International Internet Speed Testing

[cyborgx]

1) ISP SPEED TEST REVIEWS

Has anyone EVER seen a review of REAL International speeds provided by Thailand's ISP's, in particular here in Bangkok?

I am sure that some Thai computer magazines must have done some.....

Everyone has their own nigtmares and gripes about their ISP(s), but I think it's time we found a consistent and reliable souce for constantly monitoring of the ISP's International speed services here. Independant would be nice too

2) DO OUR OWN SPEED TEST DATABASE

I personally have semi-retired from the IT profession, but I'm sure some bright young spark could put something together. I would be happy to contribute some of my bandwidth to the effort. All it would require is a small application running 24/7 and doing speed tests to reputable sites like www.dslreports.com and logging the results, or even publishing them directly to a web database.

Initially nothing complicated would be required, perhaps just ISP Name, ISP Package, Local City, Remote test Location, UP/DOWN speed.

We poor overpaying underserviced mugs could then just select our city, and package to see what other's are getting, then if you see others are getting the same speeds as you, it will be easy to see if it is worth switching to another ISP's package.

Whilst raw speed tests do not cover all the problems of protocol/traffic type choking etc. done by ISP's it will at least be a very good starting point, and would probably quickly make the ISP's stand up and be counted (we'd probably have to put checks in place to make sure they didn't try to seed out database with inflated results for their company!).

I have 2 Load balanced ADSL lines (True 2.5Mb/1Mb and CSLoxinfo 1Mb/512Kb), and FREQUENTLY find that the speeds are nothing short of abismal! From my initial ADSL line with True (which I got when they first started the service) their International speed has gone from over 2Mbps down to an average about 400kbps, and in the last couple of days both ISP's have been consistently as low as 50kbps! I should go back to a dial up modem!

The more I think of the database, the more I like the idea. There are some comparison options at www.dslreports.com and www.speedtest.net, but they are not particularly clear, and there is not enough detail to be able to view and compare results in a way that would be really useful, and as helpful to everyone as the idea above.

I'd even probably be prepared to pay for the domain and hosting!

So, anyone prepared to write the small application and setup the webpage?

How many of you would be prepapred to participate?

[Firefoxx]

Speed tests are posted here fairly frequently. The one that's most relevant for international tests (for people who frequent US servers) is from speakeasy.net. They have plenty of locations to choose from, and their results are usually accurate (for checking speed to the US).

The thing about doing a review of Thai ADSL speeds for any particular time is that it won't be accurate, except for at that time for that particular location it is tested. Speeds can fluctuate very widely on a daily basis, and can fluctuate from week-to-week or year to year, etc. Not only that, the local congestion of your area also greatly affects your speed. If you have a few greedy bastards hogging up the international bandwidth allocated for your area, your speed will be slow. If not, it may be fast. I've witnessed both these things personally, and can say that it's luck of the draw. Your ISP's link might be behaving and you might have good neighbors, and so your speeds will skyrocket to near 100%. Or, your ISP's link might be congested, and your neighbors might be bittorrent abusers, and so your speed will be a trickle. This is all with the same ISP and package, just different times and different locations.

[Prasert]

Gathering the information you want can be achieved with two things: SNMP and MRTG

SNMP stands for simple network management protocol and is a built-in feature on most networking equipment. One can send a query to such a device and ask for a specific item, e.g. the average bandwidth usage from the last 5 minutes.

MRTG is a well-known freeware package, which creates graphs from 2 input values. Most seen is incoming versus outgoing traffic, but one can modify it to process any values. MRTG uses non-growing databases which contain information from the last 24 hours, a 30-minute average of the last 7 days, a daily average from the last month, and an average over the year.

Sofar the technical part, which is not too hard to setup. Now the practical part:

SNMP is built-in in decent networking equipment. An obvious example is Cisco, which is hardly used in home environments. One requires more than average knowledge to set it up (knowledge beyond the capabilities of the average ADSL engineer), so you'll see the majority using the well-known white TOT modem-router-thing or maybe Linksys or DLink (Linksys has SNMP but it's useless for retrieving data), and those can't be used for MRTG.

But.....

The speed one experiences on an adsl line, is not only determined by the bandwidth of the adsl line. Of course it matters if the line is 512/1024 or 256/512, but more important is the router used: the number of packages-per-second the router can handle is very important and so is the number of simultaneous NAT connections.

So if you have 2 connections in the same place (on the same dslam), one on a decent router and one on a crappy router, you'll see a very big difference in quality.

[cyborgx]

Thanks for the opinions guys. Sorry, I didn't explain fully that I was an IT Pro for about 25 years, so have a reasonable idea of what I am thinking of.

SNMP is a nice idea, but as you say is complicated to setup, and requires technical expertise.

Firefoxx: I have seem some of the posts, bus as you say yourself, they are sporadic, and varied so no real conclusions can be taken from them.

As I said, my idea is a small program running on participating machines that send do regular speed tests (perhaps every hour) and send the results to a central database. The data such as ISP Name, ISP Package (i.e. Down/Up speeds), Local City, Remote test Location can all be drop down lists or detected by the program during the speed test. All it needs to add is the test results and send to website.

Whilst I agree the results will vary a lot, the average for each package can be calculated, as well as any imaginable selective view made be done on the database e.g. average TRUE 2.5/512 between 6pm and 6am.

The website could show league tables based upon anything we like, but all I am looking for at first is collecting some consistent and accurate data, and perhaps doing an initial league table of best AVERAGE dowload speeds (any package). I am sure that's what most people would like to see.

