Safaricom’s GPON FTTH is Crippled by Misinformed Decisions and Shareholder Greed

Warning: This will get technical; I’ll do my best to simplify things. However, you’ll be surprised to learn that Safaricom is working on something that could wipe its sins clean. Long Post! (Estimated read time: 15 minutes for slow readers; 12 minutes for average readers).

We all know what happened on February 2021; No, not COVID-19 vaccine. Safaricom finally introduced reasonable data plans, got us excited, and then boom! Data caps. Reason? “We are trying to address the reseller problem.” Many (including myself) were livid. We posted a barrage of irate comments on social media, but I’m sure Peter Ndegwa barely flinched :smirk:. Some “techie” websites even supported Safaricom, claiming that almost no one will feel the effect. This was a direct insult to me :rage:. How could someone say I should be okay with 500GB or 1TB a month when only 10 high-quality Blu-ray movies can easily deplete the latter cap in two days?

Others even suggested that the internet should only be used for “productive” things like education and zoom calls. The audacity :unamused:. I don’t blame them, though. Internet is relatively a new concept in Kenya, and even some decisionmakers at Safaricom might be surprised that movies can be as large as 100GB a file.

Enough of ranting. Back to the point. What is FTTH? GPON/XGPON Standard? Splitting vs Contention Ratio? Can Safaricom’s OLT infrastructure support data-cap“less” FTTH? Why is Safaricom investing in SDN-capable OLT? Well, brace yourselves.

This post will help you understand why Safaricom can do better rather than target “resellers” with data caps in pursuit of profits.

FTTH Basics

Fibre to the Home (FTTH) basically comprises of three main components: Optical Line Terminal (OLT), Optical Distribution Network (ODN), and Optical Network Unit (ONU) (also known as Optical Network Terminal (ONT)). The diagram is as shown below.

ONT/ONU generally refers to your router at home that receives the fibre from ISP. Conversely, OLT is a sort of colossal switchboard that houses several fibre ports. OLT is the origin of your local fibre, owned by supplier ISP, in most cases.

ODN comprises the fibre line infrastructure, optical splitters (in Passive Optical Network (PON)), and fibre cabinets/hubs for distribution. I’m primarily exploring ODN and OLT elements of FTTH in this post.

Safaricom PON Standard

The most common PON standards today are GPON, XG-PON, XGS-PON, and NG-PON2 (see Appendix for more info on speed). The G stands for Gigabit, X for ten, while S denotes symmetrical. Contrary to popular opinions, most fibre networks do not provide symmetrical uploads and downloads. Only standards marked with “S” can offer that. However, ISPs can provide balanced downlink and uplink depending on the number of subscribers on a specific node or line. GPON provides the “least” uploads/downloads (2488 Mbps down, 1244 Mbps up), while NG-PON2 offers the highest bandwidth (can be configured up to 100Gbps up/down). Safaricom uses GPON standard as far as I know; anyone working there currently can offer more insights if the ISP has upgraded.

Optical Line Terminal (OLT)

The brain of FTT“X” (X is a variable here, can refer to H, B, C, etc.) is the OLT, responsible for providing different fibre network standards to clients. As stated earlier, Safaricom uses GPON standard, which is capped at 2.5Gbps down and 1.2Gbps uplink per fibre port (these numbers are rounded off, in reality, it’s 2488/1244). However, OLT devices (such as Safaricom’s Huawei MA5800-X17) are designed to be easily upgradable to the next tier standard.

Splitting PON Networks

The best way to offer fibre to consumers is through Active Optical Network (AON), also known as Point to Point (P2P). In this case, there is almost no attenuation (loss of data). Therefore, if you are connected to a 2.5Gbps OLT port, you share it with no one. ISPs can obviously cap your bandwidth to 1Gbps to make you pay more for extra. However, P2P is expensive because every subscriber requires a port in the OLT. If you serve thousands of customers, you need a lot of OLT equipment and fibre investment to satisfy demand. Hence, P2P is usually reserved for businesses. Still, I doubt Safaricom offers P2P anywhere, maybe to itself (offering a “dedicated 5Mbps uncontested business package” is not an AON thing, that’s 0.002% of what a typical OLT port can handle). However, someone here suggested that Liquid Telecom Kenya boss has a 1Gbps connection to his house. That might be a real P2P. To reduce costs, ISPs resort to the almighty PON (also known as Point to Multipoint-P2MP).

PON vs P2P
PON (P2MP) vs. P2P

PON network involves offering internet to multiple clients using a single fibre port on an OLT. For example, rather than Safaricom offering everyone real estate on their OLT, they save costs on fibre and equipment by making people share a single 2.5Gbps port on OLT. This sharing should not be confused with the contention ratio; we’ll get to that in a moment. Sharing fibre is done through passive optical switches known as splitters. The most common splitters with minimal attenuation are PLC splitters. Safaricom splits its FTTH fibres in a 1:64 ratio, meaning that 64 people share a single fibre from a splitter direct to the OLT. I don’t think they have upgraded yet.

