Minister gets loved up on Akamai stats.

clohamon

Our very own Bridget Jones gets torn between suave, charming Akamai and solid, dependable OECD. Let's hope it ends well.

http://www.kildarestreet.com/wrans/?id=2012-05-02.590.0&s=%22Telecommunications+Services.%22

International comparisons on broadband access are complicated and difficult to assess. Comparisons by the OECD, for example, are based on published speeds offered by the largest service providers in each country. These comparisons do not take account of high bandwidth circuits which are negotiated bilaterally between the customer and service provider in each case. In addition, the country comparisons are based on a ranking of advertised headline speeds only, which do not take account of actual speeds experienced by customers. Akamai, http://www.akamai.com/ , an infrastructure provider to the Internet industry with servers located in 650 cities and 72 countries, on the other hand, publishes quarterly data which, among other things, measures and compares average and peak speeds available to customers across 49 countries.


The OECD broadband survey for September 2011, positions Ireland 30th in a ranking of countries based on the fastest advertised broadband speed. This compares to Akami’s report for September 2011 ranking Ireland 13th globally and 7th within Europe for the highest average peak connection speeds provided. Similarly in the case of average speeds, Ireland is 21st in the ranking of highest average advertised speeds by the OECD compared to an Akami ranking of 8th globally and 4th within Europe for its average connection speed of 7Mbps. This compares to 8.5 Mbps provided in the Netherlands, the highest ranked of 23 European countries and South Korea at 16.7Mbps, the highest ranked of the 49 countries compared. It also has to be borne in mind these are average speeds by country which do not reflect actual speeds available in particular areas or regions of the countries compared. Therefore, insofar as comparable broadband statistics on average and peak speeds provided are available, Ireland hovers about the limit of the top quartile of countries compared when measured on speeds provided and performs far better in a comparison of actual speeds measured than it does in a comparison of marketed headline speeds.

bealtine

clohamon wrote: »

Our very own Bridget Jones gets torn between suave, charming Akamai and solid, dependable OECD. Let's hope it ends well.

http://www.kildarestreet.com/wrans/?id=2012-05-02.590.0&s=%22Telecommunications+Services.%22

One false positive and the Minister crows about it...The Akamai reports are very confusing/confused and generally should be ignored. Akamai is mostly concerned with corporate customers and they will usually have reasonable pipes into their premises anyway. Their metrics have very few ordinary customers as part of their measuring. As far as I know 3 is not in the Inex so wouldn't be included in the Akamai figures for instance

All other more sensible metrics show Ireland languishing behind Laos, Nicaragua and Papua New Guinea for upload speeds. This I would believe given the Dept's love affair with mobile midband

I guess the IoffL quarterly report, due sometime in June, will be ignored by the Minister, as usual. He only wants positive spin and will clutch at any straw to get some positive spin.

watty

Any Ireland stats will be horribly misleading now that UPC is rolling out 30Mbps and 100Mbps widely.

jbkenn

Minister "Rabbitte in the headlights" is, like all his predecessors, another gormless gob$hite who has held the office, safe in the knowledge that he will retire on a gold plated pension.
I expect nothing from this time served idiot, therefore, I will never be disappointed.
If you have any regard for your sanity, you will ignore any utterance from,
Minister Rabbitte
the DCENR
Comreg
If the welded them all together, you would not get the combined intelligence of a marrowfat pea.

watty

I think pea plants can turn toward the light.

bealtine

watty wrote: »

I think pea plants can turn toward the light.

They put on the green jersey too...

nickhilliard

bealtine wrote: »

The Akamai reports are very confusing/confused and generally should be ignored.

