I would certainly hope your servers have more than 16 gigabits of RAM.
As with most Intel chipset releases, there really isn't much to get excited about here. "Native" USB3 might be exciting to someone somewhere, but from a practical standpoint we've been getting USB3 on motherboards for the last couple years anyway, and extra PCIe lanes are for the most part only interesting to nutball gamers.
Z-series chipsets are enthusiast products. Basically all this stuff just integrates features that are already on $150+ motherboards into Intel's chips.
The full size of my backup is currently 33TB. It's almost all video, so there's no deduplication or meaningful compression that can be done. What I had been doing was slicing the storage volumes up so I could have two or three copies of my data. I had four machines and 64 physical drives in a 1000 square foot apartment (and one room that was not less than 90F year round...) devoted to all of this. It didn't make sense to add more, larger consumer drives to the mix - I'd just have to find a place to put them, power them and cool them and to do so within the limitations of the space that I occupy.
Since I have my tape changer, I've been able change my storage strategy to better and more securely accommodate my need to expand my available storage. I still have my data in nice, redundant zpools, but now I don't feel like I need to buy multiple disks for every bit of data I want to store, I have a worthwhile off-site storage option and it's costing me less than adding spinning disks would.
About a year ago, staring at never-ending rsyncs between four boxes containing ~12TB of data apiece, I decided that it would be cheaper and easier for me to move to tape rather than continually duplicate data across RAID5 volumes and hope I never have a disk failure and a hard error on any of the remaining drives. I managed to get a Quantum Superloader (LTO4) and a dozen tapes for about $1600. There has been a learning curve with the setup, but there's just no other practical way to deal with tens of terabytes of information.
I was able to move to a single storage machine and switch off a bunch of noisy, hot, power-hungry systems. I was glad to make the switch and I wish I had done it sooner,.
Ages ago I was teaching a bunch of people how to work scanners in a training session. We scanned a whole bunch of stuff and most people were clearly able to see that commercially printed content doesn't look appreciably different when scanned at 600dpi or 1200dpi. Eventually I had the bright idea to try to scan a $20 bill since they're actually fine fabric and not paper. It scanned fine at 600dpi and previewed OK at higher settings, but every time I tried to scan it at a higher setting, the area of the bill would be replaced by black pixels in the finished image. My students and I decided it was probably an anti-counterfeiting measure and after about 40 minutes of experimentation with things like discoloring the bills, tearing them so they no longer resembled whole bills (we used a couple $1s for that), zooming in on small areas etc. we determined that whatever was going on was actually pretty tough to fool.
A better remote doesn't fix the on-screen interface issues with the various devices in question. Cable box + Tivo + DLNA client (WD TV Live/Boxee/Popcorn Hour/whatever) = user interface hell no matter what remote you have.
Personally, my media viewing is handled through XBMC with a cheap keyboard/trackpad thing. I never change any of the settings on my receiver. I just run everything through the computer. That means that I've forgone live TV entirely, but everything I want is already either available in my apartment or being obtained via some extralegal method. XMBC gives me consistency and is extensible for my needs, but of course it's fundamentally menu and pointer based and only works for the duration of time that I can stay in XBMC (unless I want to deal with less-than-ideal desktop interfaces on the host machines, I suppose). Which doesn't help at all if I want to listen to an SACD or something.
So anyway, my initial thought was that something Kinect might be a good step in the right direction. Presumably we could come up with a decent set of device control gestures that could be made to work with a wide variety of devices. They could be communicated on screen. Voice recognition? I agree that it's probably silly.
The better thing would be some kind of overlay that sits on top of everything else and can order devices about as need be. DLNA is supposed to do some of that (press mute on the DVD remote and in theory it mutes whatever is making sounds, whether it's the TV or a receiver or a soundbar, assuming that device speaks DLNA as well), but it's inconsistent and doesn't fix other interface irregularities.
Sadly, I don't think media production is ready for it. I'd be happy if 4k displays remained exotic or stuck at desktop PC sizes if it meant that 1080p were widespread all the way down to tablet size displays.
The article is right about a couple things: TV UIs suck and remotes suck even more.
