Slashdot Mirror

Elsevier may have enough clout with take-down notices to 3rd-party service providers (and might be able to weather the backlash blizzard that will follow) -- but not with institutions self-archiving their own research output. I take this as yet another cue to push 100% for immediate institutional deposit mandates and the Button from all institutions and funders. Since 2004 Elsevier formally recognizes their authors' right to do immediate, unembargoed OA self-archiving on their institutional website. And even if they ever do try to rescind that, closed-access deposit is immune to take-down notices. (But I don't think Elsevier will dare arouse that global backlash by rescinding its 9-year policy of endorsing unembargoed Green OA -- they will instead try to hope that they can either bluff authors off with their empty-double-talk about "systematicity" and "voluntariness" or buy their institutions off by sweetening their publication deal on condition they don't mandate Green OA) See: http://j.mp/OAngelic

We already knew at the time that Windows Phone wasn't taking off.

Obviously it depends on where you are at. Here are some older tests I found done with http://opensignalmaps.com/ Android app
http://j.mp/kfQCtR

Thorium: http://j.mp/sGoUz0

I looked around for data when writing this story, The Evolution of Digital Natives -- The Touch Generation http://j.mp/pyKxXF, and quotes from medical experts seemed to be based on old data. Appears nothing has changed. Meanwhile, children are looking at magazines and touching them expecting them to act like iPads http://aol.it/o4eCnZ.

"When you look at our car, you think that the only technology on it is the solar panels, but that's definitely not the case," said team manager Emil Hewage. "This car actually drives and thinks at the same time. It requires masses and masses of data to run the car properly." Video and photos http://j.mp/rdvmgJ.

I've recently started using Tundra, a very light-weight and scalable (CPU core wise) build system, with very fast and reliable dependency analysis written in C, using pthreads/winthreads and taking advantage of SIMD instruction sets. It supports multiple build configurations and variants side-by-side, multiple disjoint build targets/deliverables, and has very DRY LUA-based DSL for project configuration. It doesn't hide CFLAG details, it gets out of the way and lets you precisely control what you're passing to the compiler for each tool chain you want to support. It's released under the GPLv3, was written by a DICE employee and is being used to build Battlefield 3.

I was able to get rid of ~25 CMakeLists.txt that made up thousands of lines of text and replaced it with a single ~350 line tundra build file, in building a 200,000+ LoC multi-platform project and combined with a very stripped down autoconf script for environment detection and config.h generation, was able to support all of my requirements. I had to hack and hack to get CMake to try to support what I needed.

https://github.com/deplinenoise/tundr,a
http://j.mp/emwZ9Z for the PDF presentation slides.

Disclaimer: I have no relations with the author of Tundra or EA/DICE.

Well obviously one of them will be along presently to point out that "nobody knows" how well WP7 is doing since release because Microsoft won't tell us. Since I know, I may as well nip that one in the bud: Abysmal is not an exaggeration. Panglozz has been scraping the Facebook user statistics weekly since November for all the major phone platforms, and has assembled that delightful analytical spreadsheet that tells us week-by-week how it's doing relative to other platforms.

Facebook user stats may not be perfect, but it's a huge sample and lines up perfectly with other reports, which seem to be bending over backwards to avoid stating the obvious truth. The phone is not selling. After six months WP7 total facebook users don't add up to two days worth of increase in iPhone and Android platforms. The user base is not there, and ultimately that's what developers care about. They don't care if it's fun to write apps for the phone. They care if there are users to use the apps - and there aren't enough to speak of. The trend is clearly in decline, so not only are the users not there, they're not ever going to be there. Writing Windows Phone apps is not going to be profitable for nearly any developer, and it's not going to make them famous either. Nokia can't save this.

Some of the numbers we've seen for WP7 are totally bogus. Obviously if nearly three times as many people downloaded the software development kit for WP7 as use WP7 for Facebook, something is amiss. Phone software development is not a 3x more common activity than Facebook posting. Somebody is trying to make it look like the thing is more popular than it actually is - perhaps by including the WP7 SDK with some other tools.

