PROFESSIONAL-CLOUD-DEVELOPER Question #34: Real Exam Question with Answer & Explanation

Question

You are analyzing your application's performance. You observe that certain Cloud Bigtable tables in your cluster are used much more than others, causing inconsistent application performance for end users. You discover that some tablets have large sections of similarly named row keys and are heavily utilized, while other tablets are running idle. You discover that a user's ZIP code is the first component of the row key, and your application is being heavily used by profiles originating from that ZIP code. You want to change how you generate row keys so that they are human readable and so that Cloud Bigtable demand is more evenly distributed within the cluster. What should you do?

Options

• AUse serially generated integer values.
• BUse a concatenation of multiple human-readable attributes.
• CUse a subset of the MD5 hash of the row contents.
• DUse UNIX epoch-styled timestamps represented in milliseconds.

Explanation

A is not correct because, depending on the distribution of the underlying values, serially generated integer values may lead to hotspotting. B is correct because using a sufficient number of delimited attributes can provide sufficient C is not correct because this is no longer a recommended best practice: it makes it difficult to troubleshoot issues. D is not correct because timestamps are poor candidates for row keys, and in this case inadvertently lead to ID collisions if multiple updates happening in the same millisecond execute in close succession. https://cloud.google.com/bigtable/docs/schema-design#types_of_row_keys https://cloud.google.com/bigtable/docs/schema-design-time- series#ensure_that_your_row_key_avoids_hotspotting

No community discussion yet for this question.