So? They really are escaping the patent infringement... in the future. One would assume any mention of escape to involve changing the status of the escapee in the future. You don't generally hear about people escaping something and changing the past, unless you're driving a fine DMC made machine.
This deal means that the WORST that can happen is that Vonage pays some fixed-price settlement/damages/whatever and then the problem is done with. Vonage won't get stuck with an injunction or paying eternal license fees to Verizon.
The issue of if Vonage can eat the settlement fee is another issue, but assuming that they can, (other than the third patent) this effectively solves their problems.
As EMI and Jobs made clear in the announcement, they feel that the vast majority of people don't care about getting DRM-free content or higher quality than 128-kbit AAC. And I believe they're right; the vast majority of people buying music on iTunes for their iPods could care less about DRM and bitrates.
So, what could they do? Remove the DRM and bump up the quality and keep the price the same. Problem with this option is that it doubles bandwidth demands, and eats significantly into their profits. Remember, Apple already takes a tiny share of the price (RIAA gets the rest), so bandwidth matters.
Except since the majority of people don't want (or rather don't care) this, they're effectively hurting their profits giving something that people generally don't care about.
Their solution is to offer the option to the minority who DO care, and charge more for it to cover bandwidth. It's the ideal compromise between Apple and consumers.
That said, I don't think the general population not caring will last. Don't underestimate the appeal of perceived quality to the average consumer. They will hear that these new tracks are "premium", that they "sound better". Consumers have proven countless times in the past that they're willing to spend more for something that they're TOLD is better, even if they can't tell the difference. I think that these new premium tracks will slowly take over as consumers decide that they are going to buy the "premium" track just because "it's better", not because they understand why.
Eventually, as bandwidth prices drop further, and as the premium tracks take over, I expect Apple will discontinue the current 128-kbit tracks and either drop the price of the premium tracks to $1, or settle on $1.29 as the new standard iTunes price.
Interestingly, if they settled on $1.29 as the new standard price, they'd be taking the easy road to doing what the RIAA has wanted them to do all along; raise the prices from $1 per song. Apple gets to have an excuse for the price increase.
The problem is that SciFi picked up permanent rights to air SG-1 on TV. That's the reason nobody else is picking it up; nobody can.
MGM has said that they wanted to shop the show around; SciFi responded saying that they will not permit it.
They might not make it, but they have exclusive rights to it on television.
That's how MGM is getting around this with the DVDs. They can release them on DVD because DVD isn't TV. Of course, SciFi will probably purchase rights to air them later.
The concept holds. The state of Washington could fit their data with zero redundancy on about 100 drives with current technology (1TB drives). That doesn't require a lot of space... If we assume the same drive-to-unit ratio of the Sun Fire X4500 (48 drives per 4U, or 12 drives per U), a single rack would hold 504TB of storage without redundancy.
The idea is to distribute it anyhow. Spread the storage all over the region (be it state, country, or world), and connect it all together with a network (internet, dark fiber, whatever works). Redundancy is not provided by RAID, but by the system itself, which makes sure that there are a certain number of copies of each file (or chunks of file) on geographically diverse locations.
I should think that Google themselves likely store many petabytes of data in their network (Tens of thousands of servers with decent drives each, I'd imagine). That seems to be on the scale of a government entity's demands, from what you describe. Now just scale that up an order of magnitude or such into a largescale distributed network, and you'd be able to handle a pretty huge amount of data in online storage.
I'm of the opinion that online storage is the most reliable; it cannot be destroyed, because any time an individual PC is destroyed and chunks of data are lost, the network automatically should take care of replicating the existing copies of those chunks to ensure there is sufficient redundancy remaining in the network.
You don't even need to do simple chunk mirroring. You could do a parity scheme, say, 2 parity chunks for every 3 data chunks, meaning less than double the storage requirements and still being able to lose any two of five chunks. Who knows how that might work over the internet, I'm just throwing out crazy ideas. When you distribute things properly, you don't need insane redundancy, because as soon as you lose one chunk, you're going to rebuild it somewhere else, and the chance of losing 3 out of 5 chunks at the same time in geographically distinct locations is infinitesimal; even a widespread disaster would be unlikely to affect three different places at the same time, and such a disaster would likely destroy any offline backups you had anyhow:P
But then, I'm rambling again.
Re:Ruse to sell more motherboards
on
eSATA Connectors
·
· Score: 1
That's great but this has nothing to do with "hardware accelerated soundcards". There are *no* PCIe soundcards available today, with any featureset, for any price.
Sure it does. The point is that if host-based (and/or USB based) audio is good enough (and it increasingly is), the vast majority of users don't need discrete soundcards at all, let alone PCI-Express ones.
We can expect it, but do we have any reason to believe it'll happen soon?
We do. They appear to have demonstrated it, assuming that the models they showed off were actually functioning.
How long does it take before the complete absense of an entire industry from transitioning to a new bus is seen as a problem with the bus, and not one company (or ten)?
The discrete soundcard "industry" has shriveled to a tiny fraction of what it once was. These days, the typical user does NOT have a soundcard in their computer. They likely use the sound chipset integrated into their motherboards. Gamers are increasingly likely to use heapdhones that include a USB soundcard.
With the market for discrete internal soundcards so weak, THAT can be seen as the real reason they haven't moved to PCI-Express yet; the market is already small and all PCs that have PCI-e slots still have PCI slots, so there is no real benefit in introducing a PCI-Express soundcard.
Obviously, since many notebooks sold today have completely transitioned to PCI-Express via ExpressCard slots (fairly recently though), there *is* the pressure for them to do it there, which is obviously why they're getting ready to introduce those kinds of cards.
