Twitter, PayPal reveal database performance

Re:Twitter, PayPal reveal database performance by aidos on Friday October 05, 2012 @05:19AM (#4616183)

"get a timestamp error if one node attempts to read data that was written by another node in the future (due to data moving across datelines)"

Honest question: Which part of a replication/db system would be using time stamps that weren't universal?

[ Parent | on HN ]

Re:Twitter, PayPal reveal database performance by blrgeek on Friday October 05, 2012 @07:12AM (#4616395)

And that's the part that Google Spanner gets around with Atomic clocks and GPS clocks.

[ Parent | on HN ]

Re:Twitter, PayPal reveal database performance by donavanm on Friday October 05, 2012 @06:19AM (#4616293)

Just the concept of using time for the logical clock is sketchy. It's not monotonically increasing, it drifts, it can go decrease, etc... Curious what their cross host/region ordering and consistency looks like.

[ Parent | on HN ]

Re:Twitter, PayPal reveal database performance by sanxiyn on Friday October 05, 2012 @06:30AM (#4616314)

"An atomic clock is not that expensive." -- http://research.google.com/archive/spanner.html

[ Parent | on HN ]

Re:Twitter, PayPal reveal database performance by jrockway on Saturday October 06, 2012 @12:21AM (#4619781)

Indeed. A GPS clock accurate to 50ns is around $25 on eBay.

[ Parent | on HN ]

Re:Twitter, PayPal reveal database performance by mikecb on Friday October 05, 2012 @01:17PM (#4617735)

Especially since there are 24 circling the earth that send out time continuously.

[ Parent | on HN ]