3 ms·

> What it does show is how much dead weight there was at Twitter and how little was lost by letting it go. If he wasn't told something vitally important, then

> What it does show is how much dead weight there was at Twitter and how little was lost by letting it go. If he wasn't told something vitally important, then the odds are the engineers he let go were ones that knew about that vital knowledge. And if you think the scale and infrastructure cost is absurd for Twitter I'd like to see the services you've worked on. Right now.

busterarm 3y ago
156,000 individual physical servers (remember, it said 5200 racks * 30) in one building and not being the only building _to serve 140 character tweets_ IS fucking absurd. This is the same company that took a nuclear-fucking-orbit dump on Rails because they didn't understand that your standard relational databases cannot handle their architecture pattern of very high write rate and very high read rate. This is the same company where the same lead engineer's NIH-syndrome led to them writing their own bespoke message queue twice. And they threw it out twice. And I do run 10^5 scale infrastructure -- globally distributed and not with any single building at some eye-watering $100M/yr run rate. It may be a highly visited site, but if you tell me they actually need that much physical stuff to _host tweets_ then I'm calling bullshit. Their quarterly operating costs _not including staff_ were over half a billion dollars. Twice what they were paying in salaries. That's a completely inverted cost relationship to how companies should operate.
- fzeroracer 3y ago
  First of all, their infrastructure is globally distributed. This was one data center of many. Second of all, this isn't just '140 character tweets'. It's images, video, data being sent at staggering speeds because people tweet and they tweet frequently. And people want instant response, they want to see tweets instantly and respond to them instantly. I highly doubt you do actually run infrastructure given that you're both misrepresenting what's actually going on here and taking the word of a person who in the article alone shows the various different ways of how he distorts reality. This isn't to say Twitter's infrastructure was great, because all services have heavy amounts of tech debt and cruft associated with years of fixes and updates. But rather that data being sent at high speeds with high reliability and at high frequency leads to high costs.
  - busterarm 3y ago
    I'm not stupid. I never said they weren't globally-distributed. I said to have that much stuff in any one building was moronic. Pre-Musk, Twitter's quarterly infrastructure bill dwarfed both Cloudflare and Akamai's combined. All without even having paying custoemrs. That's utterly ridiculous. I'm not trusting Elon Musk to believe this either, it's all available to look up on EDGAR.

busterarm 3y ago

156,000 individual physical servers (remember, it said 5200 racks * 30) in one building and not being the only building _to serve 140 character tweets_ IS fucking absurd. This is the same company that took a nuclear-fucking-orbit dump on Rails because they didn't understand that your standard relational databases cannot handle their architecture pattern of very high write rate and very high read rate. This is the same company where the same lead engineer's NIH-syndrome led to them writing their own bespoke message queue twice. And they threw it out twice. And I do run 10^5 scale infrastructure -- globally distributed and not with any single building at some eye-watering $100M/yr run rate. It may be a highly visited site, but if you tell me they actually need that much physical stuff to _host tweets_ then I'm calling bullshit. Their quarterly operating costs _not including staff_ were over half a billion dollars. Twice what they were paying in salaries. That's a completely inverted cost relationship to how companies should operate.

fzeroracer 3y ago
First of all, their infrastructure is globally distributed. This was one data center of many. Second of all, this isn't just '140 character tweets'. It's images, video, data being sent at staggering speeds because people tweet and they tweet frequently. And people want instant response, they want to see tweets instantly and respond to them instantly. I highly doubt you do actually run infrastructure given that you're both misrepresenting what's actually going on here and taking the word of a person who in the article alone shows the various different ways of how he distorts reality. This isn't to say Twitter's infrastructure was great, because all services have heavy amounts of tech debt and cruft associated with years of fixes and updates. But rather that data being sent at high speeds with high reliability and at high frequency leads to high costs.
- busterarm 3y ago
  I'm not stupid. I never said they weren't globally-distributed. I said to have that much stuff in any one building was moronic. Pre-Musk, Twitter's quarterly infrastructure bill dwarfed both Cloudflare and Akamai's combined. All without even having paying custoemrs. That's utterly ridiculous. I'm not trusting Elon Musk to believe this either, it's all available to look up on EDGAR.