Peering Wars

This was actually a quite old story issued by Renesys several months ago and it was also presented at NANOG 43 (June 2008). The story took a case of network-depeering and the actors were Cogent and Telia. At that time, both Cogent and Telia are considered as "near" Tier-1 providers. Why "near"?

• Cogent gets (got?) transit from NTT (formerly Verio) to reach AOL (ATDN?).
• Telia appeared to get transit from Verizon to reach certain networks.

Additionally, it is worth to mention that

• Tier-1 providers act like a cartel and have no incentive to add members
• "Near" Tier-1 providers can try to buy their way into the club (via "paid peering")

Renesys said that Cogent and Telia want to be "real" Tier-1, and from this point, the story began:

• On 27/Feb/2008 Renesys stopped seeing evidence of Telia's transit routes and Renesys promotes Telia to "Tier-1" (However, it could still be paying for some of these interconnections)
• On 13/03/2008 Cogent depeers Telia, claiming breach of contract; Renesys observes routes from Telia to Cogent via Verizon for 12 hours, these routes then disappear, partitioning the Internet for some single homed customers behind Cogent and Telia (they could not reach one another)
• On 28/03/2008, peering link is restored and the Internet is once again whole!

Renesys's Analysis
The following statements are taken (almost) entirely from Renesys.

Why did this happen?

• The peering disputes (Telia-Cogent) tend to be about peering ratios; i.e. one party is carrying the other's traffic longer distances
• Did Cogent view Telia as in a weaker position? (How many European customers want to reach Cogent hosted content?)

Facts after peering link has been restored:

• Number of routes advertised to Telia is increased; Telia reaches 2934 more prefixes via Cogent
• Number of routes advertised to Cogent is decreased; Cogent reaches 635 fewer prefixes via Telia

"This was the result of reengineering by both parties. It surely seems like this was about traffic ratios"

Lessons learned

• Being a (near) Tier-1 is not easy; you depend on everyone else in the cartel and you will be punished if you are perceived to be in a position of weakness
• Peering relationships are tricky; depend on both objective measures and perceptions; disputes can take a long time to resolve; the only driver is market pressure
• Being single-homed is dangerous!

Other notes

• An increasing number of businesses and organizations are now multi-homed
• Many more peering relationships, lessening the impact of any one loss
• True global Tier-1s would probably never consider depeering; impact would be wide ranging; it would invite government involvement
• Tier-1s are becoming less relevant, as smaller players peer around them (This shall be the reasons, why Tier-1s have always to defend their positions with expanding their coverage and acquire new customers !!)

Credits: Renesys Corp. (Martin A. Brown; Alin Popescu; Earl Zmijewski)

Telkom's Backbone

Several times ago, I mentioned Telkom's national backbone in this post. According to records published by KOMINFO, as of Q4/2005 Telkom owns 10,816 km fiber optic cable. The backbone network is distributed as follows:

• Surabaya-Banjarmasin-Makassar : 2,088 km (constructed in 2000)
• Trans-Java : 2,605 km (2000)
• Jakarta-Medan-Batam: 2,990 km (2004)
• Batam-Singapore-Bangkok: 1,043 km (2004)
• Dumai-Melaka: 161 km (2005)
• Banjarmasin-Balikpapan : 955 km (2005)
• Makassar-Palu: 974 km (2005)

This information can be compared with that I have mentioned in other posts. This and those information can be complement to each other. For instance, for "Batam-Singapore-Bangkok" entry mentioned above, I would guess that it refers to TIS cable, which shall actually be excluded in the calculation since it is a consortium cable.

Palapa Ring (6)

Below are some updates for the Palapa Ring project as of September 2008. It seems that there are some route adjustments; and that now the project schedule is more realistic, taking into account current market situations.

• 2 FP for repeatered system
• 12 FP for inland and repeaterless system
• Construction begin on April 2009;
  
• RFS final (all segments) on September 2011
• Section 1 Mataram - Parigi
  
• Section 2 Palu-Sorong
  
• Section 3 Makassar-Sorong
  
• Section 4 Makassar-Mataram/Kupang
  
• Section 5 Sorong-Jayapura
  
• Section 6 Fakfak-Merauke
  

Palapa Ring Map & Status as of Q3M/2008 (Source/Courtesy: KOMINFO)

