Get object from AWS S3 as a stream

In .NET 4, you can use Stream.CopyTo to copy the content of the ResponseStream (that is a Amazon.Runtime.Internal.Util.MD5Stream) to a MemoryStream.

GetObjectResponse response = await client.GetObjectAsync(bucketName, keyName);
MemoryStream memoryStream = new MemoryStream();

using (Stream responseStream = response.ResponseStream)
{
    responseStream.CopyTo(memoryStream);
}

return memoryStream;

Where client.GetObjectAsync(bucketName, keyName) is an alternative to calling GetObject with the request you are creating.

Even cheaper way is to use pre-signed URLs to objects in S3. That way you can return expiring URLs to your resources and do not need to do any stream copies. Very low memory is required for that so you can use a very small and cheap VM.

This approach would work for a few resources and only a few clients. With more requests you may hit AWS API limits.

using (client = new AmazonS3Client(Amazon.RegionEndpoint.USEast2))
{
    var request = new GetPreSignedUrlRequest()
    {
        BucketName = bucketName,
        Key = keyName,
        Expires = DateTime.UtcNow.AddMinutes(10),
        Verb = HttpVerb.GET, 
        Protocol = Protocol.HTTPS
    };
    var url = client.GetPreSignedURL(request);
    // ... and return url from here. 
    // Url is valid only for 10 minutes
    // it can be used only to GET content over HTTPS
    // Any other operation like POST would fail.
}