My mom can't operate a modern TV. I mean like not AT ALL. If it's anything more challenging than volume up or down, it's too much. She doesn't get it. There's a bunch of stuff we plug in and want to use now - DLNA clients, DVRs, Home Theater receivers, cable boxes, game machines - and it all works differently and needs some stupid or weird different control, both on-screen and in terms of the control device. The revolution will be the people who make some kind of master overlay and master remote (I love my Harmony but it doesn't go far enough) that handles everything.
Maybe that means a mic or a kinect that lets us talk or gesture. Maybe it means having a little display on a tablet. I don't know. I just know that what we have now is a huge mess.
I suspect they're coming that way from whomever is supplying my vendors. I've seen "new" drives containing data from a wide variety of vendors over the years including Amazon, Newegg, Provantage and CDW. For all I know they're coming out of Seagate or Samsung's factory that way. I buy drives in large enough quantities to get sealed cases full of drives rather than random one-off units someone shoved in a static bag and wrapped in bubble wrap, so I tend to think the boxes I'm getting have probably been unmolested since they got unloaded from whatever boat they came off of in California.
I've gotten drives I purchased as new from Amazon and Newegg with exsiting Windows installations on them. In fact, I'd say I see it maybe once in every 30 drives I get. I buy enough drives that I see six or seven such drives in a typical year. Once I got a drive that was clearly part of a Windows SoftRAID before I formatted it.
Personally, I send those drives back. They clearly aren't new and they're not fit for sale in that state. I'm not paranoid enough to go looking at the SMART data for power on hours but when I run across drives like that it makes me think I should. Amazon will pay return shipping on drives in that condition. That is a good reason to buy drives from Amazon.
I've had the good fortune to review a number of tablets in the last few weeks. Personally, I like the 7" form factor and prefer Samsung's Galaxy line over others I've tried, but given the requirements I think I'd probably take a long, hard look at a Lenovo Thinkpad Tablet. They're the only tablets I've seen with a USB root hub and their screens, while not quite as good as Samsung's, is better than the the Motorola, Acer, Toshiba and Asus 10" tablets I've tried. The Tegra CPU isn't top of the line but in my experience that's rarely a limiting factor for top-tier Android tablets. I also found myself really appreciating the pen input.
For the most part, the experience of one tablet vs. another is very similar, but there is quite a lot of variability in screen quality and there's actually a some differences in weight and thickness that might or might not be relevant. As with all Thinkpad-branded products, I also found the build quality to be impressively high
I think the answer is "it depends what you want to do." I had the opportunity to purchase and review 9 different media streaming devices. This is what I came away with:
WD TV Live HD: Has a wonderful interface for streaming services like Youtube and Pandora, and it does DLNA very well. Very good support for music without playlists (many DLNA streamers just want to play single tracks or tracks on a DLNA-shared playlist). But I couldn't get one to negotiate a 1080p connection with my bog-standard 1080p Toshiba TV despite switching cables and several different firmware revisions on two different units. I also found that navigating the user interface was slow and occasionally non-responsive.
Seagate has a very similar set of products. I only took a cursory look at a cheap FreeAgent Theater, found that it errored out when I tried to stream something from my DLNA server and set it aside.
Roku XD: Streams the stuff it streams. Doesn't technically do DLNA or allow access to local network content, though there is a DLNA-like application you can run on a Windows machine to sorta stream some content. I was really unimpressed. But at least it does 1080p.
Popcorn Hour C200: DLNA client, DLNA server. Has a local drive bay and USB ports to add storage. Supports every audio and video format known to man. Will auto-rip content if you put an optical drive in it. Has a great collection of free streaming services and network connectivity options... and a dreadfully slow user interface and no access to any sort of premium streaming options.
Boxee: I actually like the UI and social hooks. I like the variety of support for network connectivity and the overall speed of the UI. But oddly enough, it doesn't do UPNP or DLNA, something I ulimately confirmed with DLink. It has excellent file format support, a decent user interface for network browsing and for music playback, and I love the remote, but there are enough weird drawbacks that it feels like an unfinished product, especially at $150 or so. I'll also say that the Windows version of the software would regularly spike an i7-2600 to 100% CPU utilization across all cores, all by itself.
LG Smart TV Upgrader (Sony and several other companies sell identical devices): Accesses premium streaming services just fine. Supports Plex, as of the most recent update; the one at my parents' house can stream movies from my apartment 700 miles away. Fuss-free DLNA support and it kind-of manages SMB support as well. Dirt cheap, but the UI is ugly, slow and somewhat non-intuitive. Music support is particularly crappy and the remote is not that good either. Still, for $50, they do what I want them to to do.
