I don't want to come across as arrogant but you seem like very biased towards Afrihost. Making posting about how bad Afrihost is with packet loss in CISP thread here making well jokes that's not really funny.
Let's address you fundamentally flawed graph with packet loss. IMCP packets sent over a period of time is not and accurate measure of loss. IF one packet gets filtered along the way the will be measured as loss. I extensively make use of my 200/100 line on Afrihost for gaming, voice and then the usual downloads. If the loss you are punting all over mybb is happening this thread will be littered with complaints because voice and gaming would be impossible. A small packet lost while gaming will likely cause disconnections, very bad lag or high latency.
Afrihost uses TelkomIP Net for transit so once they hand it over to Telkom in JHB for transit the path is exactly the same being used as Telkom ISP, Telkom LTE and other business related broadband from Openserve. There is no complaints so that part is fine. Traffic from Rosebank IPC to NAP is still all Openserve wether you are with CISP or Afrihost. That basically leaves only a few routers between Openserve handoff at NAP to AFrihost or any other ISP depending on how they transit the traffic from Rosebank which if I understood PBCool correctly still had to happen with Openserve.
Afrihost is still one of the biggest ISPs in SA way bigger than CISP so very likely mistake is being and and will still be made that is why I always encourage people to post enough info. Just like the complaints in the CISP thread which ironically is WAY more than Afrihost most are related to loss on the cable providers side. Loss on international transit will affect so many people downdetector will even spike with complaints. Loss on Afrihost portion of the network will also spike a large number of complaints everywhere which is not happening.
I suggest you investigate where the loss come in and how it affects your graphs before you try to taint a ISP name. Just saying take it or leave it.
I just see loss or latency or don't see it. If I see it then it indicates a fault.
There is no flaw in using ICMP and no-one should be blocking or filtering it. ICMP packets send over a period of time are accurate reflections of loss. In this case no-one has confirmed that they are indeed filtering ICMP. Its an assumption, and if true is a bad practice.
I will agree that ICMP packet measure might not be a reflection of user experience as that is totally dependent on the thousands of apps and service out there and how they are being used. Some react differently and others have protocol mitigations to over loss. As an example, my afrihost account serves a purpose as a backup during day time where I have noticed a problem. The joke was made because I haven't had an answer from afrihost on the query and neither have any of the other guys who are reporting the same issue. I'm not saying the whole network is ****, just confirming the symptoms seen by the others. I did not initiate the post, just replying to people confirming I also observe it.
The ISP takes traffic from the Metro circuit terminating the IPC and takes responsibility for transit including peering at NAP and INX. Most ISPs use DFA circuits from the metro handover to the transit or peering partner, not Telkom as you state. My PPPoE account which has an afihost address of 169.0.243.154 AS37611 connects via an OpenServe fibre to a PPPOE server 41.170.72.193 ASN AS36937 which is liquid. I'll send a message to the contact for Liquid on AfriNIC confirming that it is their address. So I don't think everything is Telkom. There are parts that are liquid.
I'm happy to take up the offer to investigate. All I need is mib read access to all the networking kit and it will be two ticks and the problems will be identified. I'll even do it mahala.
PS: No need to be precious, since you have been looking and my message history you'll also see I have has my balls bust by tens of thousands of messages over at iBurst and have developed a thick skin.
Sign up with Google
