Re:Summary of the Amazon DynamoDB Service Disruption
by ckozlowski on Wednesday September 23, 2015 @07:58AM
(#10264465)
I just asked the question in a separate post before seeing yours, but I questioned this as well. Would it have been better to have a separate instance of DynamoDB independent from the one customers are writing to? I understand this is added cost/overhead, but it would have resulted in isolating the failure to the customer side only, I'd think.
> ... subsequent impact to other AWS services that depend on DynamoDB ...
And the metadata service is part of DynamoDB:
> The membership of a set of table/partitions within a server is managed by DynamoDB’s internal metadata service.