Jakarta (IP) Bandwidth Price

I followed some traces left by Pak Meneer and was delivered to his records at Opisboy. This is claimed to be an "Informal ISP Technical ( dir-pji-like :p )" forum. Comments ? :-D Well, for non-local readers, "dir-pji" could be understood as "indonesian ISP Executives/VIPs" or as the forum of ISPs' owners/high-rank officials. It is "dir-pji-like" ? Personally, I would say "better-than-dir-pji" :-D Reasons? he..he..he. Don't take it seriously or even start thinking about that!! :-D

Back to bandwidth. According to Pak Meneer, the condition is as follows:

• Jakarta - Singapore connection costs around USD 150 th./month (STM-1 closed circuit, clear channel); or
• Jakarta - Singapore connection costs USD 1.5 mn/10 years (STM-1 closed circuit, clear channel, IRU scheme)
• IP transit in SG costs around USD 60 - 100 /Mbps/month

From this numbers, he estimates that the cost of (budget necessary for) IP transit at Jakarta would be approx. USD 1,068/Mbps/month for normal scheme, or approx. USD 181/Mbps/month for IRU scheme.

This is more-or-less justified by several recent media reports (quoted information given by Indosat/Fadzri Santosa, APJII/Sylvia or others) which give the figure of around USD 1,000 /Mbps/month. Sylvia and Fadzri might talk on different contexts e.g. Fadzri mentioned closed-circuit leasing bandwidth and Sylvia talked about IP transit. But nevertheless, if we look at the numbers given by Meneer Tutut, cost component for transit doesn't contribute significantly. If Indosat plays in this price level, then I guess, Matrix and Moratel/XL would offer a bit cheaper price.

Regarding IRU scheme, I know that Matrix offers this scheme. But I doubt that Indosat does (will do)?? So, the information above may specific to Matrix. Right pak ?? :-D Some friends of mine have told me that Matrix has done a very good "marketing move". They may soon increase their currently-almost-full-booked 60 Gbps capacity. :-D However this might a spoiler, not official!

Last issue, many people use USD/Mbps metric for bandwidth price. I would personally prefer USD/X-port metric e.g. USD 2,000/E1-port rather than USD 1,000/Mbps. Why? Simply, because not all line bandwidth are useable for user data. If we are talking about STM-1 line for instance, if it is channelized, its useable bandwidth is only 126 Mbps. If it is unchannelized, by using PoS for instance, its useable bandwidth might be in the range of 130 - 140+ Mbps.

That's all. Sorry this time without pics. :-D

Jakarta Metro Backbone

It seems that everybody in Jakarta want to have his own metro :-) Currently many companies have built their metro infrastructure to support their business, particularly in big cities such as the Greater Jakarta, Surabaya, Medan, Bandung etc. These include :

• well known telcos (TELKOM/TSEL, ISAT, XL, other?)
• successfull ISPs (BizNET,CBN, other?)
• VSAT network operators (CSM, Lintasarta, other ?); and
• backbone operators (ICON+,PowerTel,Napinfo,Moratel, other ?)

Recently, PT. iForte Solusi Infotek announced, that they have finished the construction of their fiber optic network which is using (and installed simultaneously with the development of) the TransJakarta busway infrastructure (map see below). They currently offer connection/access services called M-Wifo which are basically WiFi (802.11n compatible) access and fiber bridging. To my guess, it might be a RoF (Radio over Fiber) system.

Other related public information about iForte:

• the fiber optic backbone spans/follows 10 TransJakarta corridors
• CAPEX : USD 10 mn.
• Length : approx. 250 km
• RFS : June 2008
• Services : fiber optic connections, WiFi mesh 802.11n connections
• Projection : +/- 2,000 cust. (connection point) each year
• VIPs: Peter Djatmiko, Andrie Tjioe
  

Map of TransJakarta Busway, Courtesy of Maximilian Doerrbecker (Chumwa), taken from Wikipedia

Bandwidth "Under Water"

As Asia NetCom (currently PACNET) introduced its idea for EAC Pacific one or two years ago, they underlined the fact that due to underserved routes from South and South-East Asia, Japan-US connections have become a bottleneck. They have shown a picture similar to one below. It has attracted my attention since it split Asia into three regions together with their connectivity parameters. At least, it was something different than we might see in the free version of TeleGeography's reports/databases :-D CMIIW!

After a closer look at the given bandwidth figures and comparing with other publications (TeleGeography etc.), I have to conclude that it isn't real i.e. ANC possibly gave the capacity based on design numbers (fully equiped system). Right ?? Wrong ??