I see this as a problem with market pressures rather than rumored technical issues.
I wonder though... It may be the supply on shelves that's the problem. Sony can cut their price, but the stores already paid for the PS3s they have on the shelves. I doubt that they're going to cut their price and sell them at a loss.
So any price cut by Sony will only take effect AFTER all the retail stores sell through their existing inventory.
This also won't counteract the fact that the PS3 is (going to be) enormously MORE expensive in Europe than in North America...
The solution to this may be as simple as saying "Don't store data on offline mediums at all".
By offline mediums I mean things like tape, optical disk, or even a solitary hard disk.
Instead, I see the future of reliable data storage as vast networks/clusters of shared storage with built-in redundancy. Look at how Google tackles their data storage needs. They have tens of thousands of highly unreliable machines, but they use them in such a way that they can store a large amount of data in a highly reliable manner.
If all "important" data were kept in such large distributed (geographically as well) or global/universal networks, there would be no need for offline storage. There'd be no need to back data up, because all data would be stored reliably enough anyhow. If you break into a Google datacenter with a shotgun and start shooting hard-drives, you're not going to cause any data loss.
By the same token, any offline backup can be physically destroyed. If you've backed up your data in an offline digital (or analog) form, those objects can be destroyed. With redundant storage on a cluster-level, the loss of any individual object isn't important.
I'm probably not conveying the idea very well, but my point is that if all information is stored in some sort of global storage network with inherent reliability (even if the actual storage is unreliable), then you're better off than offline backups and have the added advantage of all data being accessible all the time.
Re:Ruse to sell more motherboards
on
eSATA Connectors
·
· Score: 1
Hardly. The typical PC wouldn't even go close to saturating a basic 33Mhz/32 bit PCI bus. They just don't do enough "stuff".
The typical PC is not the only scenario though. A lot of enthusiasts are/were also stuck with it. And even the average consumer using the typical PC probably occasionally runs into the scenario where one device is interfering with their PCI-based soundcard.
33/32 PCI, as seen in most consumer level systems, tops out at around 120MB/sec aggregate bandwidth. *Most* people wouldn't come anywhere close to that, except in contrived (and hence rare) benchmarking situations. Copying a file over the network to another machine is a reasonable example of one of the few ways this might happen, but average consumer level hard disks won't sustain much more than about 50MB/s (except in ideal conditions) and it's unlikely the machine on the other end will be much better.
GigE is increasingly a consumer application, to the point where the typical PC has a decent chance of having onboard ethernet. I don't know why GigE on PCI caps out at about 40% of the actual bandwidth PCI supplies; my guess is there's some overhead on top of simply passing the packets back and forth, and/or the benchmarks were using PCI-based disk controllers.
The average drive is getting faster and faster, and this will rapidly continue, especially with the introduction of perpendicular recording. Current top-end drives such as the Hitach 1TB drive, or the WD Raptor, currently post average speeds of ~75MB/s, significantly higher than where PCI-based GigE network cards hit their wall (About 50-60MB/s).
These capacities of drives, while high end now, will in short order be in the typical consumer's PC. If we consider the typical PC to today have a 250GB drive (Yes, I know Dell currently is offering 500GB drives in PCs under $1000, but that's a promotion), and considering how rapidly perpendicular recording is pumping up capacities much faster than something like Moore's Law might predict, it will only be a few years before 1TB drives are typical. And with every passing day, such high speed drives inch closer and closer towards typical as they become cheaper and cheaper.
So, even today, a high-end user isn't going to have much trouble hitting PCI's limitations. It's not limited to simply contrived benchmarks.
The Raptor is a pretty common drive itself. I wouldn't even call it enthusiast level, simply high-end. An enthusiast today looking for the maximum speed might combine two or more Raptor drives in RAID, at which point that RAID card, with a theoretical maximum sustained speed of about 180MB/s if both drives are reading from the outside of their platters, and average speeds of perhaps 150MB/s, PCI is again going to be the limiting factor.
As for GigE saturating the PCI bus, motherboard and chipset manufacturers already found a solution; they moved the GigE chipsets off the PCI bus and onto dedicated or non-PCI buses. And of course, we now have cheap PCIe GigE cards to solve that for add-in cards, as well as for RAID, and other such uses.
OK, so this post is getting a bit rambling, but my point is that you didn't (don't) have to go very far from the "typical" PC to reach a user who can saturate the PCI bus during typical usage.
Re:Ruse to sell more motherboards
on
eSATA Connectors
·
· Score: 1
So according to that article, there are no PCIe soundcards because it's a different technology and Creative can't just slap a PCI-Express connector on their existing hardware with minimal R&D effort. BOO HOO. Creative and their hardware accelerated soundcards are becoming increasingly irrelevant as host-based audio progresses... Especially with Vista removing support for hardware accelerated audio in general (without workaround hacks in the drivers for OpenAL support) and providing much of the functionality in software.
We can expect PCI-Express soundcards from Creative soon, perhaps starting with their ExpressCard version of the X-Fi for notebooks (ExpressCard uses either PCIe for internal slots or USB2.0 for external slots).
Re:Ruse to sell more motherboards
on
eSATA Connectors
·
· Score: 1
Most cards, yes. But in recent years, it has become a problem. Gigabit NICs and RAID controllers don't have much trouble saturating the bus. PCI-based gigabit NICs tend to top out at about 400mbit, which is odd considering that's significantly less than PCI's own limits... I guess there's some overhead going on there.
There were also many conflicts because of the shared bus. Who here hasn't had trouble with stuttering audio on a PCI soundcard because some other device in the bus was causing trouble?
