[QUERY] Performance difference between UploadBlobAsync and OpenWriteAsync

[shaaradmsft]

Library name and version

Azure.Storage.Blobs 12.19.1

Query/Question

Hi, I am trying to figure out the reason behind the performance difference between UploadBlobAsync and OpenWriteAsync methods.

Here is the minimal test code that I have -

using Azure.Storage.Blobs;
using Azure.Storage;
using System.Runtime.CompilerServices;
using System.Diagnostics;
using Azure.Storage.Blobs.Specialized;

namespace BlobUploadTest
{
    internal class Program
    {
        static async Task Main(string[] args)
        {
            var accountName = "accountname";
            var accessKey = "accessKey";
            var containerName = "containerName";

            var keyCred = new StorageSharedKeyCredential(accountName, accessKey);
            var blobClient = new BlobServiceClient(new Uri($"https://{accountName}.blob.core.windows.net"), keyCred);

            var containerClient = blobClient.GetBlobContainerClient(containerName);

            var sw1 = new Stopwatch();
            sw1.Start();

            {
                using var fs1 = new FileStream(@"C:\temp\LocalExport\dataFile.data", FileMode.Open, FileAccess.Read);

                await containerClient.UploadBlobAsync("dataFileUploadBlobAsync.data", fs1, CancellationToken.None);
            }

            sw1.Stop();
            Console.WriteLine($"Time taken by UploadBlobAsync is {sw1.ElapsedMilliseconds} ms");

            var sw2 = new Stopwatch();
            sw2.Start();

            {
                using var fs2 = new FileStream(@"C:\temp\LocalExport\dataFile.data", FileMode.Open, FileAccess.Read);

                var blockClient = containerClient.GetBlockBlobClient("dataFileOpenWriteAsync.data");

                using var blobUploadStream = await blockClient.OpenWriteAsync(true, null, CancellationToken.None);

                await fs2.CopyToAsync(blobUploadStream);
            }

            sw2.Stop();
            Console.WriteLine($"Time taken by OpenWriteAsync is {sw2.ElapsedMilliseconds} ms");
        }
    }
}

Both the ways to upload blobs use a file stream that reads from a locally present file (to make sure the incoming stream isn't involving any network I/O etc.) and upload it to a blob store container.

This program was run on an Azure VM (with release configuration built, if it matters) in the East US region, and it was pointed to a storage account in the East US region. The blob being uploaded was of size 256MBs. The output was -

Time taken by UploadBlobAsync is 31936 ms
Time taken by OpenWriteAsync is 165537 ms

Can you please help me understand why there's a significant difference in the two upload methods? I checked that if no BlockBlobOpenWriteOptions are given to OpenWriteAsync, the default BufferSize is 4MB. So not sure what else could be a contributing factor to the slowness of OpenWriteAsync.

Environment

Host:
  Version:      8.0.4
  Architecture: x64
  Commit:       2d7eea2529
  RID:          win-x64

.NET SDKs installed:
  No SDKs were found.

.NET runtimes installed:
  Microsoft.AspNetCore.App 6.0.29 [C:\Program Files\dotnet\shared\Microsoft.AspNetCore.App]
  Microsoft.NETCore.App 6.0.29 [C:\Program Files\dotnet\shared\Microsoft.NETCore.App]
  Microsoft.NETCore.App 8.0.4 [C:\Program Files\dotnet\shared\Microsoft.NETCore.App]

[github-actions]

Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @xgithubtriage.