S3 Performance

Question 1

What is an S3 prefix?

Answer 1

• An S3 prefix is the part of the file path between the bucket name and the object (ex. ‘mybucketname/folder1/subfolder1/myfile.jpg’ has prefix ‘/folder1/subfolder1’)

Question 2

What does KMS have to do with performance?

Answer 2

• Uploading and downloading files count towards the KMS quota

_{The exact amount for the quota is region-specific, but is either 5,500; 10,000; or 30,000 requests per second.}

Question 3

Can you request a quota increase for KMS?

Answer 3

Yes

Question 4

What is Multi-Part Upload?

Answer 4

• The Idea behind Multi-part uploads is to parallelize an upload (and thus increase its efficiency)
• This is done by splitting a big file into parts and then uploading the individual parts in parallel

Question 5

What are S3 Byte Range Fetches?

Answer 5

• S3 Byte Range Fetches allow you to speed up downloads by parallelizing them.
• You parallelize by specifying exact byte ranges
• If there is a failure in the download, it is only for a specific byte range.

Question 6

What would you use if you just want to download partial amounts of a file?

Answer 6

S3 Byte Range Fetches with just the relevant bytes

Question 7

At what file size is Multi-Part Upload recommended?

Answer 7

100MB

Question 8

At what file size is Multi-Part Upload required?

Answer 8

5GB

Question 9

What do S3 prefixes have to do with performance?

Answer 9

• They are important for performance because requests are measured per second per prefix
• So, y​ou can get better performance by spreading your reads across different prefixes

Question 10

What is a good common way to ensure lots of different prefixes in S3?

Answer 10

Go by timestamp of upload

_{(there are many right answers here)}