The downsides of PON are attenuation and reduced bandwidth. In Safaricom’s case, a 1:64 split should be within a 5KM radius to avoid high data losses. 1:32 split is the sweet spot, so I’m surprised why Safaricom went with 1:64. An OS2 single-mode fibre has around 0.4dB loss per KM, but this attenuation adds up quickly due to distance, splits, and splicing.

Split Ratio Vs. Contention Ratio

You have heard many people claim Safaricom uses a 1:5 contention ratio. This ratio only tells you how many people are competing for your provisioned bandwidth. The focus should be on the fibre split ratio, which, as you have seen above, stands at 1:64 (AFAIK). This means that, in a GPON, 2.5Gbps down and 1.2Gbps up is subdivided 64 times. Of course, this is not always the case. Some split ports might be empty, waiting for people to dial *400 # to “get fibre.” Also, some users might opt for the lowest fibre plan, meaning that split bandwidth is not equal in reality.

It is essential to know that split doesn’t determine the speeds you get; OLT determines all this.

The easy/crude way to calculate contention ratio, assuming everyone has a similar plan, is this: if 64 people buy 100Mbps and share 2.5/1.2 GPON, contention down is (64X100)/2488=2.6 meaning 1:2.6. However, Safaricom institutes a router-based contention ratio intended to balance out subscribers depending on plans (and, of course, milk more money from you).

Why Safaricom can do Better

With technicalities out of the way (finally), you can see that the fibre network is solely dependent on OLT capacity. The fibre strand itself (single mode OS2) has almost unlimited theoretical bandwidth support. Once an ISP lays the ODN, future upgrades are only on the OLT side. Reason? Lasers used to transmit data are expensive, and moving from, for example, GPON to XGS-PON is costly. Nevertheless, a company like Safaricom can easily afford them if it wanted to.

Safaricom uses Huawei’s MA5800 X17 OLT that costs around $25,000 a piece (AFAIK). This OLT can support up to XG-PON (10Gbps per port asymmetrical), but Safaricom chooses to use GPON for now. I’m guessing the reason for this is power demands, cooling, and maintenance costs avoidance.

As indicated in the “splitting ratio vs contention ratio” section, Safaricom needs to find a better way to deal with resellers. Its fibre infrastructure is mostly more than capable of handling demand. The easiest way to handle an unusual demand is just to upgrade its OLT to XG-PON, heck, even XGS-PON and NG-PON2.

About the alleged resellers, Safaricom should understand that people who buy internet from them just want a connection. It is up to Safaricom to invest in the infrastructure and ensure that most parts of urban areas are covered to diminish resellers’ “opportunities.” Yes, fibre infrastructure is expensive, but genuine users should not be made to pay for a few bad apples.

It is also possible that some decisionmakers at this telco cannot wrap their heads around how people can use over a terabyte in a month. To them, users only stream, play mobile games, and maybe rant all-over Twitter. Hence, they might assume that anyone out of the ordinary is just a reseller. Their double-data offer in 2020 might have reinforced this idea since people working from home were supposed to consume gobs of data. It’s time such mindsets end. I cannot share a fibre with 64 people, contend my bandwidth with 5 others, and still put up with a data cap for Ksh12,000!

There is Hope

Safaricom is not all rotten, though. I was surprised to learn that the telco is investing in Open Access Network (OAN). Last year, Safaricom called for potential suppliers to apply for supplying Software-Defined Networking (SDN) OLT. SDN-based OLT can decouple ISPs from the physical OLT and allow multiple ISPs to use the same equipment and ODN infrastructure. See what Safaricom said about its vision for FTTX below; the whole RFI can be found here.

This means that in the future, it is possible that Airtel, Telcom, etc., will be able to supply fibre to us through Safaricom-built infrastructure (That’s how South Africa’s FTTH work btw). The new SDN-based OLT will also support NG-PON2 up to 100gig per second, meaning more goodies and more competitors. I feel bad to say this, but, thanks Saf :smirk:. However, we might have to wait for maybe five more years until Ndegwa milks enough of our cash.

Thoughts are mine. Data sources available upon request




Very informative, I paused at “Ndegwa milks enough…” :grin:

1 Like

Good Piece. Though you couldnt be more wrong… Sufferingcon have never been known to cede ground/ competitive advantage. Look at what happened to Unlimited Mobile Internet, sharing of M-Pesa shops…

:joy::joy: After reading their report I was convinced they genuinely have good intentions


They are the market leader/ mover… Why would they suddenly develop a conscious/ good intentions…

1 Like

Nice write up . On the OAN topic, imo, the reason why most big companies are jumping on it, is because of licencing fees. The problem most companies are facing is once you choose a vendor for your deployment, you have to stick to them all the way. This has led to these vendors increasing licencing fees once they have someone locked in or charging very high prices for upgrades because they know that a competitors equipment is incompatible with the already existing hardware. By decoupling the hardware from software it enables these companies to be able to upgrade to what ever hardware and from whichever vendor they see fit. all they have to do is to configure it and it will be compatible with their existing infrastructure. it also enables them in some cases to develop their own software for the hardware enabling them to support it for however long they want instead of it being on the vendors terms

1 Like

Nice post would you be up to do one on the business end of Internet Exchanges points (IXPs)? Curious to know if this is the reason for the data caps as data originating locally(increasing with bump up to speeds and new subs) transits to the different networks …

Thanks for the feedback. I wanted to include info about IXPs but that was beyond the scope of this article. If I get time in the coming weekends I will explain IXPs and why local content producers need to keep their data on local servers.

