Lol, so we are same boat, 30M, free upgrade 100M, then slowed down today midnight.
They really have a crappy QoS, I guess it is like this.

1. Multi-user, open port for connection take long time, old issues since Streamyx
2. Unlike Streamyx QoS implement based on public IP, not private IP, so you are sharing the QoS with others. like 10M only for CloudFlare HK, then dropped to whatever users are using, like 30 users then 10M / 30.

Else, unable to explain why public IPv6 are affected.

I recommend you to make a report, more report related to private IP, much quicker on fixing the issues.

This post has been edited by BenYeeHua: Dec 17 2021, 11:01 AM

Can be, I am doing testing to see how muck going to Equinix now, but most of the thing are broken, Lazada, Taobao, Tencent, CloudFlare HK.

It is not funny when the line broken at the same time you are getting private IP switch on, failed to know it is the private IP or the global exchange, even my UniFi are self test and UniFi Portal are broken, live chat with CS jitter in between 0.1s - 2s...
Are they not hosted on Malaysia?

This post has been edited by BenYeeHua: Dec 17 2021, 12:06 PM

Strange thing is, after reset port, it is fine for 1 min, 10M full QoS speed like normal, then dropped, and yes it is Equinix SG.
So it should be UniFi detect heavy drop on Equinix, and implemented heavy QoS to stop it getting worst.

----
For UniFi portal and self-test, it is their side problem, 500 lol...



And funny enough, it start together midnight, when the switch to Private IP are turned on.
If the switch is not happen today, I sure will confirm it is global issues, but now?

Still testing on local CDN to ensure it is really QoS issues, not something caused by Private IP switch over la..
After the switch on, my QQ are having connection drop, and teamviewer also dropped a few times, so it is very strange that it happen together.

Also they closed my report for abnormal speed, so they found nothing on their side.

This post has been edited by BenYeeHua: Dec 17 2021, 12:18 PM

The strange part is, it happen when Private IP switched on today midnight, coincidence or ???
---
Here is affected thing.
Equinix - affected
Capital Online(tested on SG only) - affected

Lazada(SG Aliyun, exchange are blackhole, so unknown) - affected

Tencent(via Equinix) - affected
Tencent(directly peer exchange to TM peer) - affected
Taobao(Equinix) - affected
CloudFlare(SG or HK, via Equinix) - affected

PCCW(HK) - Unaffected, first time

Found something interesting, before that CloudFlare is HK, then redirect back to SG, but both are Equinix, so it do nothing at all.

---
Also good news, Tencent are having Malaysia server now, 128.1.160.x, from SG server becoming Malaysia server, no need to worry about Tencent product unable to use now, including WeChat(unsure for China Wechat which is 微信's account)
Thanks, Wechat Malaysia user, you bring Tencent server to Malaysia.

All Tencent account login will be local, just group picture will be PCCW HK and gchat.qpic.cn(DNS depend).


I open report, already received SMS inform me it is resolved, yes, resolved.

This post has been edited by BenYeeHua: Dec 17 2021, 12:37 PM

Yes, one more testing from me.

Tencent - China Mobile(via Equinix SG) - affected

So it is just unlucky that today they switch me over to private IP, while Equinix burst on same time, so I can't know who is causing trouble...
----
Ya, last year only used SG server, after Wechat becoming WeChat Global and WeChat HK with different law applied, then they can safety to implement the law on the Malaysia server, and enable MY server to Malaysia users.

But still, if need tech support, they will move it back to SG and HK server for processing and accessing.

So, one of the problem of CDN is law applied on data, a few China friends I know are switching over WeChat global to prevent tracking from China company, they are having heavy big data issues like Facebook and Google, also iOS callkit for native call la.

Still unsure how's it works on QQ account, I am still signed to China account agreement, unlike WeChat global, but they allowed using MY server for that.
For Xiaomi and Huawei, still using HK server for my China registered account, each time login will need to redirect me back to China, before success on login.

---
So, how much traffic can be handled by CloudFlare Malaysia?
Why lot of servers just using SG CDN, not local CDN?

