24/7/365.
If PSU B goes out, a large number of your disk drives and fans go offline; so you lose data and maybe burn up.
And PSU A powers the mainboard...
The disks used by Backblaze are Seagate 7200.11.
These are desktop drives, not designed for 24x7 operation like in a backup server, MTBF stats are based on 1200 power-on hours per year.
Hello... silent data corruption. The last thing you need is for the time to come when you want to restore your backup, and you find latent defects on some of your disks means there are errors in the bits that got saved.
The hard stuff is multi-touch. And pen computing.
Also, there's no point in apps trying to display custom mouse pointers or cursors.
So for most purposes, they could abstract and mostly make it look just like a mouse.. one "tap" generates an event as if you had clicked in that spot, two taps generates an event as if you had double clicked.
Putting your finger to the screen and moving around = mouse movement.
For click and drag, may want pressure sensitivity, a button, two fingers, or something...
Well, it's distinctive from other software apps, but it's not in general a distinctive interface.
Everyone with a physical bookshelf at home has seen Books and DVDs sitting in a bookshelf. That essentially means the UI isn't "novel", as in he didn't create a new type of interface.
He took something everyone's seen before and made his UI look just like it.
And his app's not the first to use that metaphor either. But it's probably the first that actually used that metaphor, did something useful, and didn't suck.
So, uh, the wood grain is very possibly the best thing he's got to complain about....
A right to freedom from arbitrary restrictions. That is, by law it will be illegal for a manufacture of any device or software program to impose artificial restrictions whose purpose is to limit the usefulness of a product and are not necessary to ensure proper operation, for example, "This unit is programmed to fail after 10 uses (even though it should work for 10000)"; "You can only change out your RAM 3 times in a month, and then your product becomes deactivated", OR.. "This program will only utilize 3 CPUs, if you want to run it on a 4 CPU system, you must contact us and ask to buy the super-deluxe version". These types of restrictions should be illegal.
A right to my content -- DVD, HD-DVD, BluRay, etc; manufacturers will be required by law to publish encryption keys, so I can access the data i've bought.
A right to jail break -- Apple and all cell phone makers will be required by law to publish all information required to defeat 'carrier locking, and A general right to re-enable any carrier disabled features of a device, such as tethering.
A right to execute software of my choice. Device makers will be required by law to publish any encryption keys required for me to prepare and run any app on my devices, as well as any encryption keys used to "sign" OS images.
I noticed that.. but then again, a lot of us are still using SAS HBAs that only support 3gb/s (or ~384 megabytes per second).
Which is (I suppose) a limitation of what will soon be considered older SAS hardware.
Still, SAS devices aren't bound to SATA 1.5 gigabit limitations... in fact, even the SATA II next gen spec can also reach 3gb/s, and SATA III will be 6gb/s also.
And 6gb/s (768 megabytes/sec) is definitely more than 500 MB/s:)
Maybe nobody would buy them at 15x, but just b/c the improvement isn't worth it to them.
A new device with an N fold increase in IOPs against the previous model does not necessarily have a market value bounded by an N fold increase in value.
If they are unique in the marketplace, and there is strong enough demand for the improvement, they can sell at a price greater than an N fold increase.
It's not as if you can buy 5 of the previous model SSD and acquire the same reduction in latency, when your writes are synchronous, you can't re-order (or pipeline/parallelize) your write ops.
Your IOPs get bound by the speed of your best SSD.
I say if you want to be into Scientology, good for you, just leave me and other non-scientologists the h*** alone.
Similarly... if you hate scientology, or think it's ignorant, good for you, but leave the scientologists alone, let them do their own things in private, don't shout from the rooftops about how they're all a bunch of f****'ers, etc...
Tolerance = Good
Calling people with certain beliefs gullible/funny/stupid is just cruel.
And potentially dangerous. (When you openly call a system of belief stupid, some of its more adamant practitioners may want to take reprisal, even to the level of violence, in extreme cases...)
The list price for Sun's 16GB flash drive (XTA7210-LOGZ18GB) is >$8,000 US.
This is write-optimized flash for use in a ZFS hybrid storage pool.
E.g. The intended use is you use it as the "log device" (Sun equivalent to ext3 journal) to accelerate the speed of synchronous writes to your storage.
And it has a small fraction of the IOPs and speeds Pliant is claiming.
When they bring this to market, we can expect it to cost 15x to 20x that price per GB, easily, at least initially.
I suggested: And any images referred must be uploaded and served from the ad network (again, no remote loading).
Meaning software running on the server would parse the image file and verify that it is valid and not containing no exploits, prior to the ad being served.