Which makes me glad that Panglozz is keeping track of this for us. It may be a little bit OCD, but it's helpful.

You may as well see the data too. Estimate of WP7 sales through scraping Facebook App active user data.

Simply put, Windows Phone 7 failed to thrive. It didn't take off. All that estimated $1B in marketing money added up to a big bucket-o-fail. It peaked at less than two percent of share on launch and is trailing off now to less than one fourth of that. Wishing that will change is not going to make it change. For the past month it's not even making up for the people giving up on Windows 6.5.

I, for one, would prefer they didn't gain any market share whatsoever. I would prefer that Microsoft fail in mobile, and that they continue to fail spectacularly by burning huge bonfires of money to no avail. Mobile is the future, and if you look at their suit against Barnes and Noble you will see that their desire for market share is not about innovation, it's not even about money. It's about control. They want to prevent all progress they don't supply. It's not enough for them to win - everybody else must lose also, including the customers. They want to stop all this neat stuff we've been getting the last few years. We like this stuff.

No, we don't need Microsoft for a vibrant competitive environment. Quite the opposite. For a vibrant competitive environment we need them to shrivel away to nothing through wasting all their money on lost causes. From the look of things they're well on their way.

Android tablets will put up the good fight yet. The Nook may save Barnes and Noble, particularly if they get really angry. We just need some tablets to hit the right price points with credible features and decent tablet-base OS. After that choice will win out over The iPad.

Your comment, it's just not true. (not my chart, btw). The integrated Facebook app is a good indicator of a mobile platform's market performance. Facebook users are common enough that they make a significant and representative statistical sample.

WP7 peaked below 1.5% market share on release, and is declining. It's now seeing about 4,300 new adopters each day worldwide, which is pathetic even for Windows Mobile. There is no way this can be described as "doing just fine." Its user base will never hit the 1.5 million units Microsoft claims are already delivered on its current trend, so somebody's about to get stuck with some dead inventory.

Its replacement Windows 8 has already been shown running at CES and the roadmap has a 1/7/2013 in-store availability scheduled. W8 being a full Windows rather than a mobile OS will of course not be compatible. Intel has committed that they will field phone platforms with it that run regular Windows applications on x86 phones. They're "all in".

So there's no reason to buy a WP7 phone. It failed to thrive, its execution date is set. There's no reason to develop apps for a phone with few users and no long-term prospects either.

Funny story: the KIN had about 8,000 sales and 300,000 Facebook likes. The integrated WP7 phone Facebook app has a little over 300,000 users now and less than 4,000 Facebook likes. It looks like buying Facebook likes has gone out of vogue with Microsoft's marketing department. But apparently hiring astroturfers to post on slashdot has not.

Seems to me that the dude has an agenda with Mozilla. His tweets are full of stuff like this..

Why we still need a strong Mozilla: to be "the Switzerland of HTML5" http://j.mp/gj6W99

Um, you can buy one of these? http://j.mp/9o6pCs

The Onion needs to update this: http://j.mp/aLzoqi with "Internet Nerd Constantly Mentioning He Doesn't Use Facebook"

Here's a nice pdf archived by wikimedia that shows where the problem is: AOL/NCSA Online Safety Study (2004).

I can't believe I'd never seen this before today.

Another article:

Meanwhile, Google says, Viacom "regularly uses so-called 'stealth marketing' to get its content onto YouTube. The goal is to create the appearance of authentic grassroots interest in the content being promoted." Google cites a marketing executive at Viacom's Paramount studio who said that clips posted to YouTube "should definitely not be associated with the studio -- should appear as if a fan created and posted it." To accomplish that, Google says that "Viacom employees have made special trips away from the company's premises (to places like Kinko's) to upload videos to YouTube from computers not traceable to Viacom."

Also, "Viacom has altered its own videos to make them appear stolen." Indeed, Google says that a former president of MTV, not named, testified that Viacom didn't take down clips from The Daily Show and The Colbert Report because "we were concerned that Jon Stewart and Stephen Colbert believed that their presence on YouTube was important for their ratings as well as for their relationship with their audience."

