Rather embarrassingly, on February 9th, Telstra suffered a significant mobile outage across the country owing to human error in handling a hardware failure in a set of redundant mobile nodes, resulting in the shut down of too many nodes and cascading failures.
To make up for this, they decided that on that Sunday (from midnight to midnight in your time zone), namely Valentines Day (February 14th), that they would offer free mobile data for all customers. Customers would automatically receive this without having to do anything, and received e-mails notifying them of this ‘We’re sorry’ day.
Of course, you can’t please everyone. Telstra is the network without equal, and is often relied upon by Australians to have coverage and performance superior to the other two. As a result, a large number of customers are business customers who didn’t really derive any benefit from the gesture. Some others, however, took full advantage of the situation.
As I really didn’t have any great need to use mobile data on Sunday, I decided to do an experiment. I wanted to know how the throughput of the shared medium of the local base station varied as a function of time-of-day. There is no easy way to do this, so instead, I decided to measure the application layer throughput using Ookla Speedtest.net’s app on my Xiaomi Redmi Note 4 (with a Category 4 LTE modem capable of 150Mbit/s download and 50Mbit/s upload). The server was locked onto Telstra Sydney to minimise transit congestion affecting the results. Readings were taken at semi-random intervals from my room in Chester Hill (just another western suburb of Sydney) around 773m from a base station with excellent 5-bar signals. I wouldn’t be able to afford to do this under regular circumstances as speed-testing a high speed connection consumes a lot of data very quickly. This was my way of “making the best” of their offer.
Unfortunately, as this was not a regular weekday, I wouldn’t be able to observe the network performance from a residential area under normal circumstances. I expected traffic to be a little lighter than normal and possibly less concentrated around business hours. As it turns out, Telstra has come out on the record to say that it saw twice as much traffic as normal, and some customers had slower than normal data speeds on 4G services.
The results obtained were plotted as a chart. As I needed some sleep between the hours of 12:30am and 7:00am, there was no available data.
Key observations include:
- Traffic loading is low during the hours of 11:00pm to 7:00am. During these hours, downstream throughput hovers around 100Mbit/s at the application level which is great for a wireless connection on a phone that only supports the more crowded and less penetrating 1800Mhz band.
- Upload bandwidth generally remained fairly stable during the day, remaining close to 30Mbit/s and only occasionally dipping below 27Mbit/s. It did not vary anywhere near as much as the download bandwidth indicating relatively low demand for upload bandwidth.
- From 10am through to 10pm, the throughput remained mostly within the 25Mbit/s to 50Mbit/s range. Occasional excursions below 25Mbit/s was experienced, but under no circumstances did it get worse than my home ADSL2+ which has just 9Mbit/s up its sleeve. The same cannot be said about the 3G network, as often we experienced throughput of 4Mbit/s during peak periods.
- Even under the “all you can eat” extreme case, performance lived up to and often exceeded the advertised throughput – namely 4G between 2 to 50Mbit/s, 4GX of 2 to 75Mbit/s (Cat4) or 2 to 100Mbit/s (Cat6) and 5 to 150Mbit/s (Cat 9).
- Packet latency appears to increase by about 90ms over the base-line at higher network loading. I wonder if this is a result of router packet queueing indicating large buffers or local LTE TDMA access delays due to slotting of transmission. The latency hovered about 120ms under load, and 30ms under no load.
Of course, such good results are unlikely to be seen in higher density areas where the number of people competing for a cell’s bandwidth may become extreme (e.g. in the City areas). However, what it does paint a picture of is how unambitious the Liberal NBN is.
NBN? Is it enough?
A lot of people who haven’t fully exploited the capabilities of the computing technology they have will always claim that nobody needs 100Mbit/s, and nobody needs even 50Mbit/s. Some even claim such rates are arbitrary.
However, here’s the truth:
A wireless 4G connection from a private telecomms company was mostly able to meet the 25Mbit/s target of the Liberal NBN around the clock, and exceed it at other times, under a free data scenario at my house. They can do it now, while I’m still waiting to be saved from my ADSL2+. Both parties have had no plans for me … not within the next three years.
Speaking of which, FTTN seems to be having issues, with some people wanting to go back to their ADSL2+ after receiving the VDSL2 service due to backhaul congestion. So much for their plans to bring cheaper speeds to the majority quicker.
I think the biggest irony here is that the new NBN proponents had been drumming up the fact that we are in love with mobile data and that fixed line is unimportant.
Might it instead be the case that mobile data has closed the gap and now is beginning to rapidly exceed the speed of fixed line services that is causing fixed line services to fall in popularity despite the higher cost of mobile data?
If indeed, that is the case, of course we would expect mobile data to take off. Back when mobile data was expensive GPRS/EDGE “sub 1Mbit/s” class, and homes had 8-20Mbit/s on ADSL1/2+, the preference for fixed broadband would be clear. Now that I have up to 100Mbit/s LTE and 9Mbit/s ADSL2+, I would probably prefer LTE if it were not for the high bandwidth cost, even with its variable latency.