Vortexbox: I set up a Vortexbox, thinking it would be an STB solution. It's a DLNA/AFP/SMB server that auto-rips stuff to FLAC or MKV and makes it available to other systems on a LAN. It's meant to be appliance-like. It worked OK for its intended purpose, but to my annoyance the install scripts assume there's only one storage drive and don't make provisions for expansion, meaning that it's basically a less-functional version of a Popcorn Hour C200.
PS3/Xbox360. Theses things can be used as streaming clients with obnoxious control devices and poor user interfaces. I'm sure it's great if you're used to it, but I found them lacking.
The great unknown for me is the Logitech Revue. They're cheap now, and I understand that they're essentially Android 3.1 devices. I would assume that I can get any sort of premium streaming on them, since all that stuff works on my Android phone, and I should have my pick of third-party media players if I don't like the ones it ships with. I don't know about the actual TV integration, but I don't really care about that aspect either.
At the end of the day, I liked the LG Smart TV Upgrader better than the others. I had too many problems with the WD TV Live HD and the Boxee needed to cost about half what it does in order to be competitive. Maybe this new focus on the hardware will fix some of the issues.
The thing that's been discussed along those lines is that Apple had a couple years to get its supply chain in order before the release of the iphone. Everybody else has been playing catch up by trying to combine off the shelf components in the most appealing fashion they can. Since no one has really found a magic recipe for building a perfect Android/WinMo phone and everyone wants to differentiate their product somehow, economies of scale and supply chain efficiencies haven't really been working in anyone's favor.
You can't? I buy them pretty regularly. If you build and sell systems, there's still a price premium for USB vs. PS/2 input devices. I'd rather buy a quality Logitech keyboard with a PS/2 plug on the end than some crappy generic that just happens to be USB. The sets I'm using these days have a PS/2 keyboard and a USB mouse and I think I pay under $75 for a box of 10 new ones.
Most mini-ITX stuff is cheap enough to just toss if there's a problem. I have a few dual core Atom ITX systems out in the world. I paid $50 for the boards with CPU. Intel branded boards, even. There's nothing else with a reasonably current CPU available at that price point and if I bought one now I'm sure I'd get an incrementally faster Atom anyway. It's not worth getting bent out of shape.
RS232 is still handy for some home AV applications. Older TVs and projectors are more likely to have VGA than either DVI or HDMI, and I don't see a problem with a dedicated port for a keyboard and mouse, especially given how common PS/2 devices are.
This is being positioned as a hobbyist platform, same as LGA1366. The affordable E-series (i7-type) Xeons don't boot on consumer-class motherboards and don't have chipset support SMP though. These guys are the only game in town for people who want to stick three video cards in something and get a top notch CPU to go with it.
As someone with a decent investment in LGA1366 stuff, I'd rather play it smart and keep everything on the "mainstream" LGA1155 for anything but the six core CPUs. The motherboards are harder to find and substantially more expensive for the dubious value of having some extra PCIe lanes and a couple extra DIMM slots.
I'm in the process now of selling off my LGA1366 machines while they still have value and replacing them with Xeon E-series equipment.
Your estimates are off somewhat, but yes, it's not a cheap hobby to have. I've probably put about $18,000 in hardware over the last 5 years, with the first of my older systems being by far most expensive. Most of it is sitting in or on top of an APC Netshelter in a closet in my extra bedroom. It all runs on one 20A circuit, albeit one that really isn't used for much of anything else.
My server systems are built with commodity parts other than disk controllers (Dell Perc5s or IBM ServerRAID M1015s) and chassis (generic Norco 20-bay units that have a SATA/SAS backplane and usually run $350 or so), rather than the much more expensive first tier OEM machines or even barebones Supermicro or Tyan rigs. This keeps cost and more importantly noise down to manageable levels compared to some of the things you're suggesting.
For what it's worth, I really don't have much else going on in my life, which makes this sort of thing possible if not exactly practical. I think it's deeply cool that I have a personal application for this kind of equipment, and it gives me something to do as a techie hobbyist besides building a newer, faster desktop every six months. I suppose the comment about the affordability of the equipment is a little bit off-side. Plenty of reasonably well paid, unmarried IT guys and engineers buy themselves impractical $50,000 cars and no one bats an eye at that or casts aspersions at the habits of their consumption.