I'd disagree with this - as a network engineer, I find them to be a highly accurate and useful indicator of network quality.

bealtine wrote: »

Akamai is mostly concerned with corporate customers and they will usually have reasonable pipes into their premises anyway. Their metrics have very few ordinary customers as part of their measuring. As far as I know 3 is not in the Inex so wouldn't be included in the Akamai figures for instance

3 Ireland has been a member of INEX since Feb 2009:

https://www.inex.ie/about/memberlist

It's certainly true that some of the larger ISPs have Akamai caches on their internal networks, but to the best of my knowledge, there are no corporates in Ireland with locally hosted Akamai nodes. They serve data to both corporate end-users and residential. Plenty of really large sites are akamised - e.g. www.mtv.com, www.apple.com, Windows Update, Adobe, Netflix, BBC iPlayer, and many thousands of others. So I think it's fair to say that the vast majority of Internet users in the country use Akamai services on a daily basis.

With a giant statistical basis like this, you can end up with incredibly accurate figures on network service quality.

Nick Hilliard
CTO, INEX

bealtine

nickhilliard wrote: »

I'd disagree with this - as a network engineer, I find them to be a highly accurate and useful indicator of network quality.

Yes on reflection I'd agree with this point based solely on the metrics they use. Those are not the metrics anybody else uses.

you can end up with incredibly accurate figures on network service quality.

All mobile operators are excluded from the figures as is outlined in the report, this is the government solution to broadband performance across the state and it is excluded from the figures so they certainly should not be shouting about it?
Other data sets, which do include the likes of mobile and more rural areas, put Ireland Inc in 50th or 60th position globally

So yes it is a fair measure of fixed line performance across the state which in effect is measuring UPC's performance gain as opposed to DSL.

nickhilliard

bealtine wrote: »

All mobile operators are excluded from the figures as is outlined in the report

yes, just as all mobile operators were classified separately in all other countries. I.e. like was being compared with like.

bealtine wrote: »

this is the government solution to broadband performance across the state and it is excluded from the figures so they certainly should not be shouting about it?

I intend to bleat about broadband speeds in ireland until the day that every home in the country has fibre.

bealtine wrote: »

Other data sets, which do include the likes of mobile and more rural areas, put Ireland Inc in 50th or 60th position globally

Can you provide a reference for this?

bealtine wrote: »

So yes it is a fair measure of fixed line performance across the state which in effect is measuring UPC's performance gain as opposed to DSL.

Probably correct. DSL is overwhelmingly limited by distance to the exchange, so any country which depends wholly or mainly on DSL will see significantly lower figures. Is there a problem with UPC's performance gain on their cable network? I'm not sure I see why this doesn't make a valid contribution to the overall connectivity speed of a country.

I'm not going to argue that rural connectivity isn't a serious problem in Ireland - it very obviously is.

Nick

bealtine

nickhilliard wrote: »

hey Nick

I intend to bleat about broadband speeds in ireland until the day that every home in the country has fibre.

As do IoffL continuously. The telcos can bring fibre to within X kms of every community in Ireland quite easily (for mobile masts for instance) and the local communities can then use their local knowledge to trench the fibre to their area. This is a model being pioneered in the UK and can easily be adopted here in Ireland. Any model that brings ubiquitous fibre across Ireland is worth it.

Can you provide a reference for this?

Speed test metrics. These figures, to me anyway, seem more realistic and inclusive:
http://netindex.com/download/allcountries/
At the time of writing this post Ireland is in 52nd position globally.

I'm not sure I see why this doesn't make a valid contribution to the overall connectivity speed of a country.

I'm not going to argue that rural connectivity isn't a serious problem in Ireland - it very obviously is.

There's absolutely nothing wrong with UPC's contribution, in fact it's great.
But UPC is extremely geographically limited and is not a general solution for Ireland's woes. We need fibre everywhere if we are to compete in the 21st century. Fibre everywhere contributes to GNP. For instance, one recent World Bank study of 120 countries found that for “every 10-percentage-point increase in penetrations of broadband services, there is an increase in economic growth of 1.3 percentage points”. So the economic imperative is there we just need the will and foresight to implement it...

