Abhi

Data Storage @CERN, formerly @ceph🦑. Opinions my own
He/Him

[Abhi @abhishekl]

@rushiagr @UC_Assist Yes exactly my issue also, the amount of time spent to reach a customer care is clearly the biggest loss

[Abhi @abhishekl]

@UC_Assist @rushiagr Having the same experience with them now, way to go with losing customers! Their chat bot is just going in circles

[Abhi @abhishekl]

@UC_Assist Well the booking was supposed to be at 5pm so it should be a joke that you rescheduled from your side and then you are asking me to pay a surge charge for your inconvenience? Scam much?

[Abhi @abhishekl]

Hey @urbancompany_UC you changed a booking without notifying me and now you want to charge a surge charge? What joke is this, and your ai help is as dumb as it comes

[Abhi @abhishekl]

@UC_Assist Hi dmed

[Abhi @abhishekl]

@UC_Assist Hi can you dm me

[Abhi @abhishekl]

@iarzhannikov @abhip_me That rare use case for std::launder, this example would've been much more fun with std get/put money

[Abhi @abhishekl]

@theliverdoc @KysersozeSoze I wonder if 1 kg of homeopathic water pills pass through airport security, liquid restrictions and all

[Abhi @abhishekl]

@jorandirkgreef Since fille byte is aligned, and assuming the write is synced to disk, ideally for a non interrupted read only 0, 4096. unless something is buggy at device driver level where 512*n is still a possibility with classical format

[Abhi @abhishekl]

@jorandirkgreef 0, 4096 or anything else in between if we're close to EOF?

[Abhi @abhishekl]

@abhi9u @arpit_bhayani Many memory allocators Tc/je malloc work quite well with THP. Also they are quite bad in returning the memory back. Subtle bugs with madvise free are cherry on top
https://t.co/kr6zv74tmP

[Abhi @abhishekl]

@abhi9u @therajiv Do you have any good recommendations for resources on understanding ASM ?

[Abhi @abhishekl]

@arpit_bhayani Even with O_DIRECT & fsync storage devices could potentially acknowledge the command when they are in the device cache, (& power failure scenarios)so other than pulling out your Distributed Systems toolbox and doing some logging/journaling you literally can't ensure persistency

[Abhi @abhishekl]

@vedang @theliverdoc Test your B12 and D levels, even with a vegetarian diet, B12 can drop to dangerously low levels

[Abhi @abhishekl]

@swapnakpanda @Abhishekcur Indeed, +1 to that. Ideal to split across multiple channels also

[Abhi @abhishekl]

@arpit_bhayani Also nice about reading kernel code is the RCU stuff sprinkled everywhere. Nice little rabbit hole into concurrency

[Abhinav Upadhyay @abhi9u]

You can find it here: https://t.co/hoqSo1dfLo (I will open it in 3 weeks)

PDF version here: https://t.co/FgMkJIXkst

[Abhi @abhishekl]

@wangbin579 The paper in qn https://t.co/Q2HEKz7uHb

[Abhi @abhishekl]

@wangbin579 Looks very interesting, at the read of abstract, similar to session guarantees paper for dist sys in db world, but still need to read the paper wholly

[Abhi @abhishekl]

@abhishekcode42 For single precision floats, you'd miss after 24 Bits, so values after ~16 million (1<<24) will be incorrectly represented, you'd start rounding to the next 2, 4... etc for every subsequent bit.