The total size of data that actively needs to be backed up is a hair over 30TB at this point. Yes, I could move to a system of using some number of hard disk drives, but there are some mitigating factors there:
1. Drive mechanics are excessively delicate, especially for the low-cost "green" drives. Given that engineers spend man-weeks trying to shave fractions of a cent off the per-unit cost of each drive, what are they doing to make "green" drives significantly cheaper than their full-speed siblings?
I'm using Hitachi 3TB 0S03230 drives, which ARE 5400rpm "green" drives in my main storage server now, but had I not had access to my tape library, I don't think I would have felt comfortable enough to buy them in the first place.
2. Drives in quantity aren't really all that portable. They're not going off site. My second set of tapes lives at my office. I run an incremental backup job every weekend and swap full sets every couple months, so I'm more or less carting backup media around all the time. I wouldn't want to do that with hard drives.
3. One of the reasons I moved to tape was to dramatically reduce my power consumption. Moving from four file servers to one-and-a-half (I have one other machine that's been repurposed for non-media needs) cut my power bill by around 40%. And no, I don't ever bother to heat my home in the winter.
At the moment, I need 38 tapes to get a full backup. I don't bother with hardware compression since my data is practically all video files. The tapes are around $20 apiece. I have 86 of them, most of which came with my autoloader, which covers me for incremental backups and a few spares in case I have a bad tape (I've only had one so far). I haven't tested a full restore due to the massive volume of data involved, but I was able to spot-recover 5TB from a full backup set without any issues. I don't really pay all that much attention to how long backups or restores take (though, gotta say, LTO is surprisingly fast). I basically load in my 16 tapes and come back in 12 hours to load in 16 more.
Slashdot Mirror
I would certainly hope your servers have more than 16 gigabits of RAM.
As with most Intel chipset releases, there really isn't much to get excited about here. "Native" USB3 might be exciting to someone somewhere, but from a practical standpoint we've been getting USB3 on motherboards for the last couple years anyway, and extra PCIe lanes are for the most part only interesting to nutball gamers.
Z-series chipsets are enthusiast products. Basically all this stuff just integrates features that are already on $150+ motherboards into Intel's chips.
I give this a hearty "meh."
That... runs counter to my experiments. What backup product were you using to obtain those kinds of results?
The full size of my backup is currently 33TB. It's almost all video, so there's no deduplication or meaningful compression that can be done. What I had been doing was slicing the storage volumes up so I could have two or three copies of my data. I had four machines and 64 physical drives in a 1000 square foot apartment (and one room that was not less than 90F year round...) devoted to all of this. It didn't make sense to add more, larger consumer drives to the mix - I'd just have to find a place to put them, power them and cool them and to do so within the limitations of the space that I occupy.
Since I have my tape changer, I've been able change my storage strategy to better and more securely accommodate my need to expand my available storage. I still have my data in nice, redundant zpools, but now I don't feel like I need to buy multiple disks for every bit of data I want to store, I have a worthwhile off-site storage option and it's costing me less than adding spinning disks would.
About a year ago, staring at never-ending rsyncs between four boxes containing ~12TB of data apiece, I decided that it would be cheaper and easier for me to move to tape rather than continually duplicate data across RAID5 volumes and hope I never have a disk failure and a hard error on any of the remaining drives. I managed to get a Quantum Superloader (LTO4) and a dozen tapes for about $1600. There has been a learning curve with the setup, but there's just no other practical way to deal with tens of terabytes of information.
I was able to move to a single storage machine and switch off a bunch of noisy, hot, power-hungry systems. I was glad to make the switch and I wish I had done it sooner,.
Ages ago I was teaching a bunch of people how to work scanners in a training session. We scanned a whole bunch of stuff and most people were clearly able to see that commercially printed content doesn't look appreciably different when scanned at 600dpi or 1200dpi. Eventually I had the bright idea to try to scan a $20 bill since they're actually fine fabric and not paper. It scanned fine at 600dpi and previewed OK at higher settings, but every time I tried to scan it at a higher setting, the area of the bill would be replaced by black pixels in the finished image. My students and I decided it was probably an anti-counterfeiting measure and after about 40 minutes of experimentation with things like discoloring the bills, tearing them so they no longer resembled whole bills (we used a couple $1s for that), zooming in on small areas etc. we determined that whatever was going on was actually pretty tough to fool.
