You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The source throughput of q15 under 1CN setting is almost the same as under 3CN setting (both setting colocated with compactors)
1 CN:
3 CN:
Two settings are using the same amount CPU in total.
SELECT
TO_CHAR(date_time, 'yyyy-MM-dd') as day,
count(*) AS total_bids,
count(*) filter (where price < 10000) AS rank1_bids,
count(*) filter (where price >= 10000 and price < 1000000) AS rank2_bids,
count(*) filter (where price >= 1000000) AS rank3_bids,
count(distinct bidder) AS total_bidders,
count(distinct bidder) filter (where price < 10000) AS rank1_bidders,
count(distinct bidder) filter (where price >= 10000 and price < 1000000) AS rank2_bidders,
count(distinct bidder) filter (where price >= 1000000) AS rank3_bidders,
count(distinct auction) AS total_auctions,
count(distinct auction) filter (where price < 10000) AS rank1_auctions,
count(distinct auction) filter (where price >= 10000 and price < 1000000) AS rank2_auctions,
count(distinct auction) filter (where price >= 1000000) AS rank3_auctions
FROM bid
GROUP BY to_char(date_time, 'yyyy-MM-dd');
We notice that the query is grouping by to_char(date_time, 'yyyy-MM-dd'). It is likely that during one period of time, the source input data all belong to the same day, and we are observing data skewness because of Two settings are using the same amount CPU in total.
The source throughput of q15 under 1CN setting is almost the same as under 3CN setting (both setting colocated with compactors)
1 CN:
3 CN:
Two settings are using the same amount CPU in total.
We notice that the query is grouping by
to_char(date_time, 'yyyy-MM-dd')
. It is likely that during one period of time, the source input data all belong to the same day, and we are observing data skewness because ofTwo settings are using the same amount CPU in total.
The plan:
So we wonder if two-phase aggregation can help in this case.
But
set rw_force_two_phase_agg to true
does not change the plan.Any other ideas to improve?
Let's also wait for the numbers of other systems.
The text was updated successfully, but these errors were encountered: