As I wrote previous post, in August 2023 number of requests to our Proxy Repository was rapidly down by ~50% and stayed more less constant. It's a huge change from almost linear growing to stagnation. All these happen in just 2 weeks. Why?
Let me remind you the chart. It shows number of requests to Proxy Repository by day.
By the way, this chart was made by analyzing ~11GB of logs and the analyze took 12 minutes to complete on my DS718+.
To answer this question I need to go back in time and check my inquiry on Synology Account. On 8th of August 2023 I've sent question to Synology Support with the title: How often Package Center send requests to third-party repository server? I was concerned to see in the logs a lot of requests from same IP's. Those request were the same and DSM keeps sending request about 20 times within 2 minutes. It's ridicules! In my opinion it should be enough to send one query per day. But what I get in answer was like a snap on my face... let me quote:
"Unfortunately, it is not possible for us to talk about internal behaviour.
I will forward your feedback to our development team."
It hurt me so much, because me and other developers spend their spare time on making DSM better and better and doing the job that Synology itself does not want to do and delegated it to community. And when the community asks how the Package Center works... it looks like it's only their internal system, that no one should know how it is working...
Well. I'll stop to comment this at this point. Let's go back to my analyze.
One week later I discovered that number of requests started to dropping down. I thought that my hosting provider started to filter out some traffic, but that was not true (at least they said they never done such thing). The traffic was dropping, but I had something else to do, so I didn't done much to investigate it. But now I back to it and see the logs from August and with bare eye I saw, that there is no much repeated requests from same IP's! It might be the result of my question to Synology, but who knows? I wanted to proove it and I wrote another analyzer to check my theory.
So I downloaded logs from July and October 2023 so just before and after August (yes, I omit September because it was not representative month). Have a look on the chart below: July (in blue) and October (orange) shows number of requests from unique IP per day. I.e. (1-2) means that from one IP we get 1 or 2 requests per day. The Y axis means how many IP's were in the range o X axis. I had to normalize the data, because in October were 40k more samples than in July, which is ~18% more. So I scaled out July data with 1,18 factor for all ranges.
For instance in July there were 3200 IP's that generates 1 or 2 requests per day (after normalize it is 3800 in the chart), while in October there were 5000 cases. The chart above shows that the number of small amount (1-2) and (3-5) of request per IP was increased, while in range (16-20) and especially in (21-25) it was significantly reduced. This proves that Synology actually did something with Package Center to decrease number of requests to third party servers.
In total there were much less requests to the repository, which is shown on the first chart above. I still don't know how Package Center works. But an interesting fact is that Synology was able to change the internal behavior of Package Center without releasing new version of DSM. It was done in background silently. I checked releases of DSM and there weren't any new on that period. What else they can do without user knowledge? I will leave that question without answer (because I don't know).