The entire NIST budget is pretty large. This is one of the most visible things they do. By threatening to cut this if their budget shrinks, perhaps they're hoping to preserve their original budget request.

That's a realistic story in light of how Sesame Street, one of PBS's most profitable (and absolutely self-sustaining) franchises, was used to campaign against budget cuts back in the day. If the NIST head was a completely unscrupulous government agent (not making any accusations of course) then it would clearly be in their interest to only offer to cut valuable high-profile programs, burying their cut-able pet projects as deeply as possible. That would be the strategy that minimized the political will to cut the budget.

[0]https://www.washingtonpost.com/politics/decision2012/big-bir...

Also, the drones may never have existed.

With that much RAM, you'd be insane not to use ECC, which is another good reason for Xeon.

Or AMD, which supports ECC across all (most?) CPU lines historically. Again, the argument is about limitations of Intel consumer lines.

Question is, is buying those historical AMD CPUs which support ECC a good idea?

So far, none of the modern ones are confirmed to support ECC.

Really? I won't use a computer without it for production work. You get one bit-flip/GB/year. With 128GB or more in most of our desktops, it's not worth the risk.

Here's how we do it:

All the databases and other data are backed up to s3. For mysql, we use the python mysql-to-s3 backup scripts.

But the machines themselves are "backed up" by virtue of being able to be rebuilt with saltstack. We verify through nightly builds that we can bring a fresh instance up, with the latest dataset restored from s3, from scratch.

This makes it simple for us to switch providers, and can run our "production" instances locally on virtual machines running the exact same version of CentOS or FreeBSD we use in production.

We routinely run 5MM simultaneous TCP/IP connections on a single 12-core box (with Erlang!).

There's no reason for Twitter to need all those employees and all that hardware. If someone can get it at a fire sale price, and reduce it to 100 employees, they can have a nice business.

Twitter's needs are several orders of magnitude greater than that, and those kind of problems have a habit of being exponential.

It's not even the same kind of problem, they've got to tailor their response to every user.

Finally, their investors simply cannot accept 'a nice little business'. They need Twitter to deliver serious ROI.

As a point of reference, even the Wikimedia Foundation had almost 300 staff and contractors as of 2015. This is admittedly a lot fewer than Twitter employs today. However, I suspect that even a minimalist Twitter without sales, etc. needs more employees and would have more of other types of costs than Wikimedia.

> As a point of reference, even the Wikimedia Foundation had almost 300 staff and contractors as of 2015. This is admittedly a lot fewer than Twitter employs today.

Exactly, that's actually a pretty good argument for why Twitter should downsize substantially.

In no way does Twitter need more than 1,000 employees.

And frankly, much of the same criticism applies to them. A tiny tiny sliver of their constantly multiplying budget is spent on actually hosting the Wikipedia.

That's somewhat fair. It's certainly true that Wikimedia has a fair number of active projects that haven't had much of an impact. Apparently there have been at least some discussions of streamlining their work although organizations universally find it hard to avoid scope creep.

I'd point out though that, according to Wikipedia :-), the Internet Archive has a staff of about 200 so a few hundred employees/contractors doesn't seem out out of line as the baseline for a non-profit information infrastructure project.

Do your 5MM connections do anything that garners the attention of CNN and Google realtime search indexing?

Do you have backend data services and data analysis jobs behing those 12-core front-end servers?

Whatsapp?

edit: which had a ratio of 1 engineer to 16 million active users when acquired http://highscalability.com/blog/2014/2/26/the-whatsapp-archi...

If Twitter only needed to serve the states, sure. But they need to sync servers globally and provide decent bandwidth to every corner of the developed world. That's a big $ technical problem.

How big would your salesforce (and supporting components) be?

What's their business model? Why would I type my thoughts there instead of, say, "google docs." Or putting text files in a dropbox.

I like the idea of quiet and uncluttered. But it's hard to trust my writing to some new web app with to viable way of monetizing it.

I guess for now their main focus is on product, once they've acquired tons of users, they may go various ways to monetize.

LAX is the second largest (by passenger boarding) airport in the country. So until ATL allows Uber, I guess this is true.

Wikipedia says the data rate from Pluto is effectively about 1 KBit/second. That's why it takes so long to get the photos....

I'd like to find more detailed information about the hardware and software on this thing. It's amazing that it was launched back in 2006. My 2006 Mac Pro is basically useless now....

I like the concept; you wonder why Dropbox didn't enter this space.