The biggest problem with eBay's new (stupid) "combo boxes" is they don't act like real combo boxes.
I can't tab over to them on my keyboard and use the arrow keys to navigate the options, like I can with a real combobox.
In fact, I can't navigate eBay without a mouse anymore... if my mouse was broken, or I was blind, I'd be screwed....
What's weird is on the "My Ebay" activity page, they use those stupid comboboxes, but on the "Messages" page, they use real comboboxes that work correctly. They also use a real combobox by the "Search" button
(What the hell...) can't they even be consistent about what and where they use these annoying javascript "widgets" ?
No... Office 2007 is bloated as hell. (Office 2003 and OO.o are quite bloated, also)
The bloat is one of the negatives; just as (for MS Office but not OO), the vendor lock-in is one of the negatives.
Nevertheless, the MS Office products are very useful, very good at what they are designed for.
The Office Products are well enough polished (in terms of graphics design) that OO.o looks poor compared to them, to MS Office users.
And in many cases, for large enterprises, the cost of upgrade to 2007, and the cost of the additional hardware, is less than the cost of switching to OO.o.
Due to required document conversion projects, and required efforts to "patch" and fix important documents that don't convert properly.
They need to take responsibility for what they publish on their own sites.
I'd like to see a class action suit against the NY Times or the ad network they use by users who were infected.
Based on NYT negligently allowing advertisers to inject code into their web site.
I can understand users getting hit with fake dialogs after clicking on an ad.
But I believe web sites have a duty to take standard precautions and avoid loading remote script code
I differentiate ad content from code.
It's not rocket science -- when the advertiser uploads their ad unit, sanitize the input, so the upload cannot contain any javascript, SCRIPT, IFARME, FRAME, or other unexpected tags or tag attributes, for that matter, or any remote loading. Only approved 'safe' HTML tags such as IMG. And any images referred must be uploaded and served from the ad network (again, no remote loading).
Again, it's not rocket science to sanitize input. There's really no excuse for not doing it, other than negligently ignoring security issues, and possible harm malicious ads can do...
Slashdot Mirror
I could see Google using it to protect their account signup/login/new service signup processes, not their search function.
What's more valuable is other people using reCaptcha technology. Google can now benefit from their use, by using the service to assist their book scanning/OCR efforts.
Hardware RAID controllers have exactly the same issue. They won't save you from the problem you're describing
The RAID5/6 write hole is closed on hardware controllers. Usually by a battery-backed cache module.
It does nothing as for silent data corruption; however.
3 arrays? Why? Do it in software, do RAID 5. How likely is it that you'll have a bad sector, or a miswrite, that hits both the stripes and the parity?.
It doesn't matter. Just one of them has to have been miswritten. With RAID5, you have no way of telling which one is correct and which one is wrong.
Parity = STRIPE D1 (XOR) STRIPE D2
You can customize your RAID implementation to read all 3 disks... it will find the bit where the bit in the parity stripe is not equal to STRIPE1_BITS (XOR) STRIPE2_BITS. It can know that the bit read is inconsistent, there must be the error.... the problem is RAID5 won't know which disk has the error..
How do you know which stripe is right and which stripe is wrong?
The answer is, you don't know unless you have double parity (RAID6), which would further impact performance, and reduce the amount of storage you get per $$$.
Using software RAID5/6 in such a scenario is even worse than RAID1 in a 3-way mirror config, due to the RAID write hole it introduces, you lose terribly, if, say the power fails. Using PCIe controllers with hardware RAID would massively increase the cost of the box.
It's not likely the same bit error occurs, but RAID5 cannot correct these errors for you at all, or RAID6 cannot correct these errors for you (once a single drive has failed). When using consumer drives, the probability of an eventual drive failure in your RAID5/6 array is high, and chances are it happens before you need your data. Different drives can have different bits in error. And when one of those drives fails (before you can detect any error), the redundancy is no longer there to even try. When you rebuild the array, the bits will still be in error.
That's why it's called silent data corruption RAID5/6 implementations do not read bits from all 3/4 stripes when a read is requested, that would be slow, even slower than it is due to the XOR computation against 2 stripes; they distribute their reads.
And look how massively the Backblaze unit has oversubscribed SATA channels with all these multipliers, software RAID rebuild performance will be quite reduced VS more common server configurations.
To be able to verify in software, you'd have to break the RAID abstraction and somehow read bits from each underlying disk, perform both XORs, compare them, alert if they are different. And somehow pull those bits from the other box that doesn't detect an error.
Well, Linux software RAID doesn't offer a way to do this... and it is a layering violation, also, kernel developers are not likely to add a feature such as that, as a result.