PCI-X solved the bandwidth problem, but never really made it into the consumer space. And yet the bandwidth demands from consumers increasingly stressed the PCI bus.
Motherboard manufacturers eventually solved the problem by moving on-board gigabit controllers from the PCI bus to either a dedicated bus (Intel did this, I think), or a different general purpose bus (HyperTransport?). Such controllers saw an immense performance improvement.
RAID controllers are a good example because modern 7200RPM drives have peak (not burst, peak) read speeds of about 90MB/s at the outside of the disk. Put two of those on a RAID controller and you've got some serious bandwidth issues.
Re:Any advantages over having only one connector?
on
eSATA Connectors
·
· Score: 5, Informative
The official specs for eSATA essentially say the cable is better suited for external use. Here's what they have to say about it:
The external cable connector is a shielded version of the connector specificed in SATA 1.0a with these basic differences:
The External connector has no "L" shaped key, and the guide features are vertically offset and reduced in size. This prevents the use of unshielded internal cables in external applications.
To prevent ESD damage, the insertion depth is increased from 5mm to 6.6mm and the contacts are mounted further back in both the receptacle and plug.
To provide EMI protection and meet FCC and CE emission requirements, the cable has an extra layer of shielding, and the connectors have metal contact points.
There are springs as retention features built into the connector shield on both the top and bottom surfaces.
The external connector and cable are designed for over five thousand insertions and removals while the internal connector is only specified to withstand fifty.
They make it pretty clear exactly what's different. The biggest difference is the cable is shielded, while internal SATA is not (or less so?). And obviously the connector being rated for a hundred times more insertions is a pretty big difference.
I should note that in recent benchmarks done by MaximumPC, eSATA did not provide substantial performance benefits over Firewire800 drives. eSATA featured a higher burst speed, but more or less equivalent average transfer rates and seek times. Unless there were specific licensing issues with Firewire 800, I would rather have seen it become the preferred drive interface; I'll take a general purpose connector that I can use for other stuff over something as specific as eSATA any day, especially when eSATA provides little benefit.
Re:Ruse to sell more motherboards
on
eSATA Connectors
·
· Score: 5, Informative
While I'm not as certain when it comes to graphics cards, PCI-Express is/was sorely needed to replace PCI for general expansion cards. PCI had a limited amount of bandwidth available that was extremely easy to saturate (A single gigabit NIC would hit a bandwidth wall at something like 400mbit). The shared nature of the PCI bus was also an extremely frustrating limitation.
By contrast, PCI-Express 1x slightly increases the bandwidth from roughly 133MB/s to 150MB/s, but more importantly each device gets that, it's not shared anymore. And of course, 2x and higher slots provide more bandwidth.
But when it comes to graphics, AGP 8x was (at the time) providing more than enough bandwidth... as for the demands of modern monsterous graphics cards (such as the 8800 GTX), for all I know they might be able to saturate an AGP 8x bus.
As I understand it anyhow, the more tangible benefits from moving from AGP to PCI-Express were increase bi-directional bandwidth (AGP was great at Host->Card, but sucked at Card->Host), and increased ease with sticking multiple PCI-Express slots on the motherboard, making modern SLI possible.
Why would you need a hardphone to use Asterisk? Grab a soft phone that supports a high quality voice codec such as Speex (Asterisk supports it), connect the two Asterisk servers with IAX, and you're set.
Personally I'd just grab two copies of Skype, forward the ports to minimize latency, and go at it; the quality and latency is good enough for live broadcasting when it is set up properly (again, with ports forwarded), and the quality between two properly configured Skype clients is significantly better than POTS. Sure, it won't be as good as Speex through Asterisk, but it is a tad easier and faster to set up, to say the last;)
They'll need some sort of emulation layer, but because they control the entire OS, they have the potential to do this with enormous performance.
They can port all *their* code (and all the APIs) to the PowerPC. The only thing left running x86 code will be the actually application code. But since they'll be making calls to native libraries, you don't get the overhead of having to emulate the entire OS.
This is actually how Apple has done it all the way back to the 68k/PowerPC switchover. And during the PowerPC/x86 switchover, they benefited from the same thing. The reason that Rosetta (Or rather, Transitive's QuickTransit) is so fast emulating a PowerPC on an Intel processor (which is normally much more difficult than the other way around) is because Rosetta isn't emulating an entire OS like PearPC and other similar projects are attempting. Rosetta just emulates the application code, and the application's API calls are made to native APIs.
There is some trickery needed to make cross-platform API calls, of course, but the performance benefits are insane:)
Not that I know anything about implementation. But the observation of Apple's two arch migrations is proof enough.
It won't strictly double bandwidth usage. There are several methods that proxy servers can employ to reduce bandwidth usage.
The first is the use of a cache. For example, if the primary use of the proxy is to visit myspace, many of the images, stylesheets, scripts, and so on will be cached. That reduces the proxy to internet bandwidth.
Another is compressing text content that passes through the proxy. HTML, javascript, XML, and stylesheets can make up a surprising portion of bandwidth usage these days. By having the proxy compress these types of content as it passes through, user to proxy bandwidth can be reduced.
There are obviously other more destructive methods; GIF images can be run through gif2png and HTML rewritten to use the new images, JPG files can be recompressed at lower quality settings, all sorts of things can be done on-the-fly to reduce bandwidth use.
I'd imagine that because of the repetitive use, the bandwidth demands with some of the above mentioned techniques would be substantially less than double.
Of course, all this discussion is moot; the instant the school realizes that the kids are using this proxy, they'll block it, and that'll be that.
