6) Why not restrict IDs to 53bits?

A Snowflake ID is composed:
 * 41bits for millisecond precision time (69 years)
 * 10bits for a configured machine identity (1024 machines)
 * 12bits for a sequence number (4096 per machine)

The factor influencing the length of the ID is the time. For a 53bit ID this
would mean only 31bits are available for the time. 31bits is only enough for
24 days (2147483648/(1000*60*60*24)) of time.

Reducing the resolution of the timestamp would prevent a K-sorted resolution
of 1 second or less.

Reducing the configured machine identity or sequence number by 1bit would
mean we couldn’t scale Twitter, or operate our infrastructure in an
uncoordinated high-available way.

Interesting ... so you have the theoretical capacity to scale to 2**22 (about 4 million) tweets per millisecond? Even 4 million tweets a second seems unrealistic, as does a single "machine" only being able to generate 4096 IDs. I think if you're really expecting this kind of volume, the FPGA vendors probably can help you out. We are talking clocks and counters, here, right, not Javascript interpreters or robust linear regressions? ;-)

Ah, well, I'll check back on you guys in 69 years to see how you're holding up. ;-)

M. Edward (Ed) Borasky
http://borasky-research.net http://twitter.com/znmeb

"A mathematician is a device for turning coffee into theorems." - Paul Erdos

Twitter developer documentation and resources: http://dev.twitter.com/doc
API updates via Twitter: http://twitter.com/twitterapi
Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list
Change your membership to this group: 

Reply via email to