First of all this is a Replicate question, not an Enterprise Manager question as the EM is only used to trigger the tastk start but has no further control. Please consider moving this topic, or better still (imho) create a new topic in the Replicate forum with more, more concrete, information.
For example, as @Steve_Nguyen asks, what is the max-file size and full load tuning? Those are easy ones but there are more and it is best to simply provide the (relevant parts) of the JSON - obfuscation any sections which you believe are sensitive and removing highly repetitive sections (like 90 out of 100 tables).
That's the Replicate information we need to try to help. Now what about the data? "a lot of data " is hardly useable information. How many gigabytes? Given the network pipe throughput maximum, what is your back-of-the-enveloppe minimum time to load and how does that relate to the actual?
When you reach the maximum of the network pipe, adding more load streams only hurts - you need enough streams to fill the pipe, but not more. That's likely closer to 5 than to 10, but it could be 20.
Splitting a task into two can create more parallelism but reduces control and increases contention. Those hundred 'less important' tables will appear as important and will slow down the 10 important ones (or visa versa). It may be better, and easier to manage, to create a single tasks and use the per-table priority options within that task.
I haven't looked at the reptask logs for at Snowflake full-load tasks in a while. The CDC tasks have a nice log lines showing the start and end for the network transfers for each CSV file. Can you get that for full-load? May with TRACE option on TARGET_LOAD? Please share the result.
Good luck, and see you in the Replicate Forum!
Hein.