It's not just rural connectivity there are plenty of urban places with substandard broadband and again the only solution to that problem is fibre to the home. This should be relatively easy to fix and probably should be left to the telcos to do but there needs to be some incentives for them to do it.
The current eircom FTTC roll out is a basically a start or a stepping stone.
The rural situation will never be addressed by the telcos so we need intervention by the DECNR or some other state agency.
See http://irelandoffline.org/2012/03/probable-eircom-fttc-locations-2/

The next logical step of that program is FTTH but that seems unlikely to happen any time soon but again is an economic imperative for Ireland's future

Bill Shock

It's not just rural connectivity there are plenty of urban places with substandard broadband and again the only solution to that problem is fibre to the home. This should be relatively easy to fix and probably should be left to the telcos to do but there needs to be some incentives for them to do it.

THe biggest disincentive for telcos is the complete lack of demand for high speed broadband. Posters on this forum can continue to bleat about the lack of broadband in large parts of the country but much of this is down to the speculative nature of the investment that is needed to deliver it.

It would be extremely interesting to know just how many of UPC's customers are signing up to their much-vaunted 100mbps service....my guess is a single figure %.

We are in the middle of an enormous recession and for many people signing up to high speed broadband is still seen as a discretionary item of expenditure. it's totally unrealistic to expect companies make the huge capital investments to bring fibre all over the country in the absence of any certainty around demand from consumers. I'd make the same argument in relation to State investment in the market also.

watty

Limerick City and suburbs that are technically outside city boundary is 3rd City. It's not universally fed by UPC cable. Plenty of places with only MMDS and poor DSL, or no DSL.

Universal fibre is cheaper and of more benefit than ONE major road scheme or LTE rollout (auction starting this summer). It's cheaper and of more benefit than Water meter or smart electricity meter universal roll out or ubiquitous electric car charging stations.

swedex

watty wrote: »

Universal fibre is cheaper and of more benefit than ONE major road scheme or LTE rollout (auction starting this summer). It's cheaper and of more benefit than Water meter or smart electricity meter universal roll out or ubiquitous electric car charging stations.

Can I ask a question, roughly how much would it cost for FTTH for every house in Ireland? 2 Billion???

Surely the ESB could roll it out easily enough and the state would own the network?

watty

A mix of FTTH, FTTK (100Mbps, some 10% only 20Mbps but minimum) and small amount of Fixed Wireless minimum 20Mbps (no mobile or Satellite) and totally universal could be done under €1.5 Billion.

The network assets/fibre/ducts/Microwave links and Masts/sites of RTE NL, ESB, Bord Gais(Aurora), the MANs/eNet, heaNET, CIE/BT-esat fibre, the NRA ducts (and possibly the Eircom street cabinet sites and duct from that to home) should all be owned by a single wholesale agency reselling to all the ISPs at price to "pay back" Universal Fibre rollout costs over 20 years.

AlmightyCushion

bealtine wrote: »

Speed test metrics. These figures, to me anyway, seem more realistic and inclusive:
http://netindex.com/download/allcountries/
At the time of writing this post Ireland is in 52nd position globally.

The speedtest results aren't very accurate. I have a 25Mb connection (I had a 30Mb connection before I moved) and it was rare for it to get close to my max speed. I usually ended up downloading from heanet or usenet afterwards to find out what my speed actually was. The Sligo server when it was there was provided by Fastcom and was barely able to provide more than 2Mb any time I tried it.

My current results on the Irish servers:

Galway: 16.87Mb
Limerick: A whopping 2.04Mb
Derry: 9.39Mb
Cork: 12.15Mb
Dundalk:A whopping 1.29Mb
Dublin - Eircom: 20.22Mb
Dublin - Vodafone: 16.24Mb
Dublin - Digiweb: 16.59Mb

When I downloaded a file on Usenet it was holding steady at 2.6/2.7MBps. So the speedtest results weren't able to match my actual download speed with only one server even getting close and some showing me to have pretty poor broadband. Speedtest doesn't really live up to it's name. I'd only really use it to check my broadband speeds when I know they're poor and to see just how poor they are.

One thing that is lowering our ranking in these results is because of the cheapness of mobile broadband. I've talked to tonnes of people who are using mobile broadband because it's less than €20 a month even though they were able to get broadband from Eircom or UPC. They just didn't fancy paying €