As a result, we should be deploying even FASTER technology if we want the technology not to be obsolete on deployment. That’s where the FTTP had it right – it had a clear upgrade path for deployment of 2Gbit/s service, like they get on Singapore’s GPON network. It also had many benefits, especially when it comes to reliability. Even if it was running behind deployment time and over-cost, the same can be said about the Liberal’s network and … guess what? It’s going to be slower.
Mind you, they wouldn’t be caught up about this, so instead they love to push articles about multi-gigabit HFC speeds, and other new bleeding-edge VDSL technologies from labs. However, this misinformation is highly disingenuous – anyone who has done any level of research knows:
- Bleeding edge results and technologies are not ready for deployment. They aren’t ready now, and they won’t be ready in a week or two.
- Real life results will always vary because of sub-optimal conditions. Companies are out to make headlines – they want eye catching results, and numbers which can be taken out of context. Under real life circumstances with poor quality copper, thin copper, bad joints, corrosion, crosstalk, etc, you will never get the same results. You can be lucky to get half. Other issues include using shared-media systems where the total throughput or single-subscriber throughput is claimed but under no other-subscriber load situations.
- New technologies have caveats. Don’t expect to get multi-gigabit HFC speeds if you don’t have the right amplifiers with the right level of gain and phase stability and bandwidth. Don’t expect multi-gigabit HFC if you don’t have enough signal to noise ratio. If you expect to run the new technology, you can say goodbye to cable TV altogether.
- New technologies are costly. The marking guys will always claim that it isn’t, but it’s always more costly than the mature technology. Many times this is because of the use of patented non-cross-vendor compatible, non-standards technologies, or a lack of competition and other equipment providers. Ultimately, many of the new technologies are just trying to get in bed with the NBN so as to ensure they have a steady profit stream.
- Cheaper is not always cheaper. Total cost of ownership is what is most important. I think that too many politicians are thinking short-term, whereas infrastructure is a long term investment into the future. That being said, what we are now getting with FTTB/FTTN is a VDSL2 service which does not provide a battery back-up unit to the end consumer, does not even provide a modem, can only provision a single data service (so no more 4-independent data services) and does not provide any POTS service (so say goodbye to voice-band modem technologies which are horribly unreliable over non-QoS managed VoIP). The new VDSL2 also varies in rate depending on the condition of your loop like ADSL2+ does, and performs slower under heavy loading.
Of course, if you hear of the gigabit HFC rates, and you also hear that Optus is trialing LTE-A in the field and actually achieving 1.41Gbit/s, the HFC achievement which still requires further maturity seems a little less impressive.
Needless to say, I’m slightly pleased to hear that the Libs won’t be getting to me anytime soon, but also displeased to know that I’ve got a long future of sitting on a slower line and listening to silly back-and-forth over the NBN.
As for whether we need high speed rates, I think one only needs to realize that good quality high-definition can only be delivered at 16Mbit/s and up, with Blu-Rays commonly providing 36Mbit/s or so, that the 25Mbit/s target is insufficient. Present internet streaming is a big quality compromise, and the difference is obvious when viewed side-by-side. While some will claim that users are signing up to streaming in droves and that’s a sign that the quality is sufficient, they are probably missing the point that they are losing out on a lot of quality while paying for the privilege and convenience. I always prefer Blu-Ray because of this when it comes to content.
Step up to 4K, and you will need about 32Mbit/s or above per-stream using the more efficient H.265 HEVC codec to get good quality (or average quality with 25Mbit/s). Of course, if you have a family sharing the connection, several people are likely to watch video at the same time and the requirements easily add-up. With such real-time services, you need excess bandwidth to accommodate peaks, and you need to sustain the rate throughout the whole feature.
Then you also have software updates, synchronizing cloud storage back-ups, running your own SIP servers, VPN, streaming surveillance video, remote administration of servers/cloud, and other general network uses. For most of this, uploads are just as important and the Liberals are non-committed to this. While I have just 1Mbit/s of upload which is rarely sufficient for even remotely administering one desktop computer comfortably, Telstra 4G offered 30Mbit/s quite easily.
Without symmetric connections, users are discriminated against and ultimately forced to be more consumers rather than producers. If the upload was fast enough and symmetric, people could and would run cloud storage servers from home instead. People would use the internet connection as a VPN between houses and have file sharing operate as quickly as it would on most local networks. People could watch satellite or terrestrial TV received at their other house over IPTV or forward high-rate SDR data for remote-base radio applications. Uploading 4k video would not be such a chore, and concerns about running out of upload bandwidth and not being able to handle one video call and a SIP call simultaneously would be gone.
Ultimately, Telstra’s network did perform as designed and the throughput was quite mind-boggling knowing just how slow mobile data was a decade ago. It gave me an opportunity to rant about just how pointless and unambitious the new NBN is.
On the whole, they claimed a total 24 hour transaction of 1841Tb of data, or an average data rate of 21.82GiB/second or 174.6Gbit/s.
Fun Fact: This is more than the total capacity of Southern Cross Cable undersea submarine cable linking Australia to NZ, Hawaii and USA when it was inaugurated in 2000 at 120Gbit/s. Of note is that through upgrades, this same cable now carries 3,600Gbit/s today with further potential to reach 12,000Gbit/s in the future.