I still wonder why local CDN are not used at all, like CloudFlare or Aliyun(Lazada) or Tencent(Shopee), both at SG via Equinix, last time I tested on some global games that using CloudFlare for the api, when it is 5xx, it showing Johor CloudFlare to me, which is strange...

ok, cloudflare SG back to normal, they are exchanged via Telstra SG now, a bit slow, but 700kb/s.

Equinix - still affected.

This post has been edited by BenYeeHua: Dec 17 2021, 01:11 PM

But still strange, as I know the traffic issues is because too many request for SG server, the CDN is designed to avoid this issues by having data cached on local CDN.
That's how it works in China, and yes it works for 90% users, some still having traffic transfer issues in between CDN and main server(the real backbone), based on their live stream many users complain about lag but solved when switch the video quality.

So if the data are cached on Malaysia/local CDN, it should greatly reducing the traffic required going to SG/oversea traffic.

It will be
1. Traffic to SG moved to Malaysia local
2. CDN request from Malaysia to SG server will be reduced, as it is cached by local CDN.

But it is becoming
1. Using MY CDN caused issues to Malaysia users.
2. But, moving CDN back to SG like Lazada and Shopee solved the issues
3. And increased chance of line break in between MY and SG.

Why???
Malaysia CDN total bandwidth is lower than SG link because of expensive money need to be paid to TM?
or
Not enough hard disk/RAM for CDN caching in Malaysia?

Yes, most request to SG and HK are going via Equinix SG, it is because in past PCCW break too, so they have moved to Equinix for balancing the traffic in between Equinix and PCCW.

But, they are putting too much on Equinix now, or having DDOS or Physical connection issues to Equinix like line broken, so it is affecting.

If your account are safe or nothing, or the stuff you are download don't need account, then try using some free VPN to redirect the traffic to solve the issues for temporary la.

I wake up in 8 AM and found out this issues, and it is not gonna fix any sooner...
The worst is during last year WFH, it last like half of years...
(should be few months la, just the feeling)

So the answer is nobody know, only TM internal staff know.

If it is legit in SG law, find someone(that you can transfer normally without packet loss) hosted a private VPN for you to use, then by using his network as proxy, you can download full speed from SG server.

Else, hmm, ask someone in your company, did using public VPN is allowed.
If yes, then find a good VPN, trial first, if fine, then pay for a month first for getting a better traffic temporary la, they should not record down your files, except you are downloading via HTTP, in this case everyone can get it...

The best will be your company's ISP are not having issues to TM, so they open VPN server for you and you connect to their VPN, problem solved.

---
CloudFlare - telstra are unable to handle huge traffic now, and start having heavy packet drop, ouch!

This post has been edited by BenYeeHua: Dec 17 2021, 02:09 PM

If you are having full speed/acceptable(like 30M) single thread download from your VPS now, yes, only some SG exchange are broken.

This post has been edited by BenYeeHua: Dec 17 2021, 02:12 PM

That's why, remote work not suitable on Malaysia, too much traffic issues, even github can't full speed download...
---
Just for CloudFlare, if you are not working on an unrooted smartphone(excluding those with pi-hole custom localhost DNS), redirect them to Malaysia CloudFlare server using hosts file.
For expert, just redirect based on hostname of the server IP address with firewall etc la.

For the IP address, tips, Lowyat.net.

This post has been edited by BenYeeHua: Dec 17 2021, 02:25 PM

Are we having the different UniFi?

My result for cloudflare SG via Telstraglobal is 100kb/s via IPv4, 25kb/s via IPv6.

Here is the traceroute.


Ya I know, written as HK Telstraglobal, but it is at SG, because lowest ping you can get from HK are 45 ms, not 26ms.

Anyways, another method is rewritten all the CloudFlare IP address provided by DNS to local CloudFlare, as not much people are doing it, it will not burst the CloudFlare MY, so, find a method and do it yourself la.

So, you means for best experience, it should not be be.