So... who's evil here?

And the eternal copyright of such essential works as "100 Polka Classics: The Greatest Accordion Collection On Earth" have turned me off of commercial music for good. There is no way that anybody involved in that making that work got paid if I pay to download it.

BTW: Track 41 isn't bad: "Yes, we have no bananas."

i wrote a post about this a couple days ago http://j.mp/aEJZnO i point to http://ie6nomore.com/ and agree with the first comment. im not going to bend over backwards to try and get ie6 when im also dealing with quirks of ie8,firefox and chrome... write it as best to standards as i can and unless the client wants to pay more let those few little rendering errors slide

Here's a post from GigaOM that basically says that Google stabbed Motorola's Droid in the back to deliver this, despite the fact that Droid has a keyboard and Nexus One does not and so it's fair game.

That was picked up by TheRegister which for some bizarre reason sees this as a reason for Moto to run home to the warm embrace of Microsoft, as if the whole Sendo Maneuver had not happened (even though they reported it), and as if Google had actually done something dastardly. Now it's in the Mainstream press [Businessweek.com] and by Tuesday they'll be trying to meme it.

I'm doing what I can but we need more fans of open systems to get out there and make fun of these idiots or they will continue to spew their nonsense and Joe Sixpack might believe them and then we won't get our cool new stuff. Slashdot is popular but there are a lot of other sites out there like cnet, zdnet, Google Groups and Yahoo where this nonsense might take root. People who know stuff need to go from here to there and put the word out that we'd like some shiny new Tegra 2 or Android slates, a Nexus One, and a Droid under our tree at Christmas and anybody that opposes that is a Luddite.

And for those for whom bit.ly is too long, the makers of bit.ly have a new domain: j.mp

Nooooooooooo..... They're doing GREAT.

The agency responsible for this brilliantly witty perspective is Crispin, Porter & Bogusky. They've done some marvellous work here. I can't wait to see more of their efforts. Their spin on Macenstein's Mac Chick of the Month (NSFW) would be the bees knees.

I thought paying Seinfeld ten million bucks for three commercials was a brilliant investment too! Without his classic humor Vista might not have done even as well as it did. He's what got the man on the street to give it a go.

They should give them three to five more years before they give up on this crew. Their genius is subtle, but you need a good pool of data before you know how strategic marketing is working. And of course they'll be needing a bigger budget. This needs a wider audience to good market penetration. Maybe they could buy some leader space on DVD's, some long spots during the fall sports classics - the jocks will eat this stuff up. I'm thinking a few spots during the Thanksgiving day parade with real captured Windows 7 Launch Party footage will be just the thing.

4Q 2009. Intel will be bringing the vitality of the Windows Mobile Experience to Mobile Linux (Moblin) by porting Microsoft Silverlight to it. We may all anticipate the usual robust stability, inherent security, device and application compatibility and outstanding performance we've traditionally seen from mobile Windows products.

An Intel spokesperson might say:

We look forward to a durable and productive partnership with Microsoft on the Linux platform. Based on the long history of Microsoft product partnership successes and their long-term commitment to driving adoption of Linux from the palmtop to HPC, Microsoft has shown itself capable and well motivated to help us achieve our goals with Moblin.

Of course I agree with you. Any organization of a reasonable size has desktop PCs coming off the regular rotation. Right now that means systems like these DC7600s. They have 3.2GHz single core processor and a gig or two of RAM and gigabit networking. 4 year old Dells have the same Spec. I have a cluster of 8 of these I use for quick stuff down in the basement - they're set for wake on LAN and netboot. Add a cheap gigabit switch and some DRBL configs, and you've got a beowulf cluster. They're useful for checking stuff out, testing VDI solutions and whatnot.

