Hush hush ni dude, later TM going to come for you xD 

i am waiting for them also. but so far all ip range in my area i been tested all is working 

Must VVIP. Location?

Yeah opendns servers and oracle-sg is pretty much cooked for today.

Target Name: resolver1.opendns.com
        IP: 208.67.222.222
  Date/Time: 9/16/2024 1:48:45 PM - 9/16/2024 1:58:45 PM

Hop  Sent  PL%    Min    Max    Avg  Host Name / [IP]
  1  103    0    0.20    0.58    0.39  192.168.88.1 [192.168.88.1]
  2  103    0    1.05    2.82    1.60  118.101.155.254 [118.101.155.254]
  3  103    5  12.28  19.48  13.59  10.55.51.35 [10.55.51.35]
  4    99  100      0      0      0  [-]
  5    99  100      0      0      0  [-]
  6  102  17  676.56  827.30  750.58  36692.sgw.equinix.com [27.111.228.90]
  7  101  10  669.77  819.14  749.92  resolver1.opendns.com [208.67.222.222]
        IP: 168.138.179.163
  Date/Time: 9/16/2024 1:49:21 PM - 9/16/2024 1:59:21 PM

Hop  Sent  PL%    Min    Max    Avg  Host Name / [IP]
  1    27    0    0.29    0.70    0.42  192.168.88.1 [192.168.88.1]
  2    27    0    1.06    2.37    1.58  118.101.155.254 [118.101.155.254]
  3    27    4    9.72  15.09  11.22  10.55.51.35 [10.55.51.35]
  4    27    0    8.73  13.88    9.50  10.55.52.31 [10.55.52.31]
  5    27    0    5.59  20.32    6.83  10.55.39.147 [10.55.39.147]
  6    26    0    8.87  12.51    9.64  10.55.100.52 [10.55.100.52]
  7    7  86  77.94  77.94  77.94  port-channel21.core2.sin1.he.net [184.105.80.45]
  8    25  12  716.77  814.78  769.47  31898.sgw.equinix.com [27.111.229.44]
  9    24    4  707.89  830.12  764.03  140.91.232.19 [140.91.232.19]

here you go

Tbh not only SG but even to discord unifi kinda bad

How many times now since last 4 years? lol I lost count already.

The thing is how you gonna relay the routing issue to TM so that the relevant person get notified?

I'm not saying you are wrong etc. I'm just pondering how we as users can tell the CSO that this is an issue worthy of additional attention in a way he/she understands too.

I have been observing this TM route via HE for over half a year now, its always been 77ms, and suddenly starting from yesterday, it went up to 700ms++.
Who should we call to fix this problem, is it HE or TM?

EDIT : Might be calling HE would be the better choice?
From oracle datacenter back to Unifi, it seems to first go in to core3.sin.he.net, but it went to tokyo I assume, this pathway is controlled by HE?
From my oracle server (168.138.179.163)

                            My traceroute  [v0.95]
instance-20231106-1329 (10.0.0.181) -> unifi.xxx.xyz 2024-09-16T11:22:30+0000
Keys:  Help  Display mode  Restart statistics  Order of fields  quit
                                      Packets              Pings
Host                                Loss%  Snt  Last  Avg  Best  Wrst StDev
1. 140.91.232.44                    0.0%    24    0.2  0.2  0.2  0.6  0.1
2. 31898.sgw.equinix.com            0.0%    24    4.1  5.8  0.9  33.4  8.8
3. (waiting for reply)
4. port-channel1.core3.sin1.he.net  95.7%    24  18.7  18.7  18.7  18.7  0.0
5. port-channel1.core3.tyo1.he.net  95.7%    24  68.5  68.5  68.5  68.5  0.0
6. telekom-malaysia-inc.lg3.switch2 17.4%    24  635.5 647.7 475.2 767.0  74.9
7. (waiting for reply)
8. 118.101.153.xxx                  17.4%    23  624.7 657.7 545.1 744.7  60.4

Does local congestion extends past into the wee hours? Or perhaps its a routing error as suggested by heLL_bOy ?

