It was down a few mins ago, sorry about that.
Our team have got things sorted out
-
Join the MyBroadband community
South Africa’s biggest forum. Discuss, discover, and connect with thousands of members.
-
Question of the Day: If you were buying a TV today, which brand would you choose?
Afrihost Uncapped ADSL Feedback (Pt3)
- Thread starter bwana
- Start date
- Status
It was down a few mins ago, sorry about that.
Our team have got things sorted out
Looks like it, I am getting a "We're Offline" message + my office just called me to see if they are capped as their AF VDSL is down.
VPN is definitely not shaped on Uncapped, and looking at your post, this would also be at a time on a Saturday when traffic is extremely low (in terms of overall demand). I would expect most services to be performing reasonably at this time.
We do see some problems when torrents are running as QoS can affect realtime services. Do things change when you cancel torrents (and also close your torrent client)? Maybe we can look at whether QoS is not handling traffic as well as it should
The VPN thing I tried while running no torrents. VPN only got usable again the next day.
Torrent speeds have remained completely useless since early last week. I get spikes where my download speed will go to max for a minute or 2, but then it will drop down again.
At the start of the month or somewhere around there I also tried to get a torrent, and decided to speed it up with Turbocharge. That also didn't affect my download speed at all. Download speeds still wouldn't go above 30kB/s.
The ClientZone shouldn't have been down, did you have problem accessing it?
That's very odd, sorry about all this
Would you mind running a few tests this evening, jotting down the time and dropping me a PM? I'd like to check the changes in shaping from our end.
Thats really odd Afriman. Particularly when switching over to other ISPs (and before you say a quick switch isn't enough to establish a steady rate, I ran the connections all day, and when I switched back to my Afrihost connection it was STILL slow even after being disconnected), makes the problem disappear entirely. Are you saying Chrome has introduced a bug that only affects the Afrihost network? Truly that would be something special. And ultimately still your responsibility to fix traffic passing across your network.
That's very odd, sorry about all this
Would you mind running a few tests this evening, jotting down the time and dropping me a PM? I'd like to check the changes in shaping from our end.
Didn't really have time last night to monitor stuff all that much. I did note that last night at around 21:50 when I checked the Torrents were able to go at 250 - 310 kB/s at least until 22:00 when I went to sleep.
The morning the same 2 torrents are getting a combined speed of between 15 and 30 kB/s while no shaping is applied. (checked between 05:00 and 05:40)
As a side note this is with 1 torrent with 28 seeds and the other with 123 seeds.
Last edited:
We've logged this issue several times but the fact is we didn't make any changes to the network, it was something that changed on Chrome's side. From what I understand, they were experimenting with their traffic using a new experiencing UDP protocol. We don't shape or act on this traffic in any way. We are seeing the errors on both shaped and unshaped accounts.
I also notice that torrents perform differently at different times of the day. Remember that a lot of seeders are probably in the US, and we probably get better speeds from a torrent when they're asleep and nice enough to seed files with their PC's left on. Once they working or active again, many will limit upload to preserve their services or switch off their seeding, so your experience will be very different.
When the clientzone indicates the network is not being shaped, the problem goes away. When the network is shaped, the problem returns. The shaping only occurs when connected to the afrihost network while using Chrome, any other ISP experiences no problems. You cannot claim it is not happening Afriman, you've seen post after post about this. Nor is it reasonable to place the ball in Chrome's court when only your network is affected. Its starting to sound like the plaintive bleating of a child caught in the act. Simply "Not having made changes to the network" does not preclude a shift in the kind of traffic that is carried by it, nor does it absolve Afrihost from fixing the problem.
My reports are very different. If you pop over to the capped thread, you'll see that several clients are reporting this on unshaped products. So I'm not at all saying that it is not happening, what we are trying to understand is what is happening and where it is happening. I am definitely aware that this seems to be happening on our network and not necessarily others, but we've done extensive investigating to find rules that conflict with QUIC traffic and we haven't found any. I also don't mean to indicate that we're going to stop looking or trying to understand this issue, but right now it's not a simple turn of a key or switch. It may be related to how the traffic uses random protocols (as someone in the capped thread suggested) and that our network uses packet signature shaping rather than port/protocol shaping. If that's the case, it will definitely take more investigation to understand how that works and what we can implement to improve the response to this type of traffic. Until then, we can only recommend that you adjust your settings until we can find a more permanent solution for our clients.
