Is there any difference in performance when we connect to S3 via S3 API versus via Hadoop Filesystem [closed] - amazon-s3

Question

Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 5 days ago.
Improve this question
I want to create a java utility to read S3 bucket information.
We can connect to s3 via native s3 APIs and the Hadoop filesystem approach.
Approach 1: Using S3 APIs
AmazonS3 s3client = AmazonS3ClientBuilder
.standard()
.withCredentials(new AWSStaticCredentialsProvider(credentials))
.withRegion(Regions.valueOf(region))
.build();
Approach 1: Using Hadoop Filesystem:
configuration.set("fs.s3a.access.key","XXXXXXXXXXX");
configuration.set("fs.s3a.secret.key","XXXXXXXXXXX");
configuration.set("fs.s3a.impl","org.apache.hadoop.fs.s3a.S3AFileSystem");
configuration.set("fs.s3a.endpoint","http://127.0.0.1:8080");
UserGroupInformation.setConfiguration(configuration);
fileSystem = new Path("s3a://"+ bucketName).getFileSystem(configuration);
Do we know when we use which approach? Which approach is more efficient to read data?
In my observation, the filesystem route is slower. But I have not found any documentation supporting the performance difference.

Answer 1

Performance shouldn't be the only factor. If you want higher performance, or at least better file operation consistency guarantees, look into S3Guard.
But if you have to create a Java client that will only ever talk to S3, and never needs to integrate with Hadoop ecosystem, or use other Hadoop compatible filesystems (HDFS, GCS, ADLS, etc), then you should use plain AWS SDK.
If you're trying to run some mocked S3 service (or MinIO) on 127.0.0.1, then that's not a proper benchmark to a real S3 service