Help requested from IT nerds

[Blackbird]

Comment from the IT nerds if you please.......

Living in the boondocks on the edge of Coromandel Harbour, our broadband was previously delivered by copper line with the dizzying speed of around 3.3 Mpbs download and <1 Mbps upload. A few months ago, Spark gave us the option of going wireless broadband from a new cell tower just across the harbour.

The difference was night and day and we were consistently getting download speeds between 40 and 60 Mpbs. Uploads were north of 15 Mpbs on a pretty consistent basis. When Spark rang to offer the service, they guaranteed 20 Mbps or better. However, during the last week on occasions, some web pages (including Kiwi Biker) have been slow to load and have sometimes timed out. Uploads and downloads for photos and other data have also been painfully slow on occasions. Whilst I’m still happy overall with the wireless service, I wanted to get some evidence if I need to talk to Spark about it.

I set up an automated speed test (http://testmy.net/) which runs hourly sampling using different size data packets. Referring to the attached graph, although the average download speed was 37 Mpbs, it was skewed by some occasional big download speeds in the middle of the night. It can also be seen that some of the download speeds during sampling were quite poor. Uploading was universally mediocre.

I don’t intend to go to Spark with all guns blazing but if I do need to talk to them at some point, I’d like to be a bit more informed beforehand. It might be that since the tower came into being, increasing user numbers has slowed performance, or at least made it more erratic. Maintenance may also be an issue. I’m also wondering what the Spark guarantee of 20Mbps actually means in real terms. I’d like to think that it actually means daylight hours and not across the whole 24 hour day. Bugger-all good having high speeds when most people are asleep. Reasonable consistency in speeds would be nice too.

I guess that’ll cover things for now but has anyone got a decent knowledge of how wireless broadband via a cell tower actually works? (in lay terms, not geekspeak please)

Cheers,

Geoff

[cods4]

Looking at that graph it seems the speed drops to zero occasionally which would indicate the connection is going down. This would most likely be caused by an issue with your router. It could be buggy or something. I would start by rebooting the modem/router and then trying it again for a few hours. If that doesn't help then I would reset the modem/router to factory settings then try again. If you still notice no difference then contact spark and they may send you a new router or look for issues at their end. The other cause could be the radio tower is oversubscribed and not able to cope and therefore dropping connections. But this is fairly unlikely.

If it was working correctly I would expect the speed to be fairly consistent for most of the day and then it would maybe slow down a bit around 6pm-10pm which would normally be when the peak traffic is.

[Blackbird]

Looking at that graph it seems the speed drops to zero occasionally which would indicate the connection is going down. This would most likely be caused by an issue with your router. It could be buggy or something. I would start by rebooting the modem/router and then trying it again for a few hours. If that doesn't help then I would reset the modem/router to factory settings then try again. If you still notice no difference then contact spark and they may send you a new router or look for issues at their end. The other cause could be the radio tower is oversubscribed and not able to cope and therefore dropping connections. But this is fairly unlikely.

If it was working correctly I would expect the speed to be fairly consistent for most of the day and then it would maybe slow down a bit around 6pm-10pm which would normally be when the peak traffic is.

Thanks for the thought and I'll file that for investigation over the next little while. The router is a Huawei supplied by Spark when we went wireless. Looking at the data which went with the graph, data transfer never drops to zero, but is a very low kpbs number. Much appreciated

[jasonu]

Comment from the IT nerds if you please.......

Living in the boondocks on the edge of Coromandel Harbour, our broadband was previously delivered by copper line with the dizzying speed of around 3.3 Mpbs download and <1 Mbps upload. A few months ago, Spark gave us the option of going wireless broadband from a new cell tower just across the harbour.

The difference was night and day and we were consistently getting download speeds between 40 and 60 Mpbs. Uploads were north of 15 Mpbs on a pretty consistent basis. When Spark rang to offer the service, they guaranteed 20 Mbps or better. However, during the last week on occasions, some web pages (including Kiwi Biker) have been slow to load and have sometimes timed out. Uploads and downloads for photos and other data have also been painfully slow on occasions. Whilst I’m still happy overall with the wireless service, I wanted to get some evidence if I need to talk to Spark about it.