No one who has anything to do with itunes has any business accepting awards for anything.
A better remote doesn't fix the on-screen interface issues with the various devices in question. Cable box + Tivo + DLNA client (WD TV Live/Boxee/Popcorn Hour/whatever) = user interface hell no matter what remote you have.
Personally, my media viewing is handled through XBMC with a cheap keyboard/trackpad thing. I never change any of the settings on my receiver. I just run everything through the computer. That means that I've forgone live TV entirely, but everything I want is already either available in my apartment or being obtained via some extralegal method. XMBC gives me consistency and is extensible for my needs, but of course it's fundamentally menu and pointer based and only works for the duration of time that I can stay in XBMC (unless I want to deal with less-than-ideal desktop interfaces on the host machines, I suppose). Which doesn't help at all if I want to listen to an SACD or something.
So anyway, my initial thought was that something Kinect might be a good step in the right direction. Presumably we could come up with a decent set of device control gestures that could be made to work with a wide variety of devices. They could be communicated on screen. Voice recognition? I agree that it's probably silly.
The better thing would be some kind of overlay that sits on top of everything else and can order devices about as need be. DLNA is supposed to do some of that (press mute on the DVD remote and in theory it mutes whatever is making sounds, whether it's the TV or a receiver or a soundbar, assuming that device speaks DLNA as well), but it's inconsistent and doesn't fix other interface irregularities.
Sadly, I don't think media production is ready for it. I'd be happy if 4k displays remained exotic or stuck at desktop PC sizes if it meant that 1080p were widespread all the way down to tablet size displays.
The article is right about a couple things: TV UIs suck and remotes suck even more.
My mom can't operate a modern TV. I mean like not AT ALL. If it's anything more challenging than volume up or down, it's too much. She doesn't get it.
There's a bunch of stuff we plug in and want to use now - DLNA clients, DVRs, Home Theater receivers, cable boxes, game machines - and it all works differently and needs some stupid or weird different control, both on-screen and in terms of the control device. The revolution will be the people who make some kind of master overlay and master remote (I love my Harmony but it doesn't go far enough) that handles everything.
Maybe that means a mic or a kinect that lets us talk or gesture. Maybe it means having a little display on a tablet. I don't know. I just know that what we have now is a huge mess.
I suspect they're coming that way from whomever is supplying my vendors. I've seen "new" drives containing data from a wide variety of vendors over the years including Amazon, Newegg, Provantage and CDW. For all I know they're coming out of Seagate or Samsung's factory that way. I buy drives in large enough quantities to get sealed cases full of drives rather than random one-off units someone shoved in a static bag and wrapped in bubble wrap, so I tend to think the boxes I'm getting have probably been unmolested since they got unloaded from whatever boat they came off of in California.
I've gotten drives I purchased as new from Amazon and Newegg with exsiting Windows installations on them. In fact, I'd say I see it maybe once in every 30 drives I get. I buy enough drives that I see six or seven such drives in a typical year. Once I got a drive that was clearly part of a Windows SoftRAID before I formatted it.
Personally, I send those drives back. They clearly aren't new and they're not fit for sale in that state. I'm not paranoid enough to go looking at the SMART data for power on hours but when I run across drives like that it makes me think I should. Amazon will pay return shipping on drives in that condition. That is a good reason to buy drives from Amazon.
Seriously?
Boot while holding down Apple-S /var/db/.AppleSetupDone
mount -uw /
rm
shutdown -h now
Bam. Administrator access and all the password resetting glory you need thereafter.
I don't even have a Mac and I know how to do it. How fucking easy does it need to be?
I've had the good fortune to review a number of tablets in the last few weeks. Personally, I like the 7" form factor and prefer Samsung's Galaxy line over others I've tried, but given the requirements I think I'd probably take a long, hard look at a Lenovo Thinkpad Tablet. They're the only tablets I've seen with a USB root hub and their screens, while not quite as good as Samsung's, is better than the the Motorola, Acer, Toshiba and Asus 10" tablets I've tried. The Tegra CPU isn't top of the line but in my experience that's rarely a limiting factor for top-tier Android tablets. I also found myself really appreciating the pen input.