Or use ZFS. It'll checksum everything, so yes, it'll know which array has the right value.
Yes, a very good answer to all silent data corruption issues, unfortunately, their hardware uses port multipliers, which don't exist in a non-development version of OpenSolaris yet. Maybe in 2010....
Although it appears they got bought by EMC.. hrm.
Deduplication can help you reduce the size requirements on the backup server.
If buying new capacity, you should probably think about buying a backup server that can be expanded to have more capacity than your existing server, depending on current server usage.
Plan for a few years down the road, when it becomes necessary to expand capacity of the main server, backup more servers... or more likely: store multiple old versions of files that changed over time.
Normally.. if you have a 500 mb video file, and someone made some edits to it and re-saved. There are now going to be two 'files' in the backup repository for a time: the old version and the new version with the edits (twice the space usage)
So storage requirements on the backup server can actually be much more than storage requirements on the server being backed up.
If it's static, you could spend a few hundred $$ and buy 100 Blu-Ray Recordable disks; burn the 12 TB in sets of 50, and vault them at an off-site location.
get a small hosted account somewhere that you can continuously rsync the 30gb or so between backup sets.
When enough new video has been added to fill a BD-R, burn two copies of it, move the copies on the rsync server to a 'to be deleted directory', vault the BD-Rs off site, and then after that, delete the files off the rsync server.
The hard drives are desktop class, not designed for 24x7 operation. Not designed for massive write traffic that server backups generates.
Latent defects on disks are a real concern.
You write your data to a disk, but there's a bad sector, or miswrite, and when you go back later (perhaps when you need the backup), there are errors on the data you are reading from the disk.
Moreover, you have no way of detecting it, or deciding which array has recorded the "right value" for that bit...
That is, unless every bit has been copied to 3 arrays.
And every time you read data, you compare all 3. (Or that you have two copies and a checksum)
Well, the complexity of this redundancy reduces the reliability overall, and it has a cost.
Maybe, but you will need at least 3 of them and some pretty smart software to have reasonable reliability. The cheapness of backblaze's pods comes from cutting a lot of corners in the hardware.
If you look at the design of Backblaze, this is not server grade equipment: they have two power supplies, but they are not redundant, and these are desktop power supplies not designed to be operated 24/7/365.
If PSU B goes out, a large number of your disk drives and fans go offline; so you lose data and maybe burn up.
And PSU A powers the mainboard...
The disks used by Backblaze are Seagate 7200.11.
These are desktop drives, not designed for 24x7 operation like in a backup server, MTBF stats are based on 1200 power-on hours per year.
Hello... silent data corruption. The last thing you need is for the time to come when you want to restore your backup, and you find latent defects on some of your disks means there are errors in the bits that got saved.
And re-post the transcript of the document received from the FOIA request, as journalists often do....
The hard stuff is multi-touch. And pen computing. Also, there's no point in apps trying to display custom mouse pointers or cursors.
So for most purposes, they could abstract and mostly make it look just like a mouse.. one "tap" generates an event as if you had clicked in that spot, two taps generates an event as if you had double clicked.
Putting your finger to the screen and moving around = mouse movement.
For click and drag, may want pressure sensitivity, a button, two fingers, or something...
Well, it's distinctive from other software apps, but it's not in general a distinctive interface.
Everyone with a physical bookshelf at home has seen Books and DVDs sitting in a bookshelf. That essentially means the UI isn't "novel", as in he didn't create a new type of interface.
He took something everyone's seen before and made his UI look just like it.
And his app's not the first to use that metaphor either. But it's probably the first that actually used that metaphor, did something useful, and didn't suck.
So, uh, the wood grain is very possibly the best thing he's got to complain about....
Taking someone's photo and making a lot of changes to it does not cause the result to no longer be a derivative work.
Presumably his wood grain was just one copyrighted images, the developers of these apps used to create an overall unique composition...
I see X as able to support all sorts of input devices... touch screen support should be standard..
We should get touch features in common apps, they should be done in a way that makes the experience superior to anything Windows can muster.
Hey, if that ever happens, it could be the year of the Linux desktop :)
It's a start, but I also want:
I'm all for it, provided I can stipulate that the information I provide or record by having such a device, may not be used against me.
Or get you sued by the person for intrusion upon seclusion, if they find out...
Better make sure your cameras are well-hidden, and you make the submissions in an anonymous way that can't be tracked back to you or your camera.
"Dangerous cult" is a cry of the oppressor.
One person's "dangerous cult" is another person's "religion".
In ancient times, people called Christianity a dangerous cult. And it led to christians being slaughtered and many civil wars.
IOW, a belief that a religion is a cult that cf. your beliefs would consider dangerous or offensive, is not a valid reason for intolerance.
