Azure Storage access 2 times slower on some Azure machines

Question

I have 3 differents subscription. (let's call it SubA, SubB, SubC)
My storage is on SubA.
I created 1 VM on SubA, one on SubB, one on SubC.
I tried any combination of Location, Affinity group, and VM Size. The result is always the same :
On SubA and SubB, the azure throughput of the following code is 2000 Transaction/S
The VM on SubC, it is consistently 6000 Transaction/S.
Here is the code :
internal static void SetThrottling()
ServicePointManager.UseNagleAlgorithm = false;
ServicePointManager.Expect100Continue = false;
ServicePointManager.DefaultConnectionLimit = 100;
private static void Azure()
SetThrottling();
var client = IndexerConfiguration.FromConfiguration().CreateBlobClient();
var container = client.GetContainerReference("throughput2");
container.CreateIfNotExists();
int count = 0;
var threads = Enumerable.Range(0, 30).Select(_ => new Thread(__ =>
while (true)
var testStream = new MemoryStream(new byte[] { 1, 2, 3, 4, 5, 6 });
// Create a random blob name.
string blobName = string.Format("test-{0}.txt", Guid.NewGuid());
// Get a reference to the blob storage system.
var blobReference = container.GetBlockBlobReference(blobName);
// Upload the word "hello" from a Memory Stream.
blobReference.UploadFromStream(testStream);
// Increment my stat-counter.
Interlocked.Increment(ref count);
})).ToArray();
foreach (var test in threads)
test.Start();
while (true)
Thread.Sleep(2000);
Console.WriteLine(count + " / s");
Interlocked.Exchange(ref count, 0);
Result on SubA :
Result on SubC : 
SubC results is independant of the Location and VM Size, as for SubA and SubB.
Tests are done on plain newly created empty VM.
Any idea ?
The only difference I have seen is that SubC is maybe not a Bizpark subscription, whereas the two previous are. (Test environment)

Answer 1

Hi,
Thank you for reaching out to us.
I am currently researching to gather more information with regards to your request.
I shall revert back to you with an update at the earliest.
Sincerely appreciate your patience.
Regards,
Shirisha Paderu