colin-scott / interactive_latencies Goto Github PK

.google.com are not available in China without VPN

Could you consider switching to https://code.jquery.com CDN ?

The two above are blank.

https://ibb.co/T07ZTsZ

Just in case the picture gets deleted, the two I'm talking about are "Send 2,000 bytes over commodity network: 22ns " and "Read 1,000,000 bytes sequentially from SSD: 31,000ns ≈ 31μs".

As a former Netrek player I am intimately familiar with the round trip latency from Europe to the US in the 90s. It was more like 300ms to the west coast for most of the 90s. Even the east coast didn't dip below 100ms until several years later.

Hi. It's a very nice website. But I have some questions about the data.
On the website,
"read 1000,000 bytes sequentially from memory is 3us", so the memory bandwidth should be 1MB/3us = 333GB/s.
"read 1000,000 bytes sequentially from SSD is 49us", so the SSD bandwidth should be 1MB/49us = 20GB/s.
"read 1000,000 bytes sequentially from disk is 825us", so the disk bandwidth should be 1MB/815us = 1.2GB/s.
From my point of view, these data are much bigger than the real products' performance.

Hi! Thanks a lot for your research publication!
Currently, the website for the interactive timeline/chart is down, could you try fixing it back up?

Thanks in advance!

Please consider adding latency numbers for NV-RAM. I understand the technology is new and is still evolving; still adding the present numbers would give a good comparative understanding to the readers.

Visit http://www.eecs.berkeley.edu/~rcs/research/interactive_latency.html and move slider to the year 2013. You will notice Send 2000 bytes over commodity network value to be 0.0ns which does not seem right. The micro-seconds value is OK.