After the quota issue is resolved, we can migrate the current findit for crash from googleplex to appspot, so the remote_api would work fine for findit for crash.

From the statistics of last 2 days:

For Fracas:
~622 total crahses/day, 2~40 crashes/30mins (Fracas sends a bunch of requests per half hour)
gitiles-blame requests(git blame): 4~15/5mins
gitiles-view requests(git changelogs and DEPS file git changelog): 5~20/5mins

mmandlis@, can you estimate how many crashes would cracas send to findit? mbarbella@, can you give me some estimation about cluterfuzz if later we integrate?

Thanks!

At the moment we run Cracas Analysis hourly and process about 10K signatures. I think we'll start with just querying FindIt about the ones "new" and "spike" and "up" on canary.
Then we'll expand to sending "new" and "spike" and "up" on additional channels. This might be several hundreds per hour. Is this reasonable? 
Thanks!

I think it's reasonable, however I still need some clusterfuzz data and ask git team. 

Thanks!

Regarding ClusterFuzz, it catches a few hundreds of crashes a day, although they might not even out in the 24 hours.

As far as the burstiness issue, we could always just have/let findit queue things up and deal with them as it has quota available. We'll want to monitor it to make sure things don't get too backlogged, but that can be fairly easily automated. In the long run we'll want to be smarter about how we issue requests to gitiles and cache responses, but I don't know that that's critical just yet.

After integration with Cracas, the quota seems fine. Keep monitoring and update the quota if needed.