Transparent memory offloading: more memory at a fraction of the cost and power
1–10 of 62 posts
Re: Transparent memory offloading: more memory at a fraction of the cost and power
#2Nice, they have reinvented virtual memory and paging to disk.
Re: Transparent memory offloading: more memory at a fraction of the cost and power
#3or "Meta re-invents swap disk"
Re: Transparent memory offloading: more memory at a fraction of the cost and power
#4or "Meta re-invents swap disk"
You should read the whole article, because TMO does differ from traditional swapping. Notably, swapping usually occurs during periods of extreme memory pressure, where as TMO will offload memory much sooner and more intelligently.
Re: Transparent memory offloading: more memory at a fraction of the cost and power
#5>Using a compressed-memory back end, TMO saves 7 percent to12 percent of resident memory across five applications. Multiple applications’ data have poor compressibility, such that offloading to an SSD proves far more effective.
that in general sounds strange. My experience across different application types over more than 2 decades has been very different (not about SSD, i mean the compression in memory have been always a very efficient approach).
> Specifically, machine learning models used for Ads prediction commonly use quantized byte-encoded values that exhibit a compression ratio of 1.3-1.4x.
yep, i haven't used compression for that. That though calls for a question - how many of those models do FB have in memory. Any chance FB instantiates such a model for every user/request? :)
Re: Transparent memory offloading: more memory at a fraction of the cost and power
#6Computer storage hierarchies are sure getting complex with CXL and NVM in the mix.
Re: Transparent memory offloading: more memory at a fraction of the cost and power
#7or "Meta re-invents swap disk"
You should read the whole article, because TMO does differ from traditional swapping. Notably, swapping usually occurs during periods of extreme memory pressure, where as TMO will offload memory much sooner and more intelligently.
So… paging then? Slightly improved paging?
Re: Transparent memory offloading: more memory at a fraction of the cost and power
#8Nice, they have reinvented virtual memory and paging to disk.
Indeed. The side describes TMO in bombastic terms, but this is just an incremental improvement to zram and swapping.
Yet, a smarter algorithm to swap intelligently was long due.
Re: Transparent memory offloading: more memory at a fraction of the cost and power
#9Earlier quoted context omitted.
You should read the whole article, because TMO does differ from traditional swapping. Notably, swapping usually occurs during periods of extreme memory pressure, where as TMO will offload memory much sooner and more intelligently.
So… paging then? Slightly improved paging?
Only one way to know for sure!
Re: Transparent memory offloading: more memory at a fraction of the cost and power
#10Nice, they have reinvented virtual memory and paging to disk.
did you even read this?