Starhub don't have a Looking Glass, Equinix do.
I put in your IP address and here is the reverse path:


I don't know what is happening either. Initially I suspect they try to load balance users across all border gateway, but in your case, you have confirmed problem with Starhub, Equinix but not SingTel.
I tried various website I know and I am trouble free. If the hypothesis is TM trying to load balance user, I should have hit one border gateway that's congested by now.

I still kind of believe it is local congestion. Here is why I say this:
In MPLS core, there are a lot of tunnels between you and border gateway. Which tunnel you take is invisible to you. The underlay network equipment is also invisible and they don't show up in traceroute. Based on my previous testing, a lot of these links cannot hit 1 Gbps, while led me to believe they are installed during HSBB rollout and never upgraded.
Now speedtest get around this by doing multiple connections to multiple different servers. This effectively randomize the path selection and your chance of using different MPLS tunnel simultaneously is high.
However, in typical usage scenario, you connect to one endpoint and use only one MPLS tunnel at a time. In peak usage like during holiday, it is very possible all the 1 Gbps links are saturated.

Maybe you have some other hypothesis?

don't they need to pay for the terabytes traversing each foreign pipes?
no urgency to reduce charges?

Just speculation, don't quote me on this

MyIX is located in AIMS, AIMS largest shareholder is TIME

Now guess who doesn't like to work with TIME

Here is the thing. I do not think they have enough bandwidth to begin with.

Here's why:
If they have enough bandwidth, they don't need to do the TM-HE-Equinix, TM-TATA-Equinix stunt. They could have just stick to TM-Equinix and still milk u-mobile and whoever for their IP Transit charges.

Or... Corruption is in play.

ahha private ip confirmed will get slow and lag
you need public ip address get load faster

mine is public ip address 175.xxx.xxx that why i got faster lowyat forum and youtube and etc

Yes I get it TM problem.
But I don't believe using HE-Equinix is cheaper when one is billed by port and another is billed by 95th percentile.

For comparison, TM is having 300Gbps to Equinix right now. How much traffic are they egressing to HE?

I get it the Capex can be cheaper for other player because they peer in MyIX. They do not need to run fiber all the way to Singapore and pay SGD for port fees.

Granted HE peers in MyIX and Cyberjaya too.

It depends if you want to calculate the infrastructure cost to Singapore as well, or just calculate the port fees.

However none of these matter because we have MyIX. There is no way to justify using HE and TATA for Equinix when MyIX solves the same problem.

I am starting to lean on some people in TM receiving extra money for doing it. There's is no technical explanation that can convince me otherwise.

If Equinix is so expensive, why build it?
Since they have already built it, might as well just upgrade the equipment on both ends and make it 800Gbps running on the same fiber.

But nope, let's pay upstream.

And for others don't care about privacy issues, get yourself a WARP+ or ZeroTrust for free.
(unsure how to setup ZeroTrust, as I saw reddit mention about ZeroTrust "MIGHT" get high priority as WARP+, so just sharing it out for others to test about it la, me lazy person. ;P
Should give better privacy, because it target business customer, still, if you have time to set it up.)

So far TM do nothing to the certain IP range that based on WARP, it is still routing to KUL, so you can going via the routing of CloudFlare and enjoy full speed for free.
(And nope, tested, it is dedicated for WARP only, not allowed to use as :80 or :443)

This might shared your phone number out by telegram, so be aware if you gonna generating key for yourself, so far I generated once, no stranger sharing the same key/account with me, and yes, faster than regular WARP, full speed for oversea slow long comic site.
https://t.me/s/warpplus
https://t.me/generatewarpplusbot
---EDIT
Yes, no privacy.


https://www.reddit.com/r/selfhosted/comment...omment/kuz1la6/

If you notice he keep repeat same comment over and over
As i don host game, cctv, or any storage on home that need acces outside
I don see any big different speed or respond, torrent for good seed also can max speed 38, 39MB/s

That 300Gbps is bi-directional. You mean using only the egress direction but not the ingress direction saves money? Care to explain how it works? Afterall TM is eyeball network, they are not Amazon who egress more than ingress.

True MyIX don't have as many tenant as Equinix but they have Amazon, Meta, Google, Cloudflare, Akamai, Microsoft, ByteDance, Tencent, Alibaba, Huawei, Netflix.

That's some significant saving of cost just from the above list.

But not only TM uses IP Transit for those traffic, they also route them around the world.