UniFi user - 100G - CloudFlare/Lazada/Shopee MY(cached) - 10G(for cache miss data and API call, lower for Lazada/Shopee) - CloudFlare SG

It should be
UniFi user - 100G - CloudFlare/Lazada/Shopee SG

Which is causing the result of overloading the SG connection now???
That's the part I don't understand, they can just keep pulling new cable to SG, but they can't pulling cable to local CDN.

Like before UniFi show up, TM just force throttle any streamyx user connected to Google MY ISP at low as 480kbps, while anyone that using TM peer as CDN server will be full speed connection.

Then when Lazada and Shopee show up, TM just force throttle all UniFi user connected to Aliyun MY and Shopee MY, all users experienced bad connection to local CDN, which causing them switch back to CDN SG as the solution.

Which is...
What???


And now, great, TM just say, all CDN you are connected are outside of country, we don't earn money from it, so, we can't do anything on it, yay~
---
So yes, it is TM fault for not giving cheap traffic to earn money from both side, but let SG ISP to earn those money from CDN and they willing to pay exchange the money, yay...

I will never understand how the brain of TM works now...

So, the next one that want to open new data center at MY.
Run Microsoft!
Run!!!

Yes, I had make switch to JP DNS, okay for bilibili now, thanks, that lovely JP DNS.
For other like Lazada, nah, forced to SG one...

Like I said before, TM want to earn money from both side.

Level too low, they also using 30M with private IP too, based on the one that called me.

And yes, they say it is fine on their side, did they just stick to FB and YouTube???...

Seek for website that using CloudFlare as their free image hosting, then using host file, exchange the ip address to this.
(PS: Because that one I browsing is not publicly available website, so I can only give the IP address... )

2606:4700:e4::ac40:ad07
This one is rerouted by TM one, currently is SG.

Then, well, just ctrl+F5 to ignore browser caching and see your download speed of your browser from CloudFlare SG.

The best will be 10Mbps at full QoS speed, the worst will be slow down of opening of website.
---
Based on one user on Sarawak, yes, he are not affected and Twitter are fine for him, but not for Sabah user, strangely..

This post has been edited by BenYeeHua: Dec 17 2021, 11:15 PM

Facebook fine for me, double check your latency, because Facebook are using CDN at MY.
If 5ms no packet loss, then yes, normal.

Yes, some are affected, I unsure why.

Like when the CS reset port for me, I tried, it is 10M full speed, then QoS like kicked in and lost control, then heavy packet drop until 30kb/s only...
Unsure it is intended for private IP or whatever...

So you are UniFi with Public IP or?
If yes, I will said Private IP have lowest QoS than Public IP one.

This post has been edited by BenYeeHua: Dec 17 2021, 11:50 PM

Great, me is

Twitter
CloudFlare SG/HK
Shopee
Lazada
Taobao
Bilibili SG

But, only Facebook, CloudFlare MY, Tencent MY/HK, Bilibili HK(HKIX)/JP are fine.

This post has been edited by BenYeeHua: Dec 17 2021, 11:59 PM

Yes, then this crap happen same time, I don't think it is coincidence, because both happen same time...
---
Asked someone on Taobao, say notify nothing on network, interesting now...

This post has been edited by BenYeeHua: Dec 18 2021, 12:07 AM

Yes, they have no clue, I also don't have clue, like Taobao is going to HK, why my MY friend are fine, but me are lagging like in hell...
Why someone FB unable to open, me is fine
Why twitter not okay for 2 person, but ok for 1 person in this thread.

So I can said that, QoS is messed up heavy when they switched people to private IP + issues on equinix, causing different packet drop to different account...

I will make a very small bet on it, when they moved user to Private IP, they lost QoS control on limiting those heavy P2P users, then this happen...?

I think QoS lost control when they moving heavy p2p user into private ip pool, so it is affecting normal user as well...

So they are packet dropping normal user, not the p2p users, or both, so all kena unusual heavy QoS on this...
This will match with the port reset, me fine for 1 mins, then heavy QoS or packet drop kicked in.

This post has been edited by BenYeeHua: Dec 18 2021, 12:32 AM

QoS unable to implement based on users account or gateway?
As I know, gateway are different for users.