Bigtable & Time Series Data

preview_player
Показать описание
Time series data is conducive to the growth of a company but you want the best performance when collecting these field data. In this episode of Cloud Performance Atlas, Colt McAnlis helps some middleware pros with their bigtable performance. Will we upgrade to the latest version of performance? Stay tuned to find out!

Рекомендации по теме
Комментарии
Автор

Great update. We used last 2 digits of the mili-second into our time series data which provided us with 99 semi-random for row-key distribution. The only issue we found is that big table was way too expensive for time series data. Unless there has been a huge drop in pricing, it really only feasible in premium services.

DerekAndersonMinnesota
Автор

In the example you showed up, as sequential Timestamp is still in the rowkey after field promotion, wont it still cause hotspotting?

parvindersingh
Автор

I’m appending a random uuid to my keys and I’m seeing hot spots for writes but the rows written are evenly distributed. Performance is very bad for me on an ssd production cluster.

griesrt