I set up an automated speed test (http://testmy.net/) which runs hourly sampling using different size data packets. Referring to the attached graph, although the average download speed was 37 Mpbs, it was skewed by some occasional big download speeds in the middle of the night. It can also be seen that some of the download speeds during sampling were quite poor. Uploading was universally mediocre.

I don’t intend to go to Spark with all guns blazing but if I do need to talk to them at some point, I’d like to be a bit more informed beforehand. It might be that since the tower came into being, increasing user numbers has slowed performance, or at least made it more erratic. Maintenance may also be an issue. I’m also wondering what the Spark guarantee of 20Mbps actually means in real terms. I’d like to think that it actually means daylight hours and not across the whole 24 hour day. Bugger-all good having high speeds when most people are asleep. Reasonable consistency in speeds would be nice too.

I guess that’ll cover things for now but has anyone got a decent knowledge of how wireless broadband via a cell tower actually works? (in lay terms, not geekspeak please)

Cheers,

Geoff

I think the term you should be using is IT experts...

[Gremlin]

Think of your connection to the outside world as a chain, or a series of links.

At the most simple, on copper, it was PC/Device > Router > Copper > Exchange > Backhaul > Internet
On wireless, it's PC/Device > Router > Tower > Backhaul > Internet

Us IT "nerds" always like cables. They offer more or less consistent performance and to a degree, are not interrupted by the cables around it - for these purposes anyway (no, I'm not including the hubs like Mediaworks / TVNZ and their kilometres of coax etc).

Wireless on the other hand has a spectrum consisting of channels, and this spectrum has a fixed range (differing for each technology etc). Your home wireless shares the same range as your neighbours range (the 2.4GHz home wireless actually only has 3 non-overlapping channels in 11).

So, getting back to the point (now I've provided some background), to do a good test, more information is needed. Are you running the test from your PC, and is it cabled to your modem, or using wireless? The slowdown could be anywhere from your device to modem, modem to tower or further afield. Obviously only items in your house are under your control. Yes, you could expect that if takeup of wireless broadband is high in your area, that there could be more interference/congestion from the other devices on the network (same as when 4G wasn't launched - the techs testing were getting blazing speeds).

As a side note, the speed guaranteed is a next hop, ie, like running a speedtest to a local ISP. You won't get that speed if you test against an international server in Australia, USA etc. Kiwibiker from memory is USA hosted? Guaranteed speed to International is mega expensive.

[Blackbird]

So, getting back to the point (now I've provided some background), to do a good test, more information is needed. Are you running the test from your PC, and is it cabled to your modem, or using wireless? The slowdown could be anywhere from your device to modem, modem to tower or further afield. Obviously only items in your house are under your control. Yes, you could expect that if takeup of wireless broadband is high in your area, that there could be more interference/congestion from the other devices on the network (same as when 4G wasn't launched - the techs testing were getting blazing speeds).

As a side note, the speed guaranteed is a next hop, ie, like running a speedtest to a local ISP. You won't get that speed if you test against an international server in Australia, USA etc. Kiwibiker from memory is USA hosted? Guaranteed speed to International is mega expensive.

Many thanks Alan, there are some really useful pointers there. The testing site I used seems to use Sydney as an end point. It doesn't use flash, just web streaming or whatever the proper term is. I used it because of its ability to run the tests whilst I was off doing other stuff and to run the tests using differing packet sizes. It makes perfect sense to speedtest locally so I'll use Ookla on a random basis to compare as well as doing another run with the original test. It's interesting that it only became an issue a short time ago, almost a step change. Thinking out of left field, I don't suppose the tower transmitter could have been knocked slightly out of kilter by the high winds last weekend? Probably not, I'd have thought.

Cheers mate!