Really Informative, what do you think about outbound traffic(outside Kenya). I feel this is usually the biggest bottleneck/expense.
Here are some of the undersea cables we have and their designed capacity

  • TEAMS: 5.2 Tbps
  • LION2: 12.33 Tbps
  • EASSy: 18.6 Tbps
  • SEACOM: 12 Tbps
    DARE - might not be that useful for outbound traffic

I am not sure how much it costs for Safaricom to use the cables, but I feel this is probably the reason for data caps, low bandwidth in most African countries
Compared to US/CA where most of the traffic is within the continent

According to Safaricom regains home fibre market lead - Business Daily there are about 600K fixed internet connections in Kenya.

If we give each home an average of 100Mbps connection, that’s 60Tbps at max usage. using a conservative, 10% average usage that’s still 6Tbps which is not including mobile data and other countries that use the same undersea cables

what are your thoughts on this?

I came across this article from Business Daily.

According to data from the Kenya National Bureau of Statistics (KNBS), the total available bandwidth capacity in 2019 was 6.2 million Mbps, a twofold increase from two million in 2016.

“We will increase the bandwidth by ten times to have faster Internet speeds for citizens. The Covid-19 pandemic has made us think and work differently. The Internet has become everything to us, we are too much becoming dependent on the Internet,” said Mr Ochieng’.

“We are at 95 per cent complete and certainly within the next two to three months, we will be able to finish the administrative work before Kenyans start enjoying the benefits.”

The current capacity is ~6 Tbps (it is possible the undersea cables are underutilized). In the next 2 to 3 months, the bandwidth should increase by ten times. That is >60 Tbps.

I am not sure if this will benefit the common wananchi but it sounds like it will.

In other news: I find it funny how Telkom Kenya always gets mentioned in these big internet infrastructure projects but their 4G internet takes 3-5 business days to download an email attachment. :joy::joy::joy:


“We seek to upgrade from the current 10 gigabits per second to 100 gigabits per second thanks to the deployment of ultrafast gigabit broadband connections” he said.

It does seem a lot of our cables were 10 Gigabit. They are upgrading those to 100 Gigabit.


Good points. However, I do not think these corporate submarine cables have a significant effect on Safaricom’s consumer side of the business. First, most consumers subscribe to the 8Mbps bundle (Safaricom’s words).

Second, tier 1 ISPs like Safaricom don’t buy bandwidth in bundles. Third, the fibre strand itself is not the bottleneck, but the sending and receiving infrastructure.

Fourth, it is also essential to remember local traffic accounts for around 70% of the total bandwidth. In 2012 it was around 30% (This means that most of the bandwidth doesn’t exit Kenya to saturate international infrastructure. Even Netflix has local cache). All these factors mean that Safaricom is just exploiting consumers for profits. I know they ain’t a charity, but come on, that’s too blatant :smirk:

Even if we theoretically assumed every consumer had 100Mbps connection, it doesn’t mean everyone is using 100Mbps at the same time or even half the time (I don’t know if I’m making sense lol). But you have made pretty good points on submarine cables, especially when corporates are scrambling for bandwidth. I’ll look into that further. thanks

1 Like

Interesting, never knew local bandwidth was that high. Just wished Ubisoft and Steam had CDNs in Kenya.

1 Like

considering the capacity i am seeing here,we shouls have the lowest package at 100Mbps and the highest at 1Gbps at the very least.


Sorry for this mistake, Safaricom is not tier 1 network. Tier 1 networks don’t pay any fees for international connection. Saf lies between Tier 2 and 3. Pays a little to probably UK’s Liberty Global, one of the only two tier 1s in UK (and 1 of the 16 globally, these don’t pay anything, they just peer freely with each other). Since African submarine fibres terminate in the UK, it’s only logical to assume Liberty Global is the best for them.


I swear I’m not even trying. 15 days in and I’m hovering at just over a TB. It’s funny how Safaricom assumes things. No, I’m not on Safaricom but a Liquid Telecom reseller, there are always alternatives, don’t compromise on your rights :smirk:

FTTH or p2p ?

Ni FTTH, walikataa kunipea P2P :rofl:

How’s Liquid Telecom, reliability and speedwise…

So far so good. Can’t complain honestly.

1 Like