Cloudflare frees up 100TB of RAM by shrinking 1.1.1.1’s DNS cache entries — 250 billion cached DNS entries at any given time means one wasted byte costs 250GB

Cloudflare frees up 100TB of RAM by shrinking 1.1.1.1's DNS cache entries — 250 billion cached DNS entries at any given time means one wasted byte costs 250GB

None of the liberated RAM will be cashed out as smaller memory configurations, though; Cloudflare plans to pour it back into larger DNS caches, which raise hit rates and cut the query traffic it sends upstream to authoritative servers.

Follow Tom's Hardware on Google News , or add us as a preferred source , to get our latest news, analysis, & reviews in your feeds.

Luke James is a freelance writer and journalist.\u00a0 Although his background is in legal, he has a personal interest in all things tech, especially hardware and microelectronics, and anything regulatory.\u00a0 ","collapsible":{"enabled":true,"maxHeight":250,"readMoreText":"Read more","readLessText":"Read less"}}), "https://slice.vanilla.futurecdn.net/13-4-25/js/authorBio.js"); } else { console.error('%c FTE ','background: #9306F9; color: #ffffff','no lazy slice hydration function available'); } Luke James Social Links Navigation Contributor Luke James is a freelance writer and journalist. Although his background is in legal, he has a personal interest in all things tech, especially hardware and microelectronics, and anything regulatory.

DS426 Crazy how many resources can be freed up by a small handful of software engineers when you're talking about hardware on this scale. Doing some rough napkin math, 1,560 DDR-5600 RDIMM's were spared at let's say an average street price of $2,500 per RDIMM resulting in a potential $3.9 million savings. Just for DNS, just in memory costs. Cloudflare is surprisingly transparent about their custom servers: https://blog.cloudflare.com/gen13-config/ Reply

KaiserTom The RAMpocalyse has been a decade and a half in the making from a severe lack of any optimization for memory usage. We've had so much surplus memory it was cheaper to code less and cache more to pretend like systems and apps were more performant than they actually were. Just relying on caching everything they need to do the work quickly, rather than coding it to work smart and efficiently. This is just one very blatant example of that stupid industry dogma coming to a head when memory finally became too expensive to ignore. And now people need to actually optimize their software to run with less of it so systems can be significantly cheaper and more efficient. Because they can't afford to double their infrastructure again on poorly coded and un-optimized production software. Reply

PEnns KaiserTom said: The RAMpocalyse has been a decade and a half in the making from a severe lack of any optimization for memory usage. We've had so much surplus memory it was cheaper to code less and cache more to pretend like systems and apps were more performant than they actually were. Just relying on caching everything they need to do the work quickly, rather than coding it to work smart and efficiently. This is just one very blatant example of that stupid industry dogma coming to a head when memory finally became too expensive to ignore. And now people need to actually optimize their software to run with less of it so systems can be significantly cheaper and more efficient. Because they can't afford to double their infrastructure again on poorly coded and un-optimized production software. Nailed it!! 👍👍 Reply

moon2 KaiserTom said: The RAMpocalyse has been a decade and a half in the making from a severe lack of any optimization for memory usage. We've had so much surplus memory it was cheaper to code less and cache more to pretend like systems and apps were more performant than they actually were. Just relying on caching everything they need to do the work quickly, rather than coding it to work smart and efficiently. This is just one very blatant example of that stupid industry dogma coming to a head when memory finally became too expensive to ignore. And now people need to actually optimize their software to run with less of it so systems can be significantly cheaper and more efficient. Because they can't afford to double their infrastructure again on poorly coded and un-optimized production software. Wrong take. Read the description of what cloudflare did. One aspect of what they did was to replace a built in collection type with one that had no capacity field. This is a literal example of what would be considered premature optimisation in 99.99999999% of all cases. The kind of person who reads this article and wails about inefficient bloated software is one who lacks the domain knowledge to understand why this is a last-resort style optimisation and never something someone should do up front as a general technique. Storing data in wire format makes sense – easier to memcopy it then. The move to place slightly larger objects in the heap by storing a pointer to them in the union type is interesting though. It worsened performance while reducing memory usage. They were ok with that pessimisation as the system overall improved through the full set of optimisations. This is an interesting example of where some targeted micro optimisations can yield real world benefits for highly specific usecases. Reply

alexbusu moon2 said: Wrong take. Read the description of what cloudflare did. One aspect of what they did was to replace a built in collection type with one that had no capacity field. This is a literal example of what would be considered premature optimisation in 99.99999999% of all cases. The kind of person who reads this article and wails about inefficient bloated software is one who lacks the domain knowledge to understand why this is a last-resort style optimisation and never something someone should do up front as a general technique. Storing data in wire format makes sense – easier to memcopy it then. The move to place slightly larger objects in the heap by storing a pointer to them in the union type is interesting though. It worsened performance while reducing memory usage. They were ok with that pessimisation as the system overall improved through the full set of optimisations. This is an interesting example of where some targeted micro optimisations can yield real world benefits for highly specific usecases. I very much agree with what you say. Reply

Key considerations

  • Investor positioning can change fast
  • Volatility remains possible near catalysts
  • Macro rates and liquidity can dominate flows

Reference reading

More on this site

Informational only. No financial advice. Do your own research.

Leave a Comment