[Akzle]

Looking at that graph it seems the speed drops to zero occasionally which would indicate the connection is going down. This would most likely be caused by an issue with your router..

or more likely the dhcp renewing his IP.

[Akzle]

.

As a side note, the speed guaranteed is a next hop, ie, like running a speedtest to a local ISP. You won't get that speed if you test against an international server in Australia, USA etc. Kiwibiker from memory is USA hosted? Guaranteed speed to International is mega expensive.

an ISPs guarantee is worth 32.3% of the average shit. throughput and uptime are basically optional unless you pay about 4k$/mo.

[Gremlin]

an ISPs guarantee is worth 32.3% of the average shit. throughput and uptime are basically optional unless you pay about 4k$/mo.

Mainly people want unlimited data at fast speed for about $50... and that's not going to happen. I'm reminded of a case in Dunedin shortly after it became gig... someone apparently did about 2TB within a few days... the poor companies providing the backhaul for that

We're a virtual ISP for our clients, controlling our address range, data and routing etc, but yes, still depend on the network beyond us working... Last memorable one was a fire in Sky Tower taking out some equipment

[russd7]

been down that path, most likely cause is the tower getting overloaded and unless you start making a shit load of noise to your service provider then you will have to put up with it, they do have the ability to up your speed and the noisiest wheel definately gets the attention. what you need to remember is that when you signed on there was probably only a hand ful of people that did but once they started bragging about their speeds then everyone else would have jumped on the wagon and the whole lot will be over loaded with people gaming and streaming etc,

[Blackbird]

been down that path, most likely cause is the tower getting overloaded and unless you start making a shit load of noise to your service provider then you will have to put up with it, they do have the ability to up your speed and the noisiest wheel definately gets the attention. what you need to remember is that when you signed on there was probably only a hand ful of people that did but once they started bragging about their speeds then everyone else would have jumped on the wagon and the whole lot will be over loaded with people gaming and streaming etc,

Appreciate the comments Russd7. Because of the location, the tower would (presumably) only service a few hundred people at most seeing as Coromandel only has a static population of around 1500 over the bigger area. With all the tourists that will start flocking in over summer, it could take a further hit so will keep monitoring.

[russd7]

Appreciate the comments Russd7. Because of the location, the tower would (presumably) only service a few hundred people at most seeing as Coromandel only has a static population of around 1500 over the bigger area. With all the tourists that will start flocking in over summer, it could take a further hit so will keep monitoring.

speaking from experience. remember that the tower may only be serving a few hundy but each house will probably have at least a couple of cell phones hooked in to the tower as well as at least one pc or laptop and probably a tablet or two, add to that the tourists coming through and all of a sudden that few hundy are probably looking more like a couple of thou devices hooked in.

we lived in an area that the only net we could get was wireless through woosh then via vodafone through their cell network, woosh were next to useless and vodafone were great to start with then got progressively worse as more and more people signed on.

we shifted now and back on cable based vdsl

[BMWST?]

you arent getting to the data cap are you?

[Blackbird]

speaking from experience. remember that the tower may only be serving a few hundy but each house will probably have at least a couple of cell phones hooked in to the tower as well as at least one pc or laptop and probably a tablet or two, add to that the tourists coming through and all of a sudden that few hundy are probably looking more like a couple of thou devices hooked in.

we lived in an area that the only net we could get was wireless through woosh then via vodafone through their cell network, woosh were next to useless and vodafone were great to start with then got progressively worse as more and more people signed on.

we shifted now and back on cable based vdsl

Yep, I'm more inclined to think that it's a volume thing. I've got an automated speed test running hourly to build data so that I can talk objectively with Spark.

you arent getting to the data cap are you?

Good thought but no! Checked today 35 Gb used from a 120 Gb allocation.

[Akzle]

I've got an automated speed test running hourly to build data so that I can talk objectively with Spark.
.

exactly no-one at spark gives a fuck about your speedtests. 0.

what's your latency nic to dns?