What is actually in the water ? The second pic gives us a hint for trans-pacific segment based on information provided by REACH.

How Much Bandwidth do Indonesian ISPs Have ?

Though there are many Internet and NAP service providers running their business since several years ago, "big Internet pipes" are only owned by the (former-telco-only) giants: Indosat, Telkom, Excelcomindo, and Telkomsel. They may have a total international IP bandwidth more than 1 Gbps. Unfortunately, as usual :-D I don't have complete, official, clear, and confirmable numbers for that. Maybe next time, I would come with a simple calculation using several rough statistics found in the media and balanced with some cases (personal experiences).

As can be seen in the mindmap below, Telkom shall have international IP connection (transit + peering) at least 2 Gbps; Telkomsel at least 1 Gbps (It is for me still a question, whether Telkomsel shares the bandwidth from its parent PT. Telkom ??); Indosat at least 2 Gbps (I found different numbers for year 2006/2007). Somebody will to advise ??

Unity Cable System

Unity is a new multi-terabit transpacific cable system with Google's taste :-D As you can see at this post, I've misunderstood that EAC Pacific would completely be a new transpac cable. It shall actually be a part of the Unity cable. To my knowledge, there are currently three competing parties offering new Asia-US connections:

• Trans-Pacific Express (TPE) : Verizon's Taste :-)
• Asia America Gateway (AAG)
• Unity : PACNET, Google etc.

Any other planned cables ?

Some public information about Unity:

• It connects Los Angeles, US to Chikura, Japan
• Length : approx. 10,000km
• Contractors: NEC Corp. + Tyco Telecommunications; the contract was signed on February 23, 2008.
• Members of Unity consortium : Bharti Airtel, Global Transit, Google, KDDI, Pacnet and SingTel
• CAPEX: approx. USD 300 mn
• RFS: Q1/2010
• Capacity: up to 8 FPs (recently deployed only 5 ?; 2 FP owned by PACNET); 96 Lambdas @ 10 Gbps (design); lit ?

Unity Route Map (Source/Courtesy Unity, NEC corp.)

PT. Power Telecom (PowerTel)

Is there any relation between PowerTel Indonesia and PowerTel New Zealand/Australia ? Honestly I've just heard about this company a few months ago and about last week I was quite surprised when several media quoted the company press release that they are ready to offer services in Jakarta Bandung and that they are in the last stage of the preparation of Java-wide fiber optic infrastructure. Well, "Gut gemacht! Gratuliere!"

Facts and quotes:

• "In 2007 the company signed a contract with government-owned railway company PT Kereta Api to install the network along its railway in a 20-year deal"
  
• "PowerTel plans to go for IPO roadshow to Middle East, Singapore, and Hong Kong. The Indonesia's telecommunication network infrastructure provider has confirmed the support from Middle East investor. There could be more Middle East based investors coming in exceeding the locals"
• "PowerTel aims to deploy the optical fiber backbone along Java island. The 3,400 km - 3 ringed - architecture will be finished by 2009"
• "PT Power Telecom (Powertel) hopes to finish construction of a 3,400-km fiber optic cable system between Jakarta and the East Java capital of Surabaya early next year at the latest"
• Network link in Greater Jakarta and Bandung (500 km; RFS June 2008?) ; Stage two comprises a 833km fibre-optic link connecting Jakarta and Surabaya (RFS September 2008 ? Complete rings RFS Q1/2009 ?)
• "Power Telecom is spending IDR 800 bn (USD 86 mn) in its Java fibre-optic deployment and aims to complete the rollout in September" (2008? maybe for link to Surabaya?); from other sources "The company started building the network, estimated to cost IDR 1.2 trillion (USD 130 mn), earlier this year"
• "The network is expected to be linked with international networks from Jakarta and Surabaya" (Surabaya Hongkong?)
  
• "Once Power Telecom has completed its Java rollout it intends to expand its infrastructure in Sumatra in 2008 and Kalimantan in the next four years"
• "The company has allocated IDR 5 trillion to finance its entire capital expenditures for the next five years"
• "To complement its fibre-optic rollout the company is also setting up Wi-Fi and core IP networks to offer high speed internet access in Jakarta, Solo, Yogyakarta, Semarang and Boyolali"
• Capacity ? (multiple 640 Gbps/FP design capacity?; 48 FPs cable?) Suppliers?
  

Sources: TeleGeography, Swa, Kontan, etc.

