We deployed a lambda@edge with the cloudfront. We see that the default limit is 1000 lambda concurrency per AWS Region.
As I understand, lambda@edge is executed at the edge location, and this edge location is attached to a region (for example London or Frankfurt). The first time cloudfront called, the lambda@edge is pulled from region to edge location and executed here.
So the limit 1000 lambda concurrency is only in the region, not at edge location ?
Your concurrency limits are per region. There's also a triggers per second limit, per region, but both of these limits can be increased upon request to AWS support.
Each Lambda@Edge invocation runs under the control of one specific AWS region near the edge location through which the viewer has connected to CloudFront, for that request. You can easily observe this because the AWS_REGION environment variable is always set to the controlling region.
Where Lambda@Edge triggers are physically executed is not actually documented -- other than "at the edge," of course -- but on November 30, 2016, AWS announced a new two-tiered architecture for CloudFront, with the existing outer tier called "global edges" and the new inner tier called "regional edges." Literally one day later, on December 1, the Lambda@Edge preview was announced... so the edges in question may be the ones in the inner tier.
