Records made through API don't execute bots in parallel

[MultiTech]

If you create a set of records through the API, and those records kick off automations when a new record is made (in the table that the API is creating records in)...

• The bots DO NOT run in parallel - instead they seem to run one by one

Which completely negates the benefit of running through automation.

------------------------------------------------------------------------------------

Let's consider a scenario:

• We're creating 100 records
• Each record kicks off a bot that runs a script that takes 15 seconds to run

If this were to run in parallel (with an enterprise license with a degree of parallelizm of 10):

• 100 x 15 seconds = 1,500 seconds
• 1,500 / 10 processing paths = 150 seconds
• 150 / 60 = 2.5 minutes

But since things run one by one, each one waiting for the full completion of the previous row before it will kick off the next bot run.....

• 100 x 15 Second = 1,500 second
• 1,500 / 60 = 25 minutes

-----------------------------------------------------------------------------------------------------------------------

You can see why parallel processing is so important.

• Now consider if I'm making 1,397 records:
  • 1397 * 15 = 20,970 seconds
  • 20,970 / 60 = 349.5 minutes
  • 349.5 / 60 = 5.825 hours

(Note exactly a direct comparison; but you get the point)

-------------------------------------------------------------------------------------------------------------------------------

I would hope that things would process in parallel since the records were made through the API.

[SkrOYC]

---

@MultiTech wrote:

I would hope that things would process in parallel since the records were made through the API

---

Well, at that point you would also be Enterprise.

Remember when we talked about this with someone on the team last year and said that API was actually not intended to be open outside Enterprise and it currently is but could be fixed anytime?

[Marc_Dillon]

I've built functionality that depends on the API records being added in serial. I'm sure others have as well. That's always a big consideration for any change, is what existing things will it break. Not that parallelism here wouldn't be nice.

I wonder if the more important issue is the parallelism of the data source itself. Appsheet can't very well add records in parallel if the source itself can only process them 1 at a time.