If you're rotating them out they're costless and if your problem works with asymmetrical nodes and you have good power, you've not only solved the computation deficit you've deferred the recycling problem as well. If you have to buy them at $1000 for five, that's $200 per core and you're probably better off buying modern quad-core desktops. The Small Form Factor ones work best for ad-hoc clustering.

But this only works if your needs are smaller - like for a pilot - and if the problem you're solving is not very granular (it can be divided only into larger chunks). For fine grained problems best approachable with shared-memory solutions, or scaling to thousands of cores, you get more bang for your buck with the GPGPU solutions because they have 600 cores on one $1,500 card.

That said, I do agree with you - the problems solvable by a rack of surplus desktops are numerous and interesting.

You've got low latency and high bandwidth. Make your storage iSCSI OpenFiler configured in cluster mode with block replication. Do use a pair of the BackBlaze boxes somebody else mentioned. Configure with RAID 6. Get enterprise support here. You're in and done at $16K capital cost, $2k labor, and annual support (24/7 4 hour response) at $6200/yr for 67TB of raw storage (~48TB net) plus whatever the network, rackspace and power costs, and it scales in volume storage at linear cost when your needs do and the more volume you have, the better performance gets. As a bonus it fits in two 4U slots.

If you want to skimp you don't have to fully populate the boxes until you need the room and can save $8K in capital costs up front. Every couple of months you have to hot-swap out some cheapo consumer grade drives so buy a few spares and configure them as hot spares and a few more for cold spares. If you have some extra Franklins, splurge on the 10G Ethernet connection from the BackBlaze box to the local network - the remote can stay on Gig-E because it's only used for writes or HA. With a little mental gymnastics and PSU field modifications you can use one BackBlaze master to control up to three BackBlaze slaves with passthrough connections only - no internal server needed. Just get the cards with some external eSATA or external SAS ports, depending on your preference. You might need to upgrade the motherboard spec on the master BackBlaze box, but it's worth the extra money. Since Openfiler support is unlimited CPU you may as well get the dual quad core Nehalem motherboard with 72GB RAM and 8 PCIe slots, or whatever's in the sweet spot this week. I do like the X5550, but if you can get a quad core for under $100 it's hard to pass up, especially combined with one of these cheap motherboards that use up to 32GB of cheap DDR2 RAM. Be careful with your PCIe slot counts when choosing motherboards.

Configure whatever machine you're using to do a backup periodically from one i-SCSI LUN on the local machine to another LUN. This gives you protection against 90% of backup needs (oops! I accidentally all all my presentations!) and will be transparently replicated to the HA site at block level without user intervention. Somewhere in here you should educate users that backup systems are not an alternative method of version control.

You could probably upgrade this with a few TB of PCIe attached SSD cache (pdf) for the million plus IOPS, guaranteed multiple 10Gbps network port saturation for an additional $40k, if you knew how, or why, or needed to.

Or you can go cheap with Linux and BSD and some scripts. You won't save any money and you won't have support. Buy the support. It's worth the money. Disclosure: I don't work for any of these folks. For the company I work for I can quote you a FC SAN. Trust me, you don't want to know what that costs for 67TB with block replication to a DR site and 24/7 4 hour support, let alone the scalable solution I've proposed here. Just assume it's "a lot".

You've got low latency and high bandwidth. Make your storage iSCSI OpenFiler configured in cluster mode with block replication. Do use a pair of the BackBlaze boxes somebody else mentioned. Configure with RAID 6. Get enterprise support here. You're in and done at $16K capital cost, $2k labor, and annual support (24/7 4 hour response) at $6200/yr for 67TB of raw storage (~48TB net) plus whatever the network, rackspace and power costs, and it scales in volume storage at linear cost when your needs do and the more volume you have, the better performance gets. As a bonus it fits in two 4U slots.

