I’m looking to evaluate MapD for a finance client. A couple of their key requirements are:
- transactions, to give consistent views of data sets
I see support for updates/upserts is planned (Update / Upsert support). Any news on when that is likely to be?
For transactions, in some use cases we would need clients to query consistent snapshots isolated from data being ingested, then make the data set visible atomically. Transactions wouldn’t have to be supported explicitly, as we can insert data with logical timestamps or transaction sequence numbers, and perform queries bounded by those timestamps which I imagine would be lightning fast. But we’d need consistency guarantees on when data becomes visible. Is a row consistently, durably, visible across a cluster once a JDBC insert statement has completed, or a Kafka consumer has processed a record?