That's the wonderful thing about Linux. Don't like the mainstream, and want to keep using an ideologically pure distribution? Nobody is stopping you. There are a whole bunch of distributions that specialize in that, and there's nobody holding a gun to your head saying "You must use Ubuntu."
Of course, this is also one of the things that keeps Linux as a niche product on the desktop. But that's another topic.
5000 CPUs isn't particularly impressive either when you consider 1U quad core quad processor machines. Up to 672 CPU cores per rack. Pretty insane power requirements as far as density is concerned, though.
Let's not forget that the Sun X4500 is a quad-processor beast itself, and you're not going to saturate all 16 (possible) cores with storage demands, unless you're processing the data locally.
A more dedicated processor such as the Cell (which would seem to be ideally suited for software rendering) would reduce the space/power/heat requirements significantly, although I'm not sure of what kind of enclosure you'd need. Assuming you're not shoving PS3s into a datacenter, the Cell comes on PCI-X cards. I'm sure there is some enclosure designed to maximize the number of PCI-X cards in a limited space.
I don't think their datacenter is bad, by any means, just that the technology exists to do what they're doing in much less space (and possibly cheaper).
Did you ever consider that many companies centralize the important data into shared user directories? A redundant centralized server with proper backups are much simpler than setting up dedicated backup solutions on EVERY workstation.
So, reinstalling the OS from scratch on a workstation certainly is a good way to perform disaster recovery; the workstation is borked, and all the user settings are server-side, so why NOT nuke the workstation?
Of course, such a company would probably also install the workstations from a ghost image. However I work for a company that does go the centralized route and yet doesn't use ghost images (we have an instruction list of what to install and how to set the machine up).
More like, in New York City, 10 million people commute to work each day (making up figures here), each traveling down the road at 50km/h. So the aggregate commute speed in New York City is 46% the speed of light!
300TB is still nothing, even with insane redundancy.
For example, assume you've got Sun Fire X4500 servers, and that you take them with the stock 500GB drives. 24TB per 4U server.
Now let's assume that you run RAIDZ2 on each server, dedicating 2 of the 48 drives to parity. That's 23TB per server. Now let's assume you want some redundancy, as in, completely separate failover capacity. You mirror every single server with hot standbys. Easy with ZFS, you can mirror the file system in real time without any issues. So we get an effective 11.5TB per 4U server.
Now, we need 300TB, so that requires 27 servers. 108U, or three 42u racks, with each rack having 6u for network infrastructure, load balancers to do the failover to the hot-standby units (or even localized UPS if you want double UPS coverage).
3 racks. With pretty decent redundancy that allows two drives per server to fail before you need to switch over to the hot standbys, and that's ignoring the fact that each X4500 server has redundant power supplies and allows hot swapping of drives and power supplies. THREE RACKS.
Explain to me why they need a 10,000 sq. ft. datacenter for three racks worth of servers? Yeah, I know, AC, UPS, generators, all that stuff takes up room, but I could fit a lot more than 300TB in a 10k sq. ft. datacenter.
Sure, I'm no expert, and I might be missing some important considerations, but even I could set up a 300TB high-uptime file storage network without needing an entire datacenter to do it.
We have the multiple election days in Canada (Three of them, Federal, Provincial, and Municipal. Referendums are extremely rare and often happen only every decade or two, and are never country-wide). Unless I'm mistaken, our voter turnout is HIGHER than the US, despite having three election days instead of one.
I wouldn't want it any other way. On Federal election night, I'm glued to the TV watching the results. There's enough information being thrown at us with just one election, why would I want three elections at the same time?!
Of course, we don't have fixed elections in Canada. Federal elections happen whenever the current government wants to, with exceptions such as there being a max length of a term, and the government collapsing due to a vote of non-confidence (which happens on average every 18 months in a minority government, of which our current government is one). So a single election day couldn't possibly work with out voting system.
Municipal elections are still held on a fixed schedule, though. Since municipalities are strictly the responsibility of the province, and not the federal government, each province sets their own municipal election dates, on a fixed schedule.
Erased markings? Not dark enough? Not problems. The pencils provided at voting booths are both more than dark enough, and lack erasers. If you're saying that ballots might be erased after the fact, this is prevented by procedure; counting is done under supervision, there is no chance to erase a ballot.
In Canada, with I'm sure a few exceptions, the simple "make a mark in the circle with the pencil" system works perfectly fine. In federal elections, at least, there haven't been (to my knowledge) any serious recent problems with miscounts and fraud issues.
One thing you have to keep in mind is that our ballots here are incredibly simpler than the US. In the US, a federal ballot has dozens of candidates for a huge number of posts, along with propositions thrown in.
In a Canadian election ballot, there are usually roughly 3 to 5 names on the ballot, each with a large white circle that is perhaps an inch wide. You put a mark in the circle with the pencil.
Vote counting in Canada after a federal election is usually completed in an hour or so, although some ridings can take longer, and different polls close at different times due to the timezones. Still, the final results are almost always known by the next day. This is why I find it very strange that in the US, counting votes for federal elections takes so much longer. I suppose it is because of the more complicated multi-choice ballots.
And notice also that in the text above it says "The sponsor is adware". It can't be clearer. It up and TELLS you it's adware in exactly those words, and gives the option to not install it.
Huh? I'm a user of MsgPlus. Every time I've ever installed it or an update, it has given me a CHOICE as to if I want to install the toolbar. There is an ENTIRE page of the installer dedicated to explaining what it is, and giving the option as to if you want to install it.
