SoylentNews is people
SoylentNews is powered by your submissions, so send in your scoop. Only 15 submissions in the queue.
SoylentNews
Submitted via IRC for TheMightyBuzzard
Twitter Inc. is planning widespread job cuts, to be announced as soon as this week, according to people familiar with the matter.
The company may cut about 8 percent of the workforce, or about 300 people, the same percentage it did last year when co-founder Jack Dorsey took over as chief executive officer, the people said. Planning for the cuts is still fluid and the number could change, they added. The people asked not to be identified talking about private company plans.
Yup, they're really pining for the fjords now.
This discussion has been archived.
No new comments can be posted.
Twitter Planning Hundreds More Job Cuts as Soon as This Week
|
Log In/Create an Account
| Top
| 26 comments
| Search Discussion
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
* m2 stares at the monitor... it looks like a hamburger...
m2 - that's a bad sign
(Score: 2) by VLM on Wednesday October 26 2016, @09:33PM
But does the scale provide value, either to the users or the bottom line, and why wouldn't sufficiently advanced IRC scale well? Isn't the market kinda saying "no"?
I'm not just saying toss twitter and load up IRC clients and connect to freenode. You could do some behind the scenes stuff.
Remember twitter PR claims are only a factor of a thousand larger than freenode actual results. Real (as opposed to PR) is probably only a hundred times bigger. Obviously freenode is not dying of over use either so freenode could tolerate a bit more traffic. I think that 100x scaling factor quite reasonable.
Its interesting to look at how low the requirements are for freenode and then extrapolate a hundred times larger.
https://freenode.net/support [freenode.net]
So 200 gigs ram instead of 2 gigs ram. Or 300 megs/sec of traffic as opposed to 3 megs/sec
I guess if you limited IRC like twitter does limiting, then traffic would be lower?
If I made 100 virtual freenode servers (or more likely given current day specs, 10 big virtual servers) and set it all up privately and had a nice looking web/app frontend that collates the 10 or so sharded freenode servers... And of course web front ends can adsorb an infinite quantity of engineer hours depending on how nuts the designers get.
I'm just saying we're talking a couple orders of magnitude larger than a hobbiest scale project. It would be like if a kids model rocket went 1/100th the way to the moon or launched a mere 50 pounds into geosynch orbit. Its a big step, just not that far of a step.
Another interesting way to look at 100x scaling factors is a couple years ago when I only had a couple years experience with these computer thingies my dad spent about a car payment on some 4116 16Kilobit dynamic ram chips. Figure I donno $50 each, and he needed eight for 16kilobytes. And this was a good deal as they were kind of exotic technology at the time. So a couple decades later 16 gigabytes of dram is about eighty bucks. So in lets say 20 years thats six orders of magnitude of storage for a quarter the price. So at least with respect to ram storage in about 2020 a raspberry pi is going to ship with enough ram for $30 to run a clone of 2016 twitter. Yeah yeah I know there's CPU and network traffic blah blah whatever. Still the point is you can measure the cost of twitter clone not just in dolllars today but as pennies in upcoming years.
Another way to look at it is non IRC performance. Doesn't craigslist send 20 million emails a day thru FOSS mailer called haraka or something? I mean high performance networking and traffic isn't a big problem anymore.
(Score: 2) by lgw on Friday October 28 2016, @12:54AM
Scale is hard. Every time you add a 0 over what you've done before, you learn a whole new world of lessons.
Twitter has 100M users to keep track of, about 700 tweets per second, and about 35000 calls per second to its API. 700 tweets per second (will all the fan out to subscribers) takes some heavy lifting, and a good ops team to keep live. 35k tps for the APIs is a mess, even if they're all passthrough to a backend DB. That's a volume where just load balancing the traffic is a specialty.
That being said, you could build it all on AWS and make ops their problem (DynamoDB won't give you any grief over 100k TPS, and the other plumbing you'd need would scale as well), but it would be pricey. Still, cheaper that way than a couple thousand engineers.