Friday, May 01
Daily News Stuff 1 May 2026
Yay Edition
Yay Edition
Top Story
- CPanel, a web hosting software suite used by 70 million sites worldwide, had a security hole big enough to drive a Kenworth through without needing to check for clearance. (Bleeping Computer)
And people were reportedly trying to exploit in February. And succeeding in March.
It was patched last night.
Lots of sysadmins having a bad day right now.
Tech News
- If you're playing the new pirate game Windrose, you might want to stop before your SSD explodes. (Tom's Hardware)
Players have reported that the game writes 30MB of date per second to disk.
Constantly.
Over 100GB per hour.
And SSDs have a finite lifespan under constant write load.
- Alphabet, Amazon, Meta, and Microsoft are planning a combined capital expenditure of $725 billion this year. (Tom's Hardware)
Which used to be a lot, but that was long ago and in another country.
- How to get 10Gb Ethernet working at home. (Giles Thomas)
Step One: Decide 2.5Gb is fine.
- Stripe has a new digital wallet for sharing with AI agents and other idiot children. (Tech Crunch)
You are asked to confirm every purchase before anything goes through so they can't run amok and drain your bank account.
- Profits at Samsung's memory division are up 49% over the same period last year. (WCCFTech)
Oh, my mistake. They're up 49 times over the same period last year.
- Don't talk about the goblin war. (Ars Technica)
AI agents are like profoundly autistic children, except that you can switch them off.
Musical Interlude
Disclaimer: Well, there's go my weekend plans.
Posted by: Pixy Misa at
06:27 PM
| Comments (2)
| Add Comment
| Trackbacks (Suck)
Post contains 264 words, total size 3 kb.
1
"And SSDs have a finite lifespan under constant write load."
Years ago, someone, maybe at Anandtech, but I don't remember who for sure, did some long-term tests and discovered that a lot of SSDs, while they were rated for something like 100TBW per TB of capacity, actually didn't start failing until more like a couple of petabytes written per TB of capacity. I don't think anyone's publicly done anything like that recently, though, and the Tom's article did mention that TLC drives are less susceptible than QLC drives. But it's still a bad design.
Also, the article "explained" that it's something to do with some database, but devolved into gibberish Treknobabble for a paragraph or so, which was really lame.
Years ago, someone, maybe at Anandtech, but I don't remember who for sure, did some long-term tests and discovered that a lot of SSDs, while they were rated for something like 100TBW per TB of capacity, actually didn't start failing until more like a couple of petabytes written per TB of capacity. I don't think anyone's publicly done anything like that recently, though, and the Tom's article did mention that TLC drives are less susceptible than QLC drives. But it's still a bad design.
Also, the article "explained" that it's something to do with some database, but devolved into gibberish Treknobabble for a paragraph or so, which was really lame.
Posted by: Rick C at Friday, May 01 2026 11:38 PM (1zWbY)
2
RocksDB uses what is called a log-structured merge tree, which allows it to perform very fast updates by writing random updates as a sequential log before merging the table data together. That was a big deal before SSDs, it's not so significant now.
But the real reason Windrose uses so much disk bandwidth is that the developers are idiots... In this particular field. There's no reason for it to write that much.
But the real reason Windrose uses so much disk bandwidth is that the developers are idiots... In this particular field. There's no reason for it to write that much.
Posted by: Pixy Misa at Saturday, May 02 2026 05:14 AM (PiXy!)
53kb generated in CPU 0.0119, elapsed 0.1152 seconds.
58 queries taking 0.1062 seconds, 366 records returned.
Powered by Minx 1.1.6c-pink.
58 queries taking 0.1062 seconds, 366 records returned.
Powered by Minx 1.1.6c-pink.