PowerTel's Java Backbone (Source/Courtesy PowerTel| powertel.co.id)

National and Regional RENs

Actually since a long time ago I want to write threads about RENs (Research and Education Networks). Why? Simply because they were the origin of the current Internet and they still have a major role in creating advances in networking technologies. This time I would not start with an indonesian case (if INHERENT or Jardiknas can be categorized as a National REN/NREN), but with european cases.

BWiN, GWiN and SurfNET

GWiN Connectivity and Traffic Statistics

Each country in Europe has one or more NRENs. These are connected at a higher level with regional european RENs. Some of you might already heard about TEN-155 or currently GEANT/GEANT2; these are RENs at european level. At country level, Germany for instance has formerly BWiN (Breitband Wissenschaftnetz), GWiN (Gigabit Wissenschaftnetz) and currenly XWiN (X Wissenschaftnetz :-)). Other examples: France has RENATER, Holland has SurfNET, UK has Janet and UKnet?. During my dissertation several years ago, I used intensively these networks to measure, calculate and simulate performance of my approaches and algorithms.

These RENs are active and very dynamic since they are intended to fascilitate research on advance networking technologies and their applications. We may take a look the case of the RE networking (r)evolution in Germany:

First Generation (1989/1990) --> X.25-WiN

• 64 kbps in core; 2.5 - 64 kbps access
• X.25 technology
• 1990 upgraded to 2 Mbps (core and access)

Second Generation (1995-1999) --> BWiN

• 155 Mbps in core; 4.8 kbps - 155 Mbps access
• ATM technology
  

Third Generation (2000-2005) --> GWiN

• 10 Gbps in core; 128 kbps - 10 Gbps access
• PoS technology
  

Fourth Generation (2006-) --> XWiN

• heterogeneous technologies based on recent advances in optical transmission/networking

At a glance, from my point of view, XWiN takes the NLR (National Lambda Rail) model to some extent.

Excelcomindo's Backbone

As of Q4/2006, Excelcomindo has installed and operated more than 2,300 km submarine and more than 5,500 km terrestrial fiber optic cable. This information can be used to check or balance my former calculation (however this was for Q4/2007, since it is based on the information from Kominfo on that time frame). I'm just thinking to derive a progress factor, something like fiber deployment rate each year each operator (km/year), just to have a rough metric for measuring operators' progressivity. Whether somebody out there (in Kominfo or in the operators) is already doing this??

Source/Courtesy: Excelcomindo, Hasnul Suhaimi (Q4/2007)

Source/Courtesy: Excelcomindo, Dian Siswarini (Q4/2007)

Mobile Network Coverage

How many BTSs are needed to completely cover a city on average? 50 ? 100 ? more/less ? If each operator has its own tower, then it would be 250 to 500 towers in a city (assuming there are 5 operators, 1 BTS each tower).

Around two years ago, when I came back from Germany after several years leaving, almost every 1 minute or maybe less of my trip from Jakarta to Bandung (also inside the cities) I saw a BTS tower!! Several times later people told me that indeed due to "cellular wars", our cities have turned into "Hutan BTS" (BTS forrests). Well this is the fact. This might:

• positively influence economic sectors (local cellular-supporting industries, maintenance outsourcing industries, PEMDA's income ? etc.)
• also negatively impact several other things e.g. introducing "optical" polution, destroying city layout (regulation is far slower than business!), injecting more electromagnetic radiations to the home (well, EMR-TTH :-)) etc.
  

I almost forget that the title is about mobile coverage. Below are two pics showing coverage areas of Indosat and Excelcomindo. I don't have that of Telkomsel right now, but intuitively, for TSEL you might see more "reds". Statistical numbers and values will be addressed in special threads, maybe in the next days.

Network Coverage (Indosat Q1/2006)

Network Coverage (Excelcomindo Q4/2006)

Looking at these, it seems that they might use a similar software ? Nokia-based ?

Indosat's Backbone (2)

I've just found another version of Indosat's backbone. For me it looks a bit better than the previous one, though some links might still be not so accurate.

Source/Coutesy: Indosat, Brata T. Hardjosubroto, 2007

Palapa Ring (5)

The last non-incumbent member of Palapa Ring consortium, PT. Powertek Utama Internusa has indirectly cancelled its membership (update to this post) . Thus the consortium currently consists of the four-biggest telcos:

• PT. Telekomunikasi Indonesia (Telkom)
• PT. Indosat
• PT. Excelcomindo Pratama
• PT. Bakrie Telecom
  

According to the consortium, Powertek hasn't fulfilled its commitment as stated in the CCMA (Consortium Construction Management Agreement) signed several times ago. The consortium has now to solve / rearrange the investment issue, since it lacks USD 45 mn caused by membership cancellations of Powertek and Infokom.