I noticed that.. but then again, a lot of us are still using SAS HBAs that only support 3gb/s (or ~384 megabytes per second). Which is (I suppose) a limitation of what will soon be considered older SAS hardware.
Still, SAS devices aren't bound to SATA 1.5 gigabit limitations... in fact, even the SATA II next gen spec can also reach 3gb/s, and SATA III will be 6gb/s also.
And 6gb/s (768 megabytes/sec) is definitely more than 500 MB/s :)
Maybe nobody would buy them at 15x, but just b/c the improvement isn't worth it to them.
A new device with an N fold increase in IOPs against the previous model does not necessarily have a market value bounded by an N fold increase in value. If they are unique in the marketplace, and there is strong enough demand for the improvement, they can sell at a price greater than an N fold increase.
It's not as if you can buy 5 of the previous model SSD and acquire the same reduction in latency, when your writes are synchronous, you can't re-order (or pipeline/parallelize) your write ops.
Your IOPs get bound by the speed of your best SSD.
I say if you want to be into Scientology, good for you, just leave me and other non-scientologists the h*** alone.
Similarly... if you hate scientology, or think it's ignorant, good for you, but leave the scientologists alone, let them do their own things in private, don't shout from the rooftops about how they're all a bunch of f****'ers, etc...
Tolerance = Good
Calling people with certain beliefs gullible/funny/stupid is just cruel.
And potentially dangerous. (When you openly call a system of belief stupid, some of its more adamant practitioners may want to take reprisal, even to the level of violence, in extreme cases...)
The list price for Sun's 16GB flash drive (XTA7210-LOGZ18GB) is >$8,000 US. This is write-optimized flash for use in a ZFS hybrid storage pool.
E.g. The intended use is you use it as the "log device" (Sun equivalent to ext3 journal) to accelerate the speed of synchronous writes to your storage.
And it has a small fraction of the IOPs and speeds Pliant is claiming.
When they bring this to market, we can expect it to cost 15x to 20x that price per GB, easily, at least initially.
That's why their product is an Enterprise SAS drive, not a SATA drive. SAS can get 3 gigabits per second.
SAS is Serial Attached SCSI, which isn't the same thing as SATA.
SATA is a consumer-level/workstation technology. Whereas SAS is for servers.
You can plug a SATA drive into a SAS port, but can't plug a SAS drive into a SATA port. :)
I suggested: And any images referred must be uploaded and served from the ad network (again, no remote loading).
Meaning software running on the server would parse the image file and verify that it is valid and not containing no exploits, prior to the ad being served.
The biggest problem with eBay's new (stupid) "combo boxes" is they don't act like real combo boxes.
I can't tab over to them on my keyboard and use the arrow keys to navigate the options, like I can with a real combobox.
In fact, I can't navigate eBay without a mouse anymore... if my mouse was broken, or I was blind, I'd be screwed....
What's weird is on the "My Ebay" activity page, they use those stupid comboboxes, but on the "Messages" page, they use real comboboxes that work correctly. They also use a real combobox by the "Search" button
(What the hell...) can't they even be consistent about what and where they use these annoying javascript "widgets" ?
No... Office 2007 is bloated as hell. (Office 2003 and OO.o are quite bloated, also)
The bloat is one of the negatives; just as (for MS Office but not OO), the vendor lock-in is one of the negatives.
Nevertheless, the MS Office products are very useful, very good at what they are designed for.
The Office Products are well enough polished (in terms of graphics design) that OO.o looks poor compared to them, to MS Office users.
And in many cases, for large enterprises, the cost of upgrade to 2007, and the cost of the additional hardware, is less than the cost of switching to OO.o.
Due to required document conversion projects, and required efforts to "patch" and fix important documents that don't convert properly.
They need to take responsibility for what they publish on their own sites.
I'd like to see a class action suit against the NY Times or the ad network they use by users who were infected.
Based on NYT negligently allowing advertisers to inject code into their web site.
I can understand users getting hit with fake dialogs after clicking on an ad.
But I believe web sites have a duty to take standard precautions and avoid loading remote script code
I differentiate ad content from code. It's not rocket science -- when the advertiser uploads their ad unit, sanitize the input, so the upload cannot contain any javascript, SCRIPT, IFARME, FRAME, or other unexpected tags or tag attributes, for that matter, or any remote loading. Only approved 'safe' HTML tags such as IMG. And any images referred must be uploaded and served from the ad network (again, no remote loading).
Again, it's not rocket science to sanitize input. There's really no excuse for not doing it, other than negligently ignoring security issues, and possible harm malicious ads can do...