Translation API - User Rate Limit Exceeded after increasing quotas substantially

[scientific marcom]

Hi there!

Under this account, we have implemented the paid Translation API and have experienced user rate limit exceeded issues.  In order to address this, we sent in a quota increase yesterday and it shows to have been implemented. Our limits are now:

• V2 and v3 Characters per day: Unlimited
  
• V2 and v3 Characters per 100 seconds: 10,000,000
  
• V2 and v3 Characters per 100 seconds: 10,000,000
  
• V2 and v3 requests for the list of supported languages per 100 seconds: 1,000
  
• v3 requests (Beta) per minute: 6,000
  
• v3 requests (Beta) per minute per user: Unlimited
  

We are using the v2 service (to our knowledge) as the actual implementation is with Lingotek which bridges their service to Google Translate as a courtesy to certain folks.  It is; however, using our Google Cloud account.

Although we are apparently exceeding something, our quotas page does not indicate that we are exceeding any limits although the per user quota usage is not displayed, so we're a bit in the dark there.

So, now we're lost in the dark trying to figure out what is going on. We are trying to translate a web site with pages with approximately 20,000 characters apiece into 10 languages.  Even when we send very few items, we are repeatedly getting these 403 errors.  

After exceeding these limits, is there some sort of delay before we are able to submit again?  Also, is there some sort of delay before the user rate limit increases are actually in place?  We have not received any messages indicating as much; however, reviewing issues in stackoverflow I'm asking these questions now.  Thank you!

Thank you for any help you can provide on this.

[Konstantin Savenkov]

This may depend on how Lingotek deals with your 20,000 byte requests. I.e. if they do segmentation and then send single-segment requests in a burst, this may easily exceed 6,000 requests per 1-minute quota. 

Typically, charts on the Quotas page provide good insights into that. You may check when you see the 403s in the API dashboard and then see what was the requests and bytes volume at that time.

Some of the quotas have a delay attached (100s), hence the optimal strategy is to shape your traffic according to the quotas and back-off once you hit the limit. Not sure in Lingotek software does that.

cheers,

Konstantin Savenkov,

Intento, Inc.

[Kan Qiao]

Hi, there is an known quota issue in our system. Please let us know your project id so we can override it for you