If you want to skimp you don't have to fully populate the boxes until you need the room and can save $8K in capital costs up front. Every couple of months you have to hot-swap out some cheapo consumer grade drives so buy a few spares and configure them as hot spares and a few more for cold spares. If you have some extra Franklins, splurge on the 10G Ethernet connection from the BackBlaze box to the local network - the remote can stay on Gig-E because it's only used for writes or HA. With a little mental gymnastics and PSU field modifications you can use one BackBlaze master to control up to three BackBlaze slaves with passthrough connections only - no internal server needed. Just get the cards with some external eSATA or external SAS ports, depending on your preference. You might need to upgrade the motherboard spec on the master BackBlaze box, but it's worth the extra money. Since Openfiler support is unlimited CPU you may as well get the dual quad core Nehalem motherboard with 72GB RAM and 8 PCIe slots, or whatever's in the sweet spot this week. I do like the X5550, but if you can get a quad core for under $100 it's hard to pass up, especially combined with one of these cheap motherboards that use up to 32GB of cheap DDR2 RAM. Be careful with your PCIe slot counts when choosing motherboards.

Configure whatever machine you're using to do a backup periodically from one i-SCSI LUN on the local machine to another LUN. This gives you protection against 90% of backup needs (oops! I accidentally all all my presentations!) and will be transparently replicated to the HA site at block level without user intervention. Somewhere in here you should educate users that backup systems are not an alternative method of version control.

You could probably upgrade this with a few TB of PCIe attached SSD cache (pdf) for the million plus IOPS, guaranteed multiple 10Gbps network port saturation for an additional $40k, if you knew how, or why, or needed to.

Or you can go cheap with Linux and BSD and some scripts. You won't save any money and you won't have support. Buy the support. It's worth the money. Disclosure: I don't work for any of these folks. For the company I work for I can quote you a FC SAN. Trust me, you don't want to know what that costs for 67TB with block replication to a DR site and 24/7 4 hour support, let alone the scalable solution I've proposed here. Just assume it's "a lot".

You've got low latency and high bandwidth. Make your storage iSCSI OpenFiler configured in cluster mode with block replication. Do use a pair of the BackBlaze boxes somebody else mentioned. Configure with RAID 6. Get enterprise support here. You're in and done at $16K capital cost, $2k labor, and annual support (24/7 4 hour response) at $6200/yr for 67TB of raw storage (~48TB net) plus whatever the network, rackspace and power costs, and it scales in volume storage at linear cost when your needs do and the more volume you have, the better performance gets. As a bonus it fits in two 4U slots.

If you want to skimp you don't have to fully populate the boxes until you need the room and can save $8K in capital costs up front. Every couple of months you have to hot-swap out some cheapo consumer grade drives so buy a few spares and configure them as hot spares and a few more for cold spares. If you have some extra Franklins, splurge on the 10G Ethernet connection from the BackBlaze box to the local network - the remote can stay on Gig-E because it's only used for writes or HA. With a little mental gymnastics and PSU field modifications you can use one BackBlaze master to control up to three BackBlaze slaves with passthrough connections only - no internal server needed. Just get the cards with some external eSATA or external SAS ports, depending on your preference. You might need to upgrade the motherboard spec on the master BackBlaze box, but it's worth the extra money. Since Openfiler support is unlimited CPU you may as well get the dual quad core Nehalem motherboard with 72GB RAM and 8 PCIe slots, or whatever's in the sweet spot this week. I do like the X5550, but if you can get a quad core for under $100 it's hard to pass up, especially combined with one of these cheap motherboards that use up to 32GB of cheap DDR2 RAM. Be careful with your PCIe slot counts when choosing motherboards.

Configure whatever machine you're using to do a backup periodically from one i-SCSI LUN on the local machine to another LUN. This gives you protection against 90% of backup needs (oops! I accidentally all all my presentations!) and will be transparently replicated to the HA site at block level without user intervention. Somewhere in here you should educate users that backup systems are not an alternative method of version control.

You could probably upgrade this with a few TB of PCIe attached SSD cache (pdf) for the million plus IOPS, guaranteed multiple 10Gbps network port saturation for an additional $40k, if you knew how, or why, or needed to.

