Large file download performance RRS feed

  • Question

  • I'm evaluating the download performance of large files (e.g. 2GB) served through a CDN with a Storage Account origin. On the first request where there is a TCP_MISS and the blob is served from the origin I get good performance - better than the raw origin URL. My naive assumption is that this is because the content is able to use a faster connection between the Edge server and the origin data center.

    On an immediate subsequent request when there is a TCP_HIT I get download speeds that are significantly slower than the raw origin URL, with the stream periodically waiting before resuming. After waiting for some time, follow up requests generally have good performance.

    Is this the expected behavior? Can you shed some light on how the blob is cached - i.e. is the slow initial cache performance due to the blob being in the process of being copied to the Edge server (if this is not copied when originally streamed)? 

    Wednesday, August 12, 2015 8:27 PM


  • As you are currently following up with our support team on this issue I will just provide generic feedback for others who may read this thread in the future. Overall the behavior described above isn't expected. To start initial investigation on this we would need example URL's where this issue is being seen along with associated HTTP Response headers that illustrate this issue. Multiple tool can be used to view and capture HTTP response headers including Fiddler, Wget, cURL, Internet Explorer Developer Tools, Chrome Developer Tools, or Firebug.

    Thursday, August 20, 2015 1:26 AM