Telekom Malaysia


TIME dot Com


Global Transit Carrier by TIME


Binariang Maxis


Celcom Axiata


DiGi via Telenor Global Services


U Mobile Sdn. Bhd.


YTL Communications - YES

Hurricane Electric BGP page mixed upstream and peering traffic in the single graph, therefore not actually accurate. Use BGPView where peering agreements and upstream transit are actually separated, with the correct weighting too according to the BGP announcement.

For example, TM upstream providers are listed here and their private peering arrangement listed here.

You are right. Hurricane Electric BGP is not always accurate as it use its internal mechanism to determine the provider peering and upstream.

Another alternative I recommended is Caida AS rank (TM example) which list providers, peers and customers for particular AS.

This post has been edited by michaelkkl: Feb 20 2021, 08:55 PM

It's suffice enough for the general public to understand and acquire a better picture of where each ISP's traffic are routing through.

I do use SecurityTrails for many other lookups, just that their maps are less intuitive and harder to figure out for most people out there. It doesn't summarize the data in a form of a nice pie chart like HE's BGP page does.

Also I think it's important that ISPs come out transparent to the public about their IP blocks announcements and peering arrangements with other networks. You can't hide important information like these from the public just as you do with coverage maps to prevent competition.

If not, unannounced blocks and traffic routes will trigger a major disaster and breakdown on the internet like a chain effect. Like the the wrong DNS resolving which happens with TM's certain range of IPs.

The public DNS resolver things the IP belongs to some other networks causing packets to route wrongly geographically.

BGP announcements, are by definition, public. BGPView's ASN weighted upstream list + upstream graph are far more accurate than Hurricane Electric's pie graph.

Maxis finally added GTT Communications which owns the Trans-Atlantic Hibernia Cable Express cable system to ease traffic into Western Europe and across the Atlantic to Eastern US/Canadian cities.

Hibernia Atlantic Submarine Cable System
https://www.submarinenetworks.com/systems/t...bernia-atlantic



Now being able to achieve as low as 236ms to eastern US cities such as Ashburn, Virginia:



This post has been edited by Candy12: Mar 3 2021, 07:01 PM

Level3 and GTT both are premium transit

so far celcom using now turn Maxis

Maxis again recently made changes to their upstream providers around last week.

Now they have added Tata Communications (AMERICA) and Hurricane Electric to their existing list as of 20/04/2021.



Noticed that latency to Amsterdam/Netherlands & UK improved tremendously with pings between 170ms-190ms but the changes also messed up their routes to other continental Western European countries which saw latency shot up as high as 100ms to countries such as Belgium, Luxembourg and Switzerland.

Here are a few examples:

M247 Ltd Belgium
speed.be.m247.ro


Broadcast Center Europe, Luxembourg
speedtest1.bce.lu


31173 Services AB Switzerland
zur-ix1-tptest1.31173.se


Please take note Maxis, your routing went bad again for Europe.

Appreciate the improvements made to routes connecting London(UK), Amsterdam(NL) & Vienna(AT), but the efforts were negated by the bad routes to Belgium(BE), Luxembourg(LU) and Switzerland(CH).

It seems these routes are now routed across the Pacific spanning America -> Paris -> Continental Europe/Benelux countries.

As you can see through the MTR results, Cogent route is the source of the bad routing.

HE / NTT Europe / GTT did an excellent job on their part reaching Europe via main IXs within latency 170-190ms range. So it should be around that range for latency connecting Belgium, France, Luxembourg and Switzerland from Malaysia. This was before Tata Communications/HE was added to the upstream peering list.

After they were added, we observed that the latency before this from 170-190ms spiked to almost 100ms more and now averaging close to 300ms between 260ms-290ms.

Do take note.

This post has been edited by Candy12: Apr 20 2021, 05:30 PM

it seem that maxis using NTT economy transit to EU base on your traceroute result.

speed.be.m247.ro via unifi ping was 200 to 210ms average under M247 peering at NL

speedtest1.bce.lu via unifi ping was 270ms to 280ms average under He.net HK >> He.net SG >> He.net Marseille >> He.net Frankfurt >> He.net LUX

zur-ix1-tptest1.31173.se via unif ping was 175ms average under DE-CIX Frankfurt peering

They obviously messed up the Benelux(French speaking Western EU) countries routing after they added Tata Comms and HE.

