An office switches to Kladde.Verlauf and brings a year of appointments with it. Office Nord expects 7,360 appointments a year (F-02, an assumption of the task), roughly 640 a month. Writing them one after another takes a while, and with an error in the middle nobody knows what arrived. This lesson shows two tools for this and when each fits.
Not run against Azure. The code was compiled with Microsoft.Azure.Cosmos 3.64.0 and Newtonsoft.Json 13.0.4. What was run are the planning of the batches, the evaluation of the results and the summary of the bulk import, against stand-ins (Exercise 1). A batch or a bulk run against the account did not take place; everything about the behavior of the service comes from Microsoft Learn, read on the day of writing.
Three ways to write several items
| Single operations | Bulk | TransactionalBatch | |
|---|---|---|---|
| Atomic | per item | no | yes, all or nothing |
| Partitions | any | any | one logical partition |
| Quantity | few | very many | at most 100 operations |
| Errors | per call | per operation | one result for all |
| Goal | low latency | high throughput | consistency of several items |