I meant to say, something like that, a SuperMicro chassis with SAS expander backplane, does of course cost a lot more if you don't buy on eBay. A new one from Provantage is around $700 or so. Still, compared to a $3,500 tape library ...

That's not say tapes don't have their place. Tape was good enough to back up my grandpa's data in 1954 and it's still good enough, sometimes. Other times, large capacity disks really do make more sense.

It's one of the SGI units. Used, they sell for half that, so I misspoke. The storage SERVERS, with motherboard and processor, are $350. So for $700 you can get the backup controller server with 16 bays plus two more 16 bay jbods to daisy chain to it. Not bad for backup. That's not what I'd use for my main enterprise storage SAN, but for backup yeah it works real well.

That is without the disks themselves, of course. Starting with four or six 3TB drives in RAID 10, you get 150-200 MB/s actual for several hundred dollars, then add spindles as needed.

> This is impossible to flatly guarantee ... no RTOS is going to assure that.

Does that include a "trivial" RTOS, or are you speaking only of an RTOS of significantly complexity? It would seem that on an MCU, the very simplest OS, the "HelloWorld.asm" of operating systems, would absolutely run each of its functions in the exact same number of cycles, every single time. On a Z80, for example, INC always takes exactly one cycle, and ADD take two cycles, every time.

On a Core processor it would be much less consistent due to pipelining, out-of-order execution, etc., but these little MCUs don't do any of that, do they?
Some of the little bit of MCU code I've written has been fairly sensitive to timing and I've figured that 4 cycles is 4 cycles, every time. Have I been doing it wrong?

Also, those pro sports salaries in the headlines create some misconceptions. You see a headline about a $30 million contract and think pro athletes are fabulously wealthy. In reality, that $30 million is the maximum that the best player is eligible to earn over three years, which is half of their career.

Taking the NFL as an example, players work very hard for several years trying to get into the pros. For the few who make it, the average salary is $1.9 million, but the average career is only six years. That's $5.4 million for their career. A lot of tech workers will make a lot more than that in their career. Especially so if they worked as hard during high school and college as the kids who become pro athletes do, waking up two hours early to work out (or study) and then staying after school for practice, etc.

It's a good job, don't get me wrong, but it's not as obscenely lucrative as a glance at the headlines might make it appear. Hollywood, on the other hand, is incredibly lucrative for the very top talent, if they stay on top for many years. If 20 million people are entertained enough by having you on a show that it's worth 5 cents per week for them, that's $1 million per week of entertainment value.

And the NHL ... is 95+% white dudes

Hockey is the national sport of Canada. Canada is 95+% white.

Well duh. When black kids want to play sports they go outside outside and play - basketball, football, whatever they just play. If they want to dunk, they dunk.

There's one Asian guy who can dunk. The rest have to dunk from the couch. Is it not obvious why little Asian high school kids aren't on the high school football team? There's a safety issue there. Look at the NFL and NFA - thousands of black guys, 34 white, and two Asian. The white guys who want to play do so via Madden 2013.

I used to do something similar. Then I made my rsynced copies bootable with qemu-kvm. I already had a datacenter, so I rsynced it there. That provides several advantages.

> If you aren't here and there is a fire, you can always
> store it around a two layers of bricks, with a fire blanket between them.

You COULD, but you probably don't, and overheating would be a concern, as would delaminating of the platters in a fire. Theft is concern as well.

With my better version, I started syncing systems for a few friends. When I needed more capacity, I bought 16 bay SAS JBOB units for $350. It's grown into quite a nice, professional system, with real protection from fire, theft, etc. but my friends still just pay $12 / month to cover the costs.

I find that the cost of auxiliary equipment, servers, is far LESS for spindles. I just bought a 16 bay SAS jbod for $350. That's up to 64TB raw. A tape library would have cost $3,500.

Sure you CAN have a human switch tapes, just as you can have a human hotswap drives from any old server you want to use for backup storage. At least at the level of about 80 TBs, spindles are a lot less expensive as well as more convenient.

If you already have humans sitting around the datacenter who have nothing better to than switch tapes, and if you have hundreds of TBs, I suspect tapes make sense in that case.

Indeed, blatant slashvertisement. The video DID mention some key points. For those who didn't feel like watching the video or reading the transcript, aside from pure advertising, they did hit four points which I refer to as the golden rules of backup:

Backups must be:
Off site: fires, thefts happen, and they happen in datacenters too.
Automated: people will stop manually copying and swapping, probably at the worst possible time.
Rotated: Not just one backup overwritten daily. If you were hacked at 11:00 PM, that midnight backup doesn't help.
Tested: Of our customers who thought they had backups, over half didn't actually have working backups when we suggested they test them.

In the spirit of blatant advertising, Clonebox provides a very similar service, at a slightly better price, and the owner is a long time /.er

I think you read that sentence backwards. I said four threads on the CPU. A quad core processor can of course do at least four threads.

Also, four threads on the CPU means there are probably at least a dozen processes waiting on IO. Four threads active on the CPU, sixteen processes active on the system.

Versus one active process for System V since it runs them sequentially.
   

> their profitability depends almost entirely on how long they can keep
> the assets they have before they have to replace them.

Exactly. Their cost is largely a matter of how long they can continue to use the equipment they purchase and install. If users switch from viewing Facebook to watching Netflix all night, ISPs need much more capacity, which means replacing XGbps plant with 10XGbps equipment before it has worn out.

As you correctly pointed out, the XGbps equipment would continue to work just fine, delivering up Facebook for years. Infrastructure sized for Facebook pictures isn't sufficient for streamlining high definition video. As you said, replacing that equipment with faster equipment is a major cost.

What you said is true for say, a building. ISPs's networks are not "you purchase it once" items.
You say "yes it does scale somewhat but only in the short run", and that's right, infrastructure costs are only particularly important for equipment that is kept in service for less than 10 years - such as networking equipment.

I was prioritizing traffic in 1997, when I owned a hosting company. You don't need a standardized RFC giving an industry-wide standard method for doing something before you start doing it. In fact, RFCs frequently codify how people have already been doing things. The 1999 RFC shows that by that time , it had become common enough that the method needed to be standardized.

Customers who want high priority get it, spammers get deprioritized. It's been done forever and it hasn't been a catastrophe.

> Second, 1TB/mo is only around 3Mbps. The average broadband internet connection these days is going to be many times that, regardless of whether they're using 1GB/mo or 1TB/mo.

If 10,000 customers average 3 Mbps each, the ISP needs 30,000 Mbps of infrastructure for them. The instantaneous peak "up to" speed in the advertisements has nothing to do with it. If they use 1 GB, that's 1 GB the infrastructure has to carry. If they use 1 TB, the infrastructure has to carry 1 TB. Infrastructure for TBs costs a lot more than infrastructure for GBs. I'm certain of this because I've purchased both.

Yes, most of it will be optical, not copper . The same price difference applies. Most Slashdot readers aren't familiar with $10,000+ switches so I i
used analogous equipment that most nerds are familiar with. A Cisco router for 100 Gbps average usage costs less than one for 1,000 Gbps.

He said "the second it's allowed". It's ALWAYS been allowed, modulo Sherman and similar existing laws.
He's claiming that the moment it's not illegal ... Well, it's not illegal today, it's never been illegal, and his predictions haven't come true.

Given that the scare mongering is clearly bullshit, then it's time to ask "who is trying to sell us net neutrality using these transparent scare tactics, and what might their actual reasons be?" Clearly the stated reason is BS because we ALREADY live in a world with no net neutrality laws and tragedy has not befallen us.