The guy's been working full time on a free product, and STILL gives the option not to install it. I don't know about you, but I trust somebody like him a lot more than the people at Debian who want to rename Firefox because the icon is free-as-in-beer instead of free-as-in-speech. At least Patchou is being reasonable.
Slashdot Mirror
So? They really are escaping the patent infringement... in the future. One would assume any mention of escape to involve changing the status of the escapee in the future. You don't generally hear about people escaping something and changing the past, unless you're driving a fine DMC made machine.
This deal means that the WORST that can happen is that Vonage pays some fixed-price settlement/damages/whatever and then the problem is done with. Vonage won't get stuck with an injunction or paying eternal license fees to Verizon.
The issue of if Vonage can eat the settlement fee is another issue, but assuming that they can, (other than the third patent) this effectively solves their problems.
As EMI and Jobs made clear in the announcement, they feel that the vast majority of people don't care about getting DRM-free content or higher quality than 128-kbit AAC. And I believe they're right; the vast majority of people buying music on iTunes for their iPods could care less about DRM and bitrates.
So, what could they do? Remove the DRM and bump up the quality and keep the price the same. Problem with this option is that it doubles bandwidth demands, and eats significantly into their profits. Remember, Apple already takes a tiny share of the price (RIAA gets the rest), so bandwidth matters.
Except since the majority of people don't want (or rather don't care) this, they're effectively hurting their profits giving something that people generally don't care about.
Their solution is to offer the option to the minority who DO care, and charge more for it to cover bandwidth. It's the ideal compromise between Apple and consumers.
That said, I don't think the general population not caring will last. Don't underestimate the appeal of perceived quality to the average consumer. They will hear that these new tracks are "premium", that they "sound better". Consumers have proven countless times in the past that they're willing to spend more for something that they're TOLD is better, even if they can't tell the difference. I think that these new premium tracks will slowly take over as consumers decide that they are going to buy the "premium" track just because "it's better", not because they understand why.
Eventually, as bandwidth prices drop further, and as the premium tracks take over, I expect Apple will discontinue the current 128-kbit tracks and either drop the price of the premium tracks to $1, or settle on $1.29 as the new standard iTunes price.
Interestingly, if they settled on $1.29 as the new standard price, they'd be taking the easy road to doing what the RIAA has wanted them to do all along; raise the prices from $1 per song. Apple gets to have an excuse for the price increase.
The problem is that SciFi picked up permanent rights to air SG-1 on TV. That's the reason nobody else is picking it up; nobody can.
MGM has said that they wanted to shop the show around; SciFi responded saying that they will not permit it.
They might not make it, but they have exclusive rights to it on television.
That's how MGM is getting around this with the DVDs. They can release them on DVD because DVD isn't TV. Of course, SciFi will probably purchase rights to air them later.
The concept holds. The state of Washington could fit their data with zero redundancy on about 100 drives with current technology (1TB drives). That doesn't require a lot of space... If we assume the same drive-to-unit ratio of the Sun Fire X4500 (48 drives per 4U, or 12 drives per U), a single rack would hold 504TB of storage without redundancy.
:P
The idea is to distribute it anyhow. Spread the storage all over the region (be it state, country, or world), and connect it all together with a network (internet, dark fiber, whatever works). Redundancy is not provided by RAID, but by the system itself, which makes sure that there are a certain number of copies of each file (or chunks of file) on geographically diverse locations.
I should think that Google themselves likely store many petabytes of data in their network (Tens of thousands of servers with decent drives each, I'd imagine). That seems to be on the scale of a government entity's demands, from what you describe. Now just scale that up an order of magnitude or such into a largescale distributed network, and you'd be able to handle a pretty huge amount of data in online storage.
I'm of the opinion that online storage is the most reliable; it cannot be destroyed, because any time an individual PC is destroyed and chunks of data are lost, the network automatically should take care of replicating the existing copies of those chunks to ensure there is sufficient redundancy remaining in the network.
You don't even need to do simple chunk mirroring. You could do a parity scheme, say, 2 parity chunks for every 3 data chunks, meaning less than double the storage requirements and still being able to lose any two of five chunks. Who knows how that might work over the internet, I'm just throwing out crazy ideas. When you distribute things properly, you don't need insane redundancy, because as soon as you lose one chunk, you're going to rebuild it somewhere else, and the chance of losing 3 out of 5 chunks at the same time in geographically distinct locations is infinitesimal; even a widespread disaster would be unlikely to affect three different places at the same time, and such a disaster would likely destroy any offline backups you had anyhow
But then, I'm rambling again.
That's great but this has nothing to do with "hardware accelerated soundcards". There are *no* PCIe soundcards available today, with any featureset, for any price.
Sure it does. The point is that if host-based (and/or USB based) audio is good enough (and it increasingly is), the vast majority of users don't need discrete soundcards at all, let alone PCI-Express ones.
We can expect it, but do we have any reason to believe it'll happen soon?
We do. They appear to have demonstrated it, assuming that the models they showed off were actually functioning.
How long does it take before the complete absense of an entire industry from transitioning to a new bus is seen as a problem with the bus, and not one company (or ten)?
The discrete soundcard "industry" has shriveled to a tiny fraction of what it once was. These days, the typical user does NOT have a soundcard in their computer. They likely use the sound chipset integrated into their motherboards. Gamers are increasingly likely to use heapdhones that include a USB soundcard.
With the market for discrete internal soundcards so weak, THAT can be seen as the real reason they haven't moved to PCI-Express yet; the market is already small and all PCs that have PCI-e slots still have PCI slots, so there is no real benefit in introducing a PCI-Express soundcard.