Or you can go cheap with Linux and BSD and some scripts. You won't save any money and you won't have support. Buy the support. It's worth the money. Disclosure: I don't work for any of these folks. For the company I work for I can quote you a FC SAN. Trust me, you don't want to know what that costs for 67TB with block replication to a DR site and 24/7 4 hour support, let alone the scalable solution I've proposed here. Just assume it's "a lot".

You've got low latency and high bandwidth. Make your storage iSCSI OpenFiler configured in cluster mode with block replication. Do use a pair of the BackBlaze boxes somebody else mentioned. Configure with RAID 6. Get enterprise support here. You're in and done at $16K capital cost, $2k labor, and annual support (24/7 4 hour response) at $6200/yr for 67TB of raw storage (~48TB net) plus whatever the network, rackspace and power costs, and it scales in volume storage at linear cost when your needs do and the more volume you have, the better performance gets. As a bonus it fits in two 4U slots.

If you want to skimp you don't have to fully populate the boxes until you need the room and can save $8K in capital costs up front. Every couple of months you have to hot-swap out some cheapo consumer grade drives so buy a few spares and configure them as hot spares and a few more for cold spares. If you have some extra Franklins, splurge on the 10G Ethernet connection from the BackBlaze box to the local network - the remote can stay on Gig-E because it's only used for writes or HA. With a little mental gymnastics and PSU field modifications you can use one BackBlaze master to control up to three BackBlaze slaves with passthrough connections only - no internal server needed. Just get the cards with some external eSATA or external SAS ports, depending on your preference. You might need to upgrade the motherboard spec on the master BackBlaze box, but it's worth the extra money. Since Openfiler support is unlimited CPU you may as well get the dual quad core Nehalem motherboard with 72GB RAM and 8 PCIe slots, or whatever's in the sweet spot this week. I do like the X5550, but if you can get a quad core for under $100 it's hard to pass up, especially combined with one of these cheap motherboards that use up to 32GB of cheap DDR2 RAM. Be careful with your PCIe slot counts when choosing motherboards.

Configure whatever machine you're using to do a backup periodically from one i-SCSI LUN on the local machine to another LUN. This gives you protection against 90% of backup needs (oops! I accidentally all all my presentations!) and will be transparently replicated to the HA site at block level without user intervention. Somewhere in here you should educate users that backup systems are not an alternative method of version control.

You could probably upgrade this with a few TB of PCIe attached SSD cache (pdf) for the million plus IOPS, guaranteed multiple 10Gbps network port saturation for an additional $40k, if you knew how, or why, or needed to.

Or you can go cheap with Linux and BSD and some scripts. You won't save any money and you won't have support. Buy the support. It's worth the money. Disclosure: I don't work for any of these folks. For the company I work for I can quote you a FC SAN. Trust me, you don't want to know what that costs for 67TB with block replication to a DR site and 24/7 4 hour support, let alone the scalable solution I've proposed here. Just assume it's "a lot".

You've got low latency and high bandwidth. Make your storage iSCSI OpenFiler configured in cluster mode with block replication. Do use a pair of the BackBlaze boxes somebody else mentioned. Configure with RAID 6. Get enterprise support here. You're in and done at $16K capital cost, $2k labor, and annual support (24/7 4 hour response) at $6200/yr for 67TB of raw storage (~48TB net) plus whatever the network, rackspace and power costs, and it scales in volume storage at linear cost when your needs do and the more volume you have, the better performance gets. As a bonus it fits in two 4U slots.

If you want to skimp you don't have to fully populate the boxes until you need the room and can save $8K in capital costs up front. Every couple of months you have to hot-swap out some cheapo consumer grade drives so buy a few spares and configure them as hot spares and a few more for cold spares. If you have some extra Franklins, splurge on the 10G Ethernet connection from the BackBlaze box to the local network - the remote can stay on Gig-E because it's only used for writes or HA. With a little mental gymnastics and PSU field modifications you can use one BackBlaze master to control up to three BackBlaze slaves with passthrough connections only - no internal server needed. Just get the cards with some external eSATA or external SAS ports, depending on your preference. You might need to upgrade the motherboard spec on the master BackBlaze box, but it's worth the extra money. Since Openfiler support is unlimited CPU you may as well get the dual quad core Nehalem motherboard with 72GB RAM and 8 PCIe slots, or whatever's in the sweet spot this week. I do like the X5550, but if you can get a quad core for under $100 it's hard to pass up, especially combined with one of these cheap motherboards that use up to 32GB of cheap DDR2 RAM. Be careful with your PCIe slot counts when choosing motherboards.