Before this 3 weeks ago, pings to M247 BE, Luxembourg BCE and 31173 Zurich was just under 185ms on average.

The funny situation is further destinations to the West side of EU such as London/Amsterdam now only below 190ms but nearer ones such as Continental Europe destinations such as Switzerland, Luxembourg and Belgium are experiencing latency close to 300ms!

I don't see Tata Comms servers anywhere involved in the MTR routes. Most normal links which gets into Europe are still routed through NTT and GTT with TI-Sparkle taking them through the Mediterranean/Italy via Turkey.

NTT must be involved in the bad routing too. Why are they routing packets destined for West Europe across the Pacific first through America then land in Paris/Amsterdam using Cogent? It was routing correctly weeks back direct to Frankfurt/Paris before reaching the Benelux countries destinations.

Did Maxis proof check their routes or even noticed it?

i just tried one of NTT SG server to speed.be.m247.ro it seem the ping 180ms to 185ms.

speedtest1.bce.lu via NTT SG 235ms to 245ms

zur-ix1-tptest1.31173.se via NTT SG 252ms to 258ms

for TATA is just peering in between not their preference IP upstream. TATA nowadays to EU also seem not so reliable like last time anymore more or less like NTT nowadays their route to EU.

GTT/Telia/Cogent still better to EU. Telekom italia i dont have much info but their premium transit always have the shortest path like GTT.

This post has been edited by heLL_bOy: Apr 20 2021, 09:05 PM

Not a fan of Tata too. Their routing is not the best and sometimes considered lousy. That's why TM was previously badly affected last time partly due to their peering with Tata and recently Google Fiber customers also made a lot of complaints with their connectivity to Europe. Google relies heavily on Tata uplinks as one of their major transit partners since they have major Indian shareholders now inside.

A better partner to replace Tata is Level 3 to complete the picture.

What you got by using NTT SG server is basically what I got 3 weeks ago before the addition of Tata Comms and HE to Maxis's list of upstreams or even better.

NTT SG for 250-260ms range to BCE, LU?
Maxis optimized GTT routes could get pings below 190ms to BCE, LU.

If I use Cloudflare WARP+ Singapore to connect to these 3 destinations even better can get under 180ms even closer to 170ms.

from what i see Maxis just having the peering with He.net and TATA and is not their UPstream provider.

there another factor you need add in that Server to your ISP route may also occur the routes/latency different which also happen sometimes.

NTT and Tata considered cheaper then the rest provider and in terms of route path and latency still losing out the rest major provider.

Level3 is quite expensive and seldom will be picked as their transit.

The current setup of Maxis's upstream links looks fine.

Tata Communications (America) actually helped them complete the links to Amsterdam(NL) and London(UK) which before this I've never gotten anywhere less than 210ms. Now I can get below 180ms/190ms to both London and Amsterdam is already considered good.

Maxis' weakest point now to Europe is not its upstream providers, it already has NTT, TI-Sparkle, GTT, & Tata to cover the entire region. What is LACKS most as observed and compared to other Malaysian ISPs is that it DOESN'T HAVE any peering partner ISPs in the BENELUX REGION.

If you take a look at:

TMnet has AS6661 POST Luxembourg

Time Global Transit Carrier has:
AS6661 POST Luxembourg
AS56665 Proximus Luxembourg S.A.

Even a small ISP such as Allo guess what? They have peering with:
AS6774 Belgacom International Carrier Services SA
AS56665 Proximus Luxembourg S.A

Maxis has NONE in the following Benelux countries and WORST it doesn't have good peering partners in Switzerland and even France(except ACORUS NETWORKS SAS AS35280) which is insignificant.

What's the use of having good Tier-1 upstream providers such as NTT, TI-Sparkle, GTT and now Tata when upon handover to the receiving EU ISP, it doesn't want to prioritize the reception of the packets originating from Maxis?