Obviously, since many notebooks sold today have completely transitioned to PCI-Express via ExpressCard slots (fairly recently though), there *is* the pressure for them to do it there, which is obviously why they're getting ready to introduce those kinds of cards.
I see this as a problem with market pressures rather than rumored technical issues.
I wonder though... It may be the supply on shelves that's the problem. Sony can cut their price, but the stores already paid for the PS3s they have on the shelves. I doubt that they're going to cut their price and sell them at a loss.
So any price cut by Sony will only take effect AFTER all the retail stores sell through their existing inventory.
This also won't counteract the fact that the PS3 is (going to be) enormously MORE expensive in Europe than in North America...
The solution to this may be as simple as saying "Don't store data on offline mediums at all".
By offline mediums I mean things like tape, optical disk, or even a solitary hard disk.
Instead, I see the future of reliable data storage as vast networks/clusters of shared storage with built-in redundancy. Look at how Google tackles their data storage needs. They have tens of thousands of highly unreliable machines, but they use them in such a way that they can store a large amount of data in a highly reliable manner.
If all "important" data were kept in such large distributed (geographically as well) or global/universal networks, there would be no need for offline storage. There'd be no need to back data up, because all data would be stored reliably enough anyhow. If you break into a Google datacenter with a shotgun and start shooting hard-drives, you're not going to cause any data loss.
By the same token, any offline backup can be physically destroyed. If you've backed up your data in an offline digital (or analog) form, those objects can be destroyed. With redundant storage on a cluster-level, the loss of any individual object isn't important.
I'm probably not conveying the idea very well, but my point is that if all information is stored in some sort of global storage network with inherent reliability (even if the actual storage is unreliable), then you're better off than offline backups and have the added advantage of all data being accessible all the time.
Hardly. The typical PC wouldn't even go close to saturating a basic 33Mhz/32 bit PCI bus. They just don't do enough "stuff".
The typical PC is not the only scenario though. A lot of enthusiasts are/were also stuck with it. And even the average consumer using the typical PC probably occasionally runs into the scenario where one device is interfering with their PCI-based soundcard.
33/32 PCI, as seen in most consumer level systems, tops out at around 120MB/sec aggregate bandwidth. *Most* people wouldn't come anywhere close to that, except in contrived (and hence rare) benchmarking situations. Copying a file over the network to another machine is a reasonable example of one of the few ways this might happen, but average consumer level hard disks won't sustain much more than about 50MB/s (except in ideal conditions) and it's unlikely the machine on the other end will be much better.
GigE is increasingly a consumer application, to the point where the typical PC has a decent chance of having onboard ethernet. I don't know why GigE on PCI caps out at about 40% of the actual bandwidth PCI supplies; my guess is there's some overhead on top of simply passing the packets back and forth, and/or the benchmarks were using PCI-based disk controllers.
The average drive is getting faster and faster, and this will rapidly continue, especially with the introduction of perpendicular recording. Current top-end drives such as the Hitach 1TB drive, or the WD Raptor, currently post average speeds of ~75MB/s, significantly higher than where PCI-based GigE network cards hit their wall (About 50-60MB/s).
These capacities of drives, while high end now, will in short order be in the typical consumer's PC. If we consider the typical PC to today have a 250GB drive (Yes, I know Dell currently is offering 500GB drives in PCs under $1000, but that's a promotion), and considering how rapidly perpendicular recording is pumping up capacities much faster than something like Moore's Law might predict, it will only be a few years before 1TB drives are typical. And with every passing day, such high speed drives inch closer and closer towards typical as they become cheaper and cheaper.
So, even today, a high-end user isn't going to have much trouble hitting PCI's limitations. It's not limited to simply contrived benchmarks.
The Raptor is a pretty common drive itself. I wouldn't even call it enthusiast level, simply high-end. An enthusiast today looking for the maximum speed might combine two or more Raptor drives in RAID, at which point that RAID card, with a theoretical maximum sustained speed of about 180MB/s if both drives are reading from the outside of their platters, and average speeds of perhaps 150MB/s, PCI is again going to be the limiting factor.
As for GigE saturating the PCI bus, motherboard and chipset manufacturers already found a solution; they moved the GigE chipsets off the PCI bus and onto dedicated or non-PCI buses. And of course, we now have cheap PCIe GigE cards to solve that for add-in cards, as well as for RAID, and other such uses.
OK, so this post is getting a bit rambling, but my point is that you didn't (don't) have to go very far from the "typical" PC to reach a user who can saturate the PCI bus during typical usage.
So according to that article, there are no PCIe soundcards because it's a different technology and Creative can't just slap a PCI-Express connector on their existing hardware with minimal R&D effort. BOO HOO. Creative and their hardware accelerated soundcards are becoming increasingly irrelevant as host-based audio progresses... Especially with Vista removing support for hardware accelerated audio in general (without workaround hacks in the drivers for OpenAL support) and providing much of the functionality in software.
We can expect PCI-Express soundcards from Creative soon, perhaps starting with their ExpressCard version of the X-Fi for notebooks (ExpressCard uses either PCIe for internal slots or USB2.0 for external slots).
Most cards, yes. But in recent years, it has become a problem. Gigabit NICs and RAID controllers don't have much trouble saturating the bus. PCI-based gigabit NICs tend to top out at about 400mbit, which is odd considering that's significantly less than PCI's own limits... I guess there's some overhead going on there.
There were also many conflicts because of the shared bus. Who here hasn't had trouble with stuttering audio on a PCI soundcard because some other device in the bus was causing trouble?