Configure whatever machine you're using to do a backup periodically from one i-SCSI LUN on the local machine to another LUN. This gives you protection against 90% of backup needs (oops! I accidentally all all my presentations!) and will be transparently replicated to the HA site at block level without user intervention. Somewhere in here you should educate users that backup systems are not an alternative method of version control.

You could probably upgrade this with a few TB of PCIe attached SSD cache (pdf) for the million plus IOPS, guaranteed multiple 10Gbps network port saturation for an additional $40k, if you knew how, or why, or needed to.

Or you can go cheap with Linux and BSD and some scripts. You won't save any money and you won't have support. Buy the support. It's worth the money. Disclosure: I don't work for any of these folks. For the company I work for I can quote you a FC SAN. Trust me, you don't want to know what that costs for 67TB with block replication to a DR site and 24/7 4 hour support, let alone the scalable solution I've proposed here. Just assume it's "a lot".

You've got low latency and high bandwidth. Make your storage iSCSI OpenFiler configured in cluster mode with block replication. Do use a pair of the BackBlaze boxes somebody else mentioned. Configure with RAID 6. Get enterprise support here. You're in and done at $16K capital cost, $2k labor, and annual support (24/7 4 hour response) at $6200/yr for 67TB of raw storage (~48TB net) plus whatever the network, rackspace and power costs, and it scales in volume storage at linear cost when your needs do and the more volume you have, the better performance gets. As a bonus it fits in two 4U slots.

If you want to skimp you don't have to fully populate the boxes until you need the room and can save $8K in capital costs up front. Every couple of months you have to hot-swap out some cheapo consumer grade drives so buy a few spares and configure them as hot spares and a few more for cold spares. If you have some extra Franklins, splurge on the 10G Ethernet connection from the BackBlaze box to the local network - the remote can stay on Gig-E because it's only used for writes or HA. With a little mental gymnastics and PSU field modifications you can use one BackBlaze master to control up to three BackBlaze slaves with passthrough connections only - no internal server needed. Just get the cards with some external eSATA or external SAS ports, depending on your preference. You might need to upgrade the motherboard spec on the master BackBlaze box, but it's worth the extra money. Since Openfiler support is unlimited CPU you may as well get the dual quad core Nehalem motherboard with 72GB RAM and 8 PCIe slots, or whatever's in the sweet spot this week. I do like the X5550, but if you can get a quad core for under $100 it's hard to pass up, especially combined with one of these cheap motherboards that use up to 32GB of cheap DDR2 RAM. Be careful with your PCIe slot counts when choosing motherboards.

Configure whatever machine you're using to do a backup periodically from one i-SCSI LUN on the local machine to another LUN. This gives you protection against 90% of backup needs (oops! I accidentally all all my presentations!) and will be transparently replicated to the HA site at block level without user intervention. Somewhere in here you should educate users that backup systems are not an alternative method of version control.

You could probably upgrade this with a few TB of PCIe attached SSD cache (pdf) for the million plus IOPS, guaranteed multiple 10Gbps network port saturation for an additional $40k, if you knew how, or why, or needed to.

Or you can go cheap with Linux and BSD and some scripts. You won't save any money and you won't have support. Buy the support. It's worth the money. Disclosure: I don't work for any of these folks. For the company I work for I can quote you a FC SAN. Trust me, you don't want to know what that costs for 67TB with block replication to a DR site and 24/7 4 hour support, let alone the scalable solution I've proposed here. Just assume it's "a lot".