Jakasusi, Jasutra & SKKL-JS

Below are several domestic submarine fiber optic cable systems owned by PT. Indosat.

Jakasusi (JAva KAlimantan SUlaweSI) :

• Jakasusi segment 1 (B1) connects Banyu Urip (Jawa Timur) to Takesung (Kalimantan Selatan)
• Jakasusi segment 2 (B2) connects Takesung (Kalsel) to Aeng Batu Batu (Sulawesi Selatan)
• Length 1,000 km
• CAPEX USD 40 mio.
• Construction period : March 2006 - October 2006
• Contractor : Alcatel (ASN)
• Subcontractors :EGS Asia Ltd (route survey, burial assessment survey - February 2006)
  
• Capacity : ?

Jasutra (JAva SUmaTeRA) :

• CAPEX: USD 6 mio.
• Contractor : Alcatel (ASN)
• System : (shall be) unrepeatered
• Capacity : ?
• Construction period : 2006 ?
  

SKKL-JS (Sistem Komunikasi Kabel Laut - Jakarta Surabaya)

• Familly/system : repeatered
• Contractor : Alcatel (ASN)
• Capacity : 5 Gbps / 1 FP
• Length (km) : 730
• Number of repeaters : 7
• Service date : 1996

Anybody knows JATISEMARAK ?

Highlight XL (Excelcomindo)

In this new "Highlight" thread, I would archieve some actual facts (of my interest) related to indonesian telcos/ISPs.

Currently, XL (Excelcomindo Pratama) is the third largest telco in the country. This is currently justified if we look at the customer base and some financial performances of the company. The current champion is Telkom Group (inc. Telkomsel) while the "vize" champion is Indosat Group. The position might change in the (near) future, especially since XL "moves" quite progressive after TM acquisition and after the former Indosat CEO Hasnul Suhaimi leads this company.

Customer base

• June 2008: 22.9 mn
• July 2008: 24+ mn
• Dec 2008F: 27+ mn

Network

• August 2008: 14,000 BTS; covers 95+% population
• 2008 as of now: 380 mn calls/day; 60 mn SMS/day; capacity 6,500 SMS/s
• fiber optic deployed 2005: submarine 2,320 km, inland 3,520 km
• fiber optic deployed 2006: submarine 2,320 km, inland 5,530 km

From many sources

International IP Bandwidth

Currently I have only some official information specific to Indosat. This was presented by Indosat several times ago (Apricot 2006 Bali ??). However, it belongs to the "biggest" pipes owned by indonesian ISPs.

Indosat's IP Transit Bandwidth:

• Batam - Singapore --> 465 Mbps
• Jakarta - Singapore --> 930 Mbps
• Jakarta - Hongkong --> 775 Mbps
• Jakarta - USA --> 620 Mbps
  

• Total --> +/- 3 Gbps
  

Indosat's International Peering Bandwidth:

• Total --> +/- 500 Mbps
  

Wiring Kalimantan

As in the case of Sumatera, the biggest three telcos are developing their optical transmission infrastructure in Kalimantan. As you can see below, they have chosen to follow terrestrial paths approximately parallel to the coastal line of Kalimantan. This might mirror the actual social and economic development in Kalimantan. Festoon optical networks could still be possible to be deployed as backup for current almost linear topology.

TIS Cable

TIS stands for Thailand Indonesia Singapore. As indicated by the name, it is an optical submarine cable spanning from Songkhla (Thailand) to Batam (Indonesia) via Changi (Singapore). TIS cable was the first submarine cable project of PT. Telkom.

• Operated since December 2003
• Length: approx. 1,100 km
• Owners: consortium of SingTel, CAT Telecom, and PT. Telkom
• CAPEX: USD 36 mio.
• Capacity: design 320 Gbps, initial 30 Gbps
• Contractor: NEC Corp.
• Construction: 11/2002-12/2003
• Repeatered and unrepeatered system
  

Indosat's Backbone

In this post I just want to put some additional information (maybe "another view") of the Indosat's backbone mentioned in the this post. The map below was presented by Indosat in Apricot conference several times ago (2006, in Bali I guess). Does Indosat use any consortium's cables (SMW, APCN) just to link to Singapore ?