For the most part, the experience of one tablet vs. another is very similar, but there is quite a lot of variability in screen quality and there's actually a some differences in weight and thickness that might or might not be relevant. As with all Thinkpad-branded products, I also found the build quality to be impressively high
I think the answer is "it depends what you want to do." I had the opportunity to purchase and review 9 different media streaming devices. This is what I came away with:
WD TV Live HD: Has a wonderful interface for streaming services like Youtube and Pandora, and it does DLNA very well. Very good support for music without playlists (many DLNA streamers just want to play single tracks or tracks on a DLNA-shared playlist). But I couldn't get one to negotiate a 1080p connection with my bog-standard 1080p Toshiba TV despite switching cables and several different firmware revisions on two different units. I also found that navigating the user interface was slow and occasionally non-responsive.
Seagate has a very similar set of products. I only took a cursory look at a cheap FreeAgent Theater, found that it errored out when I tried to stream something from my DLNA server and set it aside.
Roku XD: Streams the stuff it streams. Doesn't technically do DLNA or allow access to local network content, though there is a DLNA-like application you can run on a Windows machine to sorta stream some content. I was really unimpressed. But at least it does 1080p.
Popcorn Hour C200: DLNA client, DLNA server. Has a local drive bay and USB ports to add storage. Supports every audio and video format known to man. Will auto-rip content if you put an optical drive in it. Has a great collection of free streaming services and network connectivity options... and a dreadfully slow user interface and no access to any sort of premium streaming options.
Boxee: I actually like the UI and social hooks. I like the variety of support for network connectivity and the overall speed of the UI. But oddly enough, it doesn't do UPNP or DLNA, something I ulimately confirmed with DLink. It has excellent file format support, a decent user interface for network browsing and for music playback, and I love the remote, but there are enough weird drawbacks that it feels like an unfinished product, especially at $150 or so. I'll also say that the Windows version of the software would regularly spike an i7-2600 to 100% CPU utilization across all cores, all by itself.
LG Smart TV Upgrader (Sony and several other companies sell identical devices): Accesses premium streaming services just fine. Supports Plex, as of the most recent update; the one at my parents' house can stream movies from my apartment 700 miles away. Fuss-free DLNA support and it kind-of manages SMB support as well. Dirt cheap, but the UI is ugly, slow and somewhat non-intuitive. Music support is particularly crappy and the remote is not that good either. Still, for $50, they do what I want them to to do.
Vortexbox: I set up a Vortexbox, thinking it would be an STB solution. It's a DLNA/AFP/SMB server that auto-rips stuff to FLAC or MKV and makes it available to other systems on a LAN. It's meant to be appliance-like. It worked OK for its intended purpose, but to my annoyance the install scripts assume there's only one storage drive and don't make provisions for expansion, meaning that it's basically a less-functional version of a Popcorn Hour C200.
PS3/Xbox360. Theses things can be used as streaming clients with obnoxious control devices and poor user interfaces. I'm sure it's great if you're used to it, but I found them lacking.
The great unknown for me is the Logitech Revue. They're cheap now, and I understand that they're essentially Android 3.1 devices. I would assume that I can get any sort of premium streaming on them, since all that stuff works on my Android phone, and I should have my pick of third-party media players if I don't like the ones it ships with. I don't know about the actual TV integration, but I don't really care about that aspect either.
At the end of the day, I liked the LG Smart TV Upgrader better than the others. I had too many problems with the WD TV Live HD and the Boxee needed to cost about half what it does in order to be competitive. Maybe this new focus on the hardware will fix some of the issues.
The thing that's been discussed along those lines is that Apple had a couple years to get its supply chain in order before the release of the iphone. Everybody else has been playing catch up by trying to combine off the shelf components in the most appealing fashion they can. Since no one has really found a magic recipe for building a perfect Android/WinMo phone and everyone wants to differentiate their product somehow, economies of scale and supply chain efficiencies haven't really been working in anyone's favor.
You can't? I buy them pretty regularly. If you build and sell systems, there's still a price premium for USB vs. PS/2 input devices. I'd rather buy a quality Logitech keyboard with a PS/2 plug on the end than some crappy generic that just happens to be USB. The sets I'm using these days have a PS/2 keyboard and a USB mouse and I think I pay under $75 for a box of 10 new ones.