I'd like to prove to you that it has nothing to do with the seeds. Unfortunately I'm at work right now so can't do anything till tonight. I can almost guarantee you that when I switch to WA that I will get my max line speed on those same torrents that are going nowhere on Afrihost.
From what I've seen so far this week, it doesn't matter if the torrent has 5 seeds or 3500+ seeds, at any given time of the day the download speeds will be pathetically slow. I only get speed on these torrents on Afrihost for very short bursts at any given time of the day.
This is after I have checked that I don't have packet loss at the time of testing, speed test gets good speeds, no other downloads are running, etc.
Anyway, I'll post screenshots here once I get home to show that its not the seeds.
I really don't think those tests are really useful, because what is the baseline assumption that we are operating on? My own tests show that it's not a 100% guaranteed. If all networks are running well, I've tested with every major ISP and the results change every time I switch accounts. Sometimes ours is better, sometimes others. I've even had cases where a shaped account outperformed an unshaped one. The nature of torrents is that there is no systematic way to compare, because each time you disconnect you are repaired. You might just be paired with someone with an incomplete file, or change from one who did to someone who has, or experience users with unlimited uploads set as opposed to limiting their upload.
I totally understand that torrents are a service that many clients use, and torrent speeds matter. But I've never supported using them as a reliable indicator of network performance and for me (unless you are comparing a shaped or throttled product to an unshaped product) comparing torrent performance isn't really indicative of much
I really don't think those tests are really useful, because what is the baseline assumption that we are operating on? My own tests show that it's not a 100% guaranteed. If all networks are running well, I've tested with every major ISP and the results change every time I switch accounts. Sometimes ours is better, sometimes others. I've even had cases where a shaped account outperformed an unshaped one. The nature of torrents is that there is no systematic way to compare, because each time you disconnect you are repaired. You might just be paired with someone with an incomplete file, or change from one who did to someone who has, or experience users with unlimited uploads set as opposed to limiting their upload.
I totally understand that torrents are a service that many clients use, and torrent speeds matter. But I've never supported using them as a reliable indicator of network performance and for me (unless you are comparing a shaped or throttled product to an unshaped product) comparing torrent performance isn't really indicative of much
So you want to try and convince me that for the past 2 weeks or so, on every single torrent that I've tried to download, I've gotten the seeds that have incomplete files or have bad upload speeds... Good luck with that one.
If for 3 years I've been able to get between 300 and 410 kB/s on my well seeded torrents when you aren't shaping, then I expect it to stay near that level. Having it drop to 25 - 30kB/s while you aren't shaping on EVERY SINGLE TORRENT is not normal. And it doesn't matter that I stop my torrents and try to start it again, or reset my router. By your logic I should then every time connect to some other seeder and finally manage to get my usual speeds, but this doesn't happen. The only time I see those speeds are for short couple of seconds/minutes bursts.
If anyone else is experiencing these performance drops, please chip in.
I can't speak for your entire experience, but I can certainly say that from my testing on both Capped and Uncapped torrents, we are capable of achieving almost line speeds on torrents, and even with demand high I've seen very high speeds on my downloads, and even better when demand is low. I'm happy to show you that evidence. My point here is that torrents have too many variables and are not a reliable indicator. Whether it's luck or related to the file you're after, I really could not say. But I am happy to post my torrent speeds to show that the network is able to deliver speeds in excess of those you've reported. At the end of the day, I don't want to deny your experience, which I can see is frustrating. But I think I'd need to steps which I can replicate in order to be able to take this further and to also negate my own experience, which is different.
omnom
Active Member
I've also had issues with Chrome over the Afrihost network on a capped account. The error I get is ERR_QUIC_PROTOCOL_ERROR. I disabled QUIC (chrome://flags/#enable-quic) and I haven't seen an issue since then. It's not clear if this is a bug in Chrome, as I tried a free WebAfrica account and had no problems -- this appears to be an issue with the Afrihost network specifically.
QUIC was discussed a stack over in the Capped thread, we can't find anything from our end on what can be caused this - but we're still looking
Did you read my post above yours?
- Status