PCI-X solved the bandwidth problem, but never really made it into the consumer space. And yet the bandwidth demands from consumers increasingly stressed the PCI bus.
Motherboard manufacturers eventually solved the problem by moving on-board gigabit controllers from the PCI bus to either a dedicated bus (Intel did this, I think), or a different general purpose bus (HyperTransport?). Such controllers saw an immense performance improvement.
RAID controllers are a good example because modern 7200RPM drives have peak (not burst, peak) read speeds of about 90MB/s at the outside of the disk. Put two of those on a RAID controller and you've got some serious bandwidth issues.
The official specs for eSATA essentially say the cable is better suited for external use. Here's what they have to say about it:
The external cable connector is a shielded version of the connector specificed in SATA 1.0a with these basic differences:
The External connector has no "L" shaped key, and the guide features are vertically offset and reduced in size. This prevents the use of unshielded internal cables in external applications.
To prevent ESD damage, the insertion depth is increased from 5mm to 6.6mm and the contacts are mounted further back in both the receptacle and plug.
To provide EMI protection and meet FCC and CE emission requirements, the cable has an extra layer of shielding, and the connectors have metal contact points.
There are springs as retention features built into the connector shield on both the top and bottom surfaces.
The external connector and cable are designed for over five thousand insertions and removals while the internal connector is only specified to withstand fifty.
They make it pretty clear exactly what's different. The biggest difference is the cable is shielded, while internal SATA is not (or less so?). And obviously the connector being rated for a hundred times more insertions is a pretty big difference.
I should note that in recent benchmarks done by MaximumPC, eSATA did not provide substantial performance benefits over Firewire800 drives. eSATA featured a higher burst speed, but more or less equivalent average transfer rates and seek times. Unless there were specific licensing issues with Firewire 800, I would rather have seen it become the preferred drive interface; I'll take a general purpose connector that I can use for other stuff over something as specific as eSATA any day, especially when eSATA provides little benefit.
While I'm not as certain when it comes to graphics cards, PCI-Express is/was sorely needed to replace PCI for general expansion cards. PCI had a limited amount of bandwidth available that was extremely easy to saturate (A single gigabit NIC would hit a bandwidth wall at something like 400mbit). The shared nature of the PCI bus was also an extremely frustrating limitation.
By contrast, PCI-Express 1x slightly increases the bandwidth from roughly 133MB/s to 150MB/s, but more importantly each device gets that, it's not shared anymore. And of course, 2x and higher slots provide more bandwidth.
But when it comes to graphics, AGP 8x was (at the time) providing more than enough bandwidth... as for the demands of modern monsterous graphics cards (such as the 8800 GTX), for all I know they might be able to saturate an AGP 8x bus.
As I understand it anyhow, the more tangible benefits from moving from AGP to PCI-Express were increase bi-directional bandwidth (AGP was great at Host->Card, but sucked at Card->Host), and increased ease with sticking multiple PCI-Express slots on the motherboard, making modern SLI possible.
Why would you need a hardphone to use Asterisk? Grab a soft phone that supports a high quality voice codec such as Speex (Asterisk supports it), connect the two Asterisk servers with IAX, and you're set.
;)
Personally I'd just grab two copies of Skype, forward the ports to minimize latency, and go at it; the quality and latency is good enough for live broadcasting when it is set up properly (again, with ports forwarded), and the quality between two properly configured Skype clients is significantly better than POTS. Sure, it won't be as good as Speex through Asterisk, but it is a tad easier and faster to set up, to say the last
They'll need some sort of emulation layer, but because they control the entire OS, they have the potential to do this with enormous performance.
:)
They can port all *their* code (and all the APIs) to the PowerPC. The only thing left running x86 code will be the actually application code. But since they'll be making calls to native libraries, you don't get the overhead of having to emulate the entire OS.
This is actually how Apple has done it all the way back to the 68k/PowerPC switchover. And during the PowerPC/x86 switchover, they benefited from the same thing. The reason that Rosetta (Or rather, Transitive's QuickTransit) is so fast emulating a PowerPC on an Intel processor (which is normally much more difficult than the other way around) is because Rosetta isn't emulating an entire OS like PearPC and other similar projects are attempting. Rosetta just emulates the application code, and the application's API calls are made to native APIs.
There is some trickery needed to make cross-platform API calls, of course, but the performance benefits are insane
Not that I know anything about implementation. But the observation of Apple's two arch migrations is proof enough.
It won't strictly double bandwidth usage. There are several methods that proxy servers can employ to reduce bandwidth usage.
The first is the use of a cache. For example, if the primary use of the proxy is to visit myspace, many of the images, stylesheets, scripts, and so on will be cached. That reduces the proxy to internet bandwidth.
Another is compressing text content that passes through the proxy. HTML, javascript, XML, and stylesheets can make up a surprising portion of bandwidth usage these days. By having the proxy compress these types of content as it passes through, user to proxy bandwidth can be reduced.
There are obviously other more destructive methods; GIF images can be run through gif2png and HTML rewritten to use the new images, JPG files can be recompressed at lower quality settings, all sorts of things can be done on-the-fly to reduce bandwidth use.
I'd imagine that because of the repetitive use, the bandwidth demands with some of the above mentioned techniques would be substantially less than double.
Of course, all this discussion is moot; the instant the school realizes that the kids are using this proxy, they'll block it, and that'll be that.
That's the wonderful thing about Linux. Don't like the mainstream, and want to keep using an ideologically pure distribution? Nobody is stopping you. There are a whole bunch of distributions that specialize in that, and there's nobody holding a gun to your head saying "You must use Ubuntu."