Most mini-ITX stuff is cheap enough to just toss if there's a problem. I have a few dual core Atom ITX systems out in the world. I paid $50 for the boards with CPU. Intel branded boards, even. There's nothing else with a reasonably current CPU available at that price point and if I bought one now I'm sure I'd get an incrementally faster Atom anyway. It's not worth getting bent out of shape.
RS232 is still handy for some home AV applications. Older TVs and projectors are more likely to have VGA than either DVI or HDMI, and I don't see a problem with a dedicated port for a keyboard and mouse, especially given how common PS/2 devices are.
This is being positioned as a hobbyist platform, same as LGA1366. The affordable E-series (i7-type) Xeons don't boot on consumer-class motherboards and don't have chipset support SMP though. These guys are the only game in town for people who want to stick three video cards in something and get a top notch CPU to go with it.
As someone with a decent investment in LGA1366 stuff, I'd rather play it smart and keep everything on the "mainstream" LGA1155 for anything but the six core CPUs. The motherboards are harder to find and substantially more expensive for the dubious value of having some extra PCIe lanes and a couple extra DIMM slots.
I'm in the process now of selling off my LGA1366 machines while they still have value and replacing them with Xeon E-series equipment.
The i7-2700k will have a launch MSRP of $331.
I fully expect I'll be able to get one at Microcenter for $280 or so.
Your estimates are off somewhat, but yes, it's not a cheap hobby to have. I've probably put about $18,000 in hardware over the last 5 years, with the first of my older systems being by far most expensive. Most of it is sitting in or on top of an APC Netshelter in a closet in my extra bedroom. It all runs on one 20A circuit, albeit one that really isn't used for much of anything else.
My server systems are built with commodity parts other than disk controllers (Dell Perc5s or IBM ServerRAID M1015s) and chassis (generic Norco 20-bay units that have a SATA/SAS backplane and usually run $350 or so), rather than the much more expensive first tier OEM machines or even barebones Supermicro or Tyan rigs. This keeps cost and more importantly noise down to manageable levels compared to some of the things you're suggesting.
For what it's worth, I really don't have much else going on in my life, which makes this sort of thing possible if not exactly practical. I think it's deeply cool that I have a personal application for this kind of equipment, and it gives me something to do as a techie hobbyist besides building a newer, faster desktop every six months. I suppose the comment about the affordability of the equipment is a little bit off-side. Plenty of reasonably well paid, unmarried IT guys and engineers buy themselves impractical $50,000 cars and no one bats an eye at that or casts aspersions at the habits of their consumption.
The total size of data that actively needs to be backed up is a hair over 30TB at this point. Yes, I could move to a system of using some number of hard disk drives, but there are some mitigating factors there:
1. Drive mechanics are excessively delicate, especially for the low-cost "green" drives. Given that engineers spend man-weeks trying to shave fractions of a cent off the per-unit cost of each drive, what are they doing to make "green" drives significantly cheaper than their full-speed siblings?
I'm using Hitachi 3TB 0S03230 drives, which ARE 5400rpm "green" drives in my main storage server now, but had I not had access to my tape library, I don't think I would have felt comfortable enough to buy them in the first place.
2. Drives in quantity aren't really all that portable. They're not going off site. My second set of tapes lives at my office. I run an incremental backup job every weekend and swap full sets every couple months, so I'm more or less carting backup media around all the time. I wouldn't want to do that with hard drives.
3. One of the reasons I moved to tape was to dramatically reduce my power consumption. Moving from four file servers to one-and-a-half (I have one other machine that's been repurposed for non-media needs) cut my power bill by around 40%. And no, I don't ever bother to heat my home in the winter.
At the moment, I need 38 tapes to get a full backup. I don't bother with hardware compression since my data is practically all video files. The tapes are around $20 apiece. I have 86 of them, most of which came with my autoloader, which covers me for incremental backups and a few spares in case I have a bad tape (I've only had one so far). I haven't tested a full restore due to the massive volume of data involved, but I was able to spot-recover 5TB from a full backup set without any issues. I don't really pay all that much attention to how long backups or restores take (though, gotta say, LTO is surprisingly fast). I basically load in my 16 tapes and come back in 12 hours to load in 16 more.
Also, don't you, or didn't you at one time work with the folks who run Voyeurweb?