True there are lots of factors that could throw out the numbers, but I'm not looking for 100% accuracy, just an overall view. We can make a fair assumption that the factors which CAN effect the data will eventually average themselves out over time if we have enough results to work with.

Making an effort to do this is a lot better than just reviewing occasional tests by other users that as you say, can be very inaccurate taken individually.

[cyborgx]

BTW Prasert, nice speed graphs on your website, but can't see any indication if they are Up, Down or total bandwidth statistics?

[monty]

I'll give my 2 cents of how I feel about internet speeds in Thailand

In my opinion the whole exercise of setting up a database with test results of different ISP's in different area's is rather futile.

The main reason resulting in bad internet speeds is simply oversubscribing, or a more technical term for this is too high a contention ratio. This stems from the simple fact that the raw product the ISP's have to buy (international bandwidth) is so grossly overpriced in Thailand due to a de-facto monopoly CAT holds on this bandwidth.

As a result the ISP's are setting very high contention(=sharing) ratio's to allow them to keep prices on the "home" packages as low as possible.

There is simply no way to offer reliable fast internet access on packages with a contention ratio of 40:1 or even higher. Not even with the best of the best network specialists and advanced routing equipment!

On these packages it is simply up to luck on what the actual situation (subscriber density, and subscriber usage habits) in your area will be. This is also a very dynamic situation which can change at anytime. As firefox posted, the addition of just one P2P crazy subscriber in your immediate area can instantly bring down your access to a crawl as he saturates the already limited available capacity!

The only viable advice one can give on these home packages is to simply try it out, if it works for you then stick with it, if not your best bet would be to move up to the more expensive business/sme packages.

On to the next point...

It would be a slightly different story on the small business/SME packages. I always use those and up till now they are preforming pretty good in my case. Now indeed it seems true that True (pun intended ) is less honest when setting contention ratio's on their more expensive packages, so much more of interest would be to compare those packages, since at that price level a more decent connection is expected.

At these price levels you should be able to get the same quality as one would expect on European home package, since the ISP's should competitively be able to keep the same contention ratio's at that price level as what is considered acceptable in the West.

And one more point...

A lot of the internet access problems can be traced back at poor routing at the CAT gateway. These problems come and go. They can affect a single ISP or simply slow down the whole internet in Thailand. These problems seem to come and go on a very random base and they can affect one or all ISP's on the same random base.

Again, a test database most probably would be useless, simply because of the randomness of these routing problems...

Unfortunately, the majority of the subscribers are on the cheap home packages and not that many cough up the cash for the more expensive ones, so I'm afraid the test base would be to small in the small business/sme

customer group to get some useful info...

[lomatopo]

A 2.5 mbps ADSL line isn't much use if the uplink from the DSLAM is an E1 (2.048 mpbs) link.

So lots of bottlenecks between your home network and websites in the U.S.A. Your traffic can be prioritized throughout the ISP's network, although it may not be being done, but once they hand it off your packets are just like everyone elses.

Agree a speed test data base, as proposed by the OP, would be of little use.

Thailand: Internet "Map"

[Firefoxx]

Thing is, I've often seen cases (with True, at least) where two sites subscribe to the exact same package (speed, type). They're not that far apart (a few KM) either. Yet they get very different international speed results (local speed results are the same, so it's not a network/line problem). One would get nearly 100% bandwidth, while the other would be lucky to get 10%, at the same time. If you have both sites sending this information to the database, which one would you believe? They're both accurate.

[CobraSnakeNecktie]

Thing is, I've often seen cases (with True, at least) where two sites subscribe to the exact same package (speed, type). They're not that far apart (a few KM) either. Yet they get very different international speed results (local speed results are the same, so it's not a network/line problem). One would get nearly 100% bandwidth, while the other would be lucky to get 10%, at the same time. If you have both sites sending this information to the database, which one would you believe? They're both accurate.

try http://www.speedtest.net/

they have a database setup so you can see what others are getting in any given country/provider

[MikeWill]

Another site to check out is: http://www.numion.com/

[sniffdog]

So suppose you have your database, what are you going to do with it? Watch it over and over again while being frustrated?

The only way to solve your problems is to go back to where you came from ....

[Guest Reimar]

So suppose you have your database, what are you going to do with it? Watch it over and over again while being frustrated?

The only way to solve your problems is to go back to where you came from ....

You're very right!!

[Firefoxx]

Cobra, you don't get my point. My point is that, even for the same provider, with the same package, at the same time, for two different (but fairly close) places, you still get very different results. These results change daily. A database on different countries and providers would be pretty much useless. Gathering any info would also be useless, since it fluctuates so much. You'd be better off putting random results in the database.

[Jai Dee]

Has anyone seen this website?

[cyborgx]

So suppose you have your database, what are you going to do with it? Watch it over and over again while being frustrated?

The only way to solve your problems is to go back to where you came from ....

Except for the stupid entry above, this is an interesting thread.

In response to the comments that the database would be useless, all I can say is that is far too negative and defeatist. If you have a better way, then please suggest one, I'd be happy to help.

Yes, as I agreed before, there will be large variations in the data, but with enough entries, definite differences will be seen between the providers, and this will show up the impacts of their sharing ratios customer priority QoS etc... This will give an average botttom line of what you can expect from your package BEFORE you try it. I don't know about everyone else here, but the bureacracy of subscibing then testing then unsubscribing, new subscription etc. is a lot more work for everyone, and is still subject to the same testing errors put forward by the naysayers.

Concerning the SME packages, I have been given an SME package by Loxley to test, but it made hardly any difference (if any at all).

I put forward this idea, partly because I would like to see the results, and thought it would be better received here.

Admin on adslthailand.com com is already interested in getting involved, and will provide the hosting too if required.