Of course, this is also one of the things that keeps Linux as a niche product on the desktop. But that's another topic.
5000 CPUs isn't particularly impressive either when you consider 1U quad core quad processor machines. Up to 672 CPU cores per rack. Pretty insane power requirements as far as density is concerned, though.
Let's not forget that the Sun X4500 is a quad-processor beast itself, and you're not going to saturate all 16 (possible) cores with storage demands, unless you're processing the data locally.
A more dedicated processor such as the Cell (which would seem to be ideally suited for software rendering) would reduce the space/power/heat requirements significantly, although I'm not sure of what kind of enclosure you'd need. Assuming you're not shoving PS3s into a datacenter, the Cell comes on PCI-X cards. I'm sure there is some enclosure designed to maximize the number of PCI-X cards in a limited space.
I don't think their datacenter is bad, by any means, just that the technology exists to do what they're doing in much less space (and possibly cheaper).
Did you ever consider that many companies centralize the important data into shared user directories? A redundant centralized server with proper backups are much simpler than setting up dedicated backup solutions on EVERY workstation.
So, reinstalling the OS from scratch on a workstation certainly is a good way to perform disaster recovery; the workstation is borked, and all the user settings are server-side, so why NOT nuke the workstation?
Of course, such a company would probably also install the workstations from a ghost image. However I work for a company that does go the centralized route and yet doesn't use ghost images (we have an instruction list of what to install and how to set the machine up).
More like, in New York City, 10 million people commute to work each day (making up figures here), each traveling down the road at 50km/h. So the aggregate commute speed in New York City is 46% the speed of light!
300TB is still nothing, even with insane redundancy.
For example, assume you've got Sun Fire X4500 servers, and that you take them with the stock 500GB drives. 24TB per 4U server.
Now let's assume that you run RAIDZ2 on each server, dedicating 2 of the 48 drives to parity. That's 23TB per server. Now let's assume you want some redundancy, as in, completely separate failover capacity. You mirror every single server with hot standbys. Easy with ZFS, you can mirror the file system in real time without any issues. So we get an effective 11.5TB per 4U server.
Now, we need 300TB, so that requires 27 servers. 108U, or three 42u racks, with each rack having 6u for network infrastructure, load balancers to do the failover to the hot-standby units (or even localized UPS if you want double UPS coverage).
3 racks. With pretty decent redundancy that allows two drives per server to fail before you need to switch over to the hot standbys, and that's ignoring the fact that each X4500 server has redundant power supplies and allows hot swapping of drives and power supplies. THREE RACKS.
Explain to me why they need a 10,000 sq. ft. datacenter for three racks worth of servers? Yeah, I know, AC, UPS, generators, all that stuff takes up room, but I could fit a lot more than 300TB in a 10k sq. ft. datacenter.
Sure, I'm no expert, and I might be missing some important considerations, but even I could set up a 300TB high-uptime file storage network without needing an entire datacenter to do it.
We have the multiple election days in Canada (Three of them, Federal, Provincial, and Municipal. Referendums are extremely rare and often happen only every decade or two, and are never country-wide). Unless I'm mistaken, our voter turnout is HIGHER than the US, despite having three election days instead of one.
I wouldn't want it any other way. On Federal election night, I'm glued to the TV watching the results. There's enough information being thrown at us with just one election, why would I want three elections at the same time?!
Of course, we don't have fixed elections in Canada. Federal elections happen whenever the current government wants to, with exceptions such as there being a max length of a term, and the government collapsing due to a vote of non-confidence (which happens on average every 18 months in a minority government, of which our current government is one). So a single election day couldn't possibly work with out voting system.
Municipal elections are still held on a fixed schedule, though. Since municipalities are strictly the responsibility of the province, and not the federal government, each province sets their own municipal election dates, on a fixed schedule.
Erased markings? Not dark enough? Not problems. The pencils provided at voting booths are both more than dark enough, and lack erasers. If you're saying that ballots might be erased after the fact, this is prevented by procedure; counting is done under supervision, there is no chance to erase a ballot.
In Canada, with I'm sure a few exceptions, the simple "make a mark in the circle with the pencil" system works perfectly fine. In federal elections, at least, there haven't been (to my knowledge) any serious recent problems with miscounts and fraud issues.
One thing you have to keep in mind is that our ballots here are incredibly simpler than the US. In the US, a federal ballot has dozens of candidates for a huge number of posts, along with propositions thrown in.
In a Canadian election ballot, there are usually roughly 3 to 5 names on the ballot, each with a large white circle that is perhaps an inch wide. You put a mark in the circle with the pencil.
Vote counting in Canada after a federal election is usually completed in an hour or so, although some ridings can take longer, and different polls close at different times due to the timezones. Still, the final results are almost always known by the next day. This is why I find it very strange that in the US, counting votes for federal elections takes so much longer. I suppose it is because of the more complicated multi-choice ballots.
The thing is, the US would never request extradition over a civil matter. This was the result of a lawsuit, not a criminal case.
And notice also that in the text above it says "The sponsor is adware". It can't be clearer. It up and TELLS you it's adware in exactly those words, and gives the option to not install it.
Huh? I'm a user of MsgPlus. Every time I've ever installed it or an update, it has given me a CHOICE as to if I want to install the toolbar. There is an ENTIRE page of the installer dedicated to explaining what it is, and giving the option as to if you want to install it.
The guy's been working full time on a free product, and STILL gives the option not to install it. I don't know about you, but I trust somebody like him a lot more than the people at Debian who want to rename Firefox because the icon is free-as-in-beer instead of free-as-in-speech. At least Patchou is being reasonable.