HI team, I'm planning to do the performance testing on Pinot Cluster (using k8s) with 7000 transaction/s which is 10GB storage / h and the data will be in the storage for 6m - 1y. Which are the recommend specs on the Pinot containers and the node in terms of cpu/memory and storage?

7000 reads or writes? What’s the format of data and whats the SLA? Also what do the queries look like?

7000 Writes, and this will be JSON format, and they are few like 3 to 5 customers will query at a time. example of query looks like :

Select LOG_TS,AGENCY_ID,SCHEMA_VERSION_NM,REQUEST_TYPE_ID,ORIGIN_APP_CD,PCC_CD,TARGET_BRANCH_CD,SSL_TERM_REGION 
From uapi3
where TEXT_MATCH(SCHEMA_VERSION_NM, '/.*v51_0/')
and (LOG_TS >= '2022-01-09T00:00:00') and (LOG_TS <= '2023-06-15T23:59:59')
group by LOG_TS,AGENCY_ID,SCHEMA_VERSION_NM,REQUEST_TYPE_ID,ORIGIN_APP_CD,PCC_CD,TARGET_BRANCH_CD,SSL_TERM_REGION limit 10000;

select count(*) from uapi.translog_fact where log_date >= cast('2023-04-01' as DATE) and log_date <= cast('2023-04-30' as DATE) and log_ts > '2023-04-01T00:00' and log_ts < '2023-04-30T23:59' and request_type_desc in ('BookingEnd') and agency_name in ('MMT-Make My Trip Hierarchy', 'IBIBO WEB Hierarchy', 'Yatra Online Hierarchy', 'Travel Boutique - Hierarchy', 'ClearTrip India Hierarchy', 'One97 Comm Hierarchy', 'Easy Trip Planners Hierarchy', 'Paytm aka One97 Comm Hierarchy') and aux1='End' and aux2 LIKE '%The segment information is now updated and the pricing%

Is 10GB per hour JSON? If so it will compress quite a bit. You can use 3 servers (replication = 3), with 8-16’cores and 64 GB, as a starting point