The 2 BIGGEST Issue Maxis has to solve now is:

I suggest:

For Belgium
Orange Belgium SA (AS47377), Belgacom International Carrier Services SA (AS6774), Interxion Belgium NV. (AS31651)

For Luxembourg
POST Luxembourg(AS6661), Proximus Luxembourg(AS56665), Datacenter Luxembourg(DataLux) S.A. (AS24611), root SA Luxembourg (AS5577)

For Switzerland
Swisscom AG aka IP-Plus (AS3303), Init7 (Switzerland) Ltd. (AS13030), AlpineDC SA (AS198385), Salt Mobile SA (AS15796), Sunrise Communications AG (AS6730)

For France
Orange S.A. (AS5511), AS12876 ONLINE S.A.S. aka Scaleway, OVH SAS (AS16276), Bouygues Telecom SA (AS5410)

For China
China Mobile International Limited (AS58453), China Telecom Next Generation Carrier Network-CN2 (AS4809), China Unicom Backbone (AS4837)

If you look at this MTR results, you will noticed that NTT Europe did a good job routing to Belgium at around 197ms but at handover to the destination's server, the ping shot up by almost 100ms extra to as high as >330ms because the receiving ISP has no peering agreement with Maxis and thus de-prioritizes the receiving packet.



This post has been edited by Candy12: Apr 20 2021, 11:58 PM

What were your results when you used NTT's own Looking Glass(from SG/MY) to do a trace to most European servers(Germany, Switzerland & UK)?
NTT Looking Glass
https://www.gin.ntt.net/looking-glass-landing/

They all took the Pacific - Atlantic path instead of ME route right?

Upon further check, I found that something's up with NTT. It affects not just Maxis but all ISPs who are now using them as their main upstream carrier.

They are now routing ALL their Europe bound traffic across the Pacific to America then via the Atlantic ocean to reach Europe. It seems they want to avoid the troublesome and risky ME route entirely which is a LOSS to Singapore, Malaysia and SEA ISPs which uses them connect to Europe.

Maxis to FDC-Frankfurt
lg.fra2-c.fdcservers.net


Maxis to FDC-Zurich
lg-zur.fdcservers.net


Maxis to FDC-London
lg.lon-c.fdcservers.net


It makes SENSE for countries in the Far East like Japan to take the Pacific->USA->Atlantic Route to reach Europe because they save on the 75-80ms return trip connecting to Singapore first then head West through the Egyptian Gulf before exiting through the Mediterranean sea entering South Italy/Toulon France into Europe Mainland.



For Japan the trip across the Pacific is just around 100-110ms and to London is about 230ms.
If we use Singapore via the ME route using the Egyptian Gulf we might experience faster latency somewhere between 170-180ms but not for the Japanese and Far East countries such as North China/South Korea/Taiwan, they have an additional distance to travel southwest bound before taking the same route so it's not quite an advantage to them compared to taking the Pacific-Atlantic route to reach Europe landing points.

I forsee that eventually Maxis will have to find another alternative upstream carrier to serve its Europe bound traffic needs if they want to take further advantage of the shorter distance/faster latency via ME.
That is why TI-Sparkle, Tata and GTT now comes into the picture.

For Allo it's going to be even worst, they rely ALMOST entirely on NTT for its single hop upstream needs. It's not going to be good for their Europe traffic.

NTT has never been a major player in the EMEA submarine route due to stiff competition from rivals.
They are a major force in East Asia and Trans Pacific routes though to US which makes them formidable in those markets.

Maxis already got premium routing to CHINA. they wont pay for high price for china routing via china telco.

for the rest i guess is less user using this region like LUX/BEL/SWISS

maybe they can improve over France location using their current GTT or Telecom Italia Transit.

NTT Asia to NTT EU doesn't have good latency unless your ISP pay for premium price for the transit.

NTT Asia are more focus on Asia and US region.

I just take a look on FDCserver. they have NTT as their transit but not sure why their routing path coming back via US to JP then your ISP.

even using NTT premium transit the ping would be less 15ms to 20ms. not sure is there something break on the cable again or what happening and usually the traffic would come back via EU

Nothing to do with premium links.

You can use NTT's own looking glass and do a traceroute to those European servers, the routes are EXACTLY the same as what Maxis and its other downstream ISP goes through as well minus the first few hops among their own internet routes.

I compared the routes between Maxis's and NTT's own premium carrier routes using:
NTT Looking Glass
https://www.gin.ntt.net/looking-glass-landing/

They're basically taking the same routes to Europe.
SG->Japan->Los Angeles(LAX)->New York(JFK)->Paris->Europe Destination

Valve Europe Gaming server is based in Luxembourg.
Belgium and Switzerland hosts many of Europe's higher learning institutions and research facilities.They're equally important.If you take a look at some of our local university ASNs, they have peering with Belgium for this reason. Monash Malaysia is one example. These Benelux countries are Europe's most centralized countries just like Singapore is to SEA region.

For China, I think it's best to stick with China Telecom's Next Gen Carrier(CN2)/Unicom International or China Mobile for access through their Great Firewall contents.

ISP will look into the pool to see how much user using this region before they adding in peering/transit. Maybe in future maxis will add LUX/BEL/SWISS region into their list.

most ISP won't pay big price for CHINA route. per 1Mbps cost $50-100USD economy class route/premium class route $150-300USD.

the only way is using third party IP transit carrier like NTT/Telstra/PCCW becos price was lower then china telco itself.

FDCservers Asian locations(SG, HK & JP) are all using NTT as their transit carrier(ISP).
Using FDC's Looking Glass for (Asian servers) is almost similar to using NTT's own looking glass.

If you use either of them to ping to Europe destinations they're all NO LONGER routing to Europe via ME anymore.
They will ROUTE across Pacific to US then the Atlantic to reach EU the other way around.

For Far East countries like Japan, South Korea/North China the difference between Pacific+Atlantic option vs SG-ME option is almost negligible.

Rough estimated calculation:

Option 1: Pacific+Atlantic Route to EU
JP->US West->US East->Paris

Average Ping: ~235ms

Option2: JP->SG->ME Route to EU

JP->SG
(1) Average Ping: ~70ms

SG->Paris
(2) Average Ping: ~160ms

70ms + 160ms = ~230ms

Given the security and safety concerns from potential terrorism and extortion threats, give or take it's better to use the Pacific+Atlantic route in view of the Japanese evaluation for their part.

China & Russia too has their way to evade and bypassing the ME. They have a Trans-Eurasia route which goes through Kazahstan-Moscow which emerges out from St Petersburg-Baltic Sea.





We're back to the proxy extortion/protection trade wars involving Aviation and Shipping industries and now it's been extended to submarine cabling.

This post has been edited by Candy12: Apr 21 2021, 07:10 PM

base on my track previously NTT to EU normally is via Amsterdam when SEA-ME-WE 5 was not down during last year till Mid-Feb which the cable cut and during this period most of traffic was via Milan. When the restoration was done half of the traffic shifted back to Amsterdam.

Not sure now why now the traffic shifted to Vienna then only Amsterdam which also cause additional latency.

The normal usual route which Maxis took to Europe last time was via TI's SEABONE route via Southern Italy into European destinations.

I'm starting to see that Maxis has now started utilizing their TATA routes (AS6453) to reach Belgium networks such as Proximus but they're not being prioritized thus the spike in latency during handover.

The packets already arrived in Belgium via Singapore using Tata's route in just about 171ms which is considered very good but upon handover to Proximus's network the packets experienced some 100ms delay due to not being prioritized as a peering partner of Maxis.



Maxis has very little peering arrangements in European region particularly Benelux countries.

Hello Guys, sorry for interrupt your conversation. I have enquiry, currently I'm using Unifi Fibre 100mbps, I seeing Celcom Fibre offer 300mbps (competitive price)

Based on you guys expertise on ISP international routing, its that safe to take Celcom Fibre now? I have concern on latency and worry apps, gaming, etc my user might experincing delay.

Hope you guys dont mind to share. Thank you.

So much has been said about Maxis. How about Celcom and Digi?

Allo's, if anyone is lookihng for it

Digi/Umobile/Redtone is leasing majority ingress traffic from TM.

no different at all if you using TM.

only celcom / maxis is far different from them since they have their own transit provider and not leasing from TM at all.

Ok noted.

What about TIME compared to unifi local and international?

TIME still ok.. bad point if your source sending the ingress traffic to Singtel then back to TIME from Europe/US might having higher latency then the rest of upstream they have.

local as usual using MYIX or direct peering.

This post has been edited by heLL_bOy: Nov 20 2021, 04:26 PM

Thats great. At least better than unifi

any more updates on Upstream Peering Providers?