After a lot of failed attempts to download the 2.15GB PCM 176K ZIP file I have finally managed to successfully download this file.
Previous attempts have been shaped once the download speed exceeded 114MBit/sec AND the downloaded file exceeded some threshold - varying between 1.0 GB and 1.5GB as found by the practical download failures! Repeated attempts on the same file were also shaped/killed at lower size thresholds too.
It would appear that Australia’s Telstra BigPond Cable network now includes a download shaper/kill mechanism that is incorrectly configured such that HTTP downloads of a large ZIP file can be shaped to a 25Kbit/sec download speed - then shortly thereafter killed-off. But - you need to meet two criteria in order for the shape&kill to occur. First the download speed must exceed around 114.6MBit/sec and as well a downloaded file must have reached 1.5GB in size before shape&kill triggers.
I am located in Melbourne,AU and have taken up the option of 100MBit/sec download speeds for an additional fee per month. For that the DOCSIS cable modem is configured such that speeds of 100MBit/sec are enabled and depending on your location in the cable network and the latencies of the network between you and the download site will depend on actual download speeds. Multiple stacked sessions are more likely to make use of the higher bandwidth than a single session from a remote download server. When downloading from a high capacity download server like PS Audio’s download server for the ONE files - the available bandwidth is such that when the cable network is lightly loaded I have reached speeds of 117MBit/sec to the cable modem - the modem itself has a 1GB/sec ethernet interface and I have upgraded my local LAN to 1GB/sec speeds throughout so am able to take advantage of the increased speed even above 100Mbit/sec the cable link is nominally rated at.
So it was somewhat disconcerting to find that 1.5GB through the download of the 2.15GB PCM file that the download session suddenly slowed to 25Kbit/sec then shortly thereafter the session died. And i also subsequently found that the speed had to exceed around 114Mbit/sec else the session continued - and as the traffic load is quite light during the morning and early afternoon in Australia such that the download speed will cycle down and up and down depending on the overall network load - with the speed gradually arcing up - but prior to having downloaded the 1.5GB the speed exceeding 114MBit/sec had no effect - it needed a combination of both to trigger this shape&kill mechanism - which was unknown to both PS Audio and myself.
By chance I noticed an article late in 2015 stating that Telstra was planning to trial a download shaping mechanism to shape P2P traffic as they don’t wish to have high levels of traffic between users as it causes added congestion. They were stated as planning to use Deep Packet Inspection techniques to determine the download type (ie torrents or gaming traffic) and then shape it accordingly. There was no stated intention to kill download or P2P traffic so at the moment I am assuming the shape&kill of the PS Audio download server traffic is collateral damage as it were of some test shaping method that should never have been configured into the live Production Bigpond Cable network but has and has also been overlooked - subsequently shaping&killing downloads that meet the faulty criteria.
This afternoon I ran some further tests - confirmed the shape&kill mechanism was still active - then established a VPN for USA origin traffic so as to bypass the shape&kill mechanism since the DPI technique is unable to determine packet type within an encrypted VPN session.
That did work - but at a significant time penalty - instead of the typical 3 minute download of this large 2.15GB file it now took 128 minutes - but at least it was successful.
Later tonight I will repeat the download - but using a direct single session (no VPN) whilst the Netflix traffic load is high which also significantly loads the cable network. With a heavily loaded cable network the peak download speed should be well below 100MBit/sec so the trigger mechanism for the shape&kill mechanism should not activate allowing me to prove another way of bypassing this problem shape&kill mechanism that won’t have such a high time penalty for the download…
My first choice though is to get Telstra to fix their defective shape&kill mechanism to not interfere with external traffic and in particular not interfere with download traffic from the PS Audio download server. In any case there should not be such a low shaped speed and nor should a user session be killed either as a result of downloading at a high speed any large file. Having paid to use the network we do not expect to have legitimate downloads killed off in this manner.
So - if you are in Australia either in Melbourne or Sydney using Telstra Bigpond Cable Network - and with the higher 100MBit/sec speed option in effect - then be aware that I do not know the extent of this download shape&kill mechanism - but I do know that until I used a VPN I have been unable to successfully download the 2.15GB PCM file from the ONE download collection - the cause of the failure due to what I surmise is a wrongly configured shape&kill mechanism. It is likely that the extent of the shape&kill is quite wide and may even extend into the ADSL network area - depending on the actual location in the network of this shaping mechanism.
Despite the heading, the creation or not of an account identity plays no part in this problematic download - and is a separate issue in itself which has yet to be examined. The priority has been to try and find a method to allow successful download of the 3 files. The download for the 1.0GB DSD file and much smaller booklet file were immediately successful which made the shaping at 1.5GB part-download and then kill a complete surprise. This is the first time I have ever had such difficulties in downloading a straightforward file from a download server with some foreign failure mechanism terminating the downloads part way through.
Thanks especially to Paul and Sean of PS Audio for their help & assistance in allowing me to multiply repeat downloads in order to try and diagnose just why the downloads were failing - and more importantly how to work around the problem until I can try and get Telstra to correct this faulty download shape&kill mechanism. These download failures made no sense as does the apparent dual trigger of a certain file size and peak download speed before the session shape and kill will be invoked.
I have also finally been able to hear the 176K PCM downloads which was the whole point of this exercise in obtaining One - although it has taken a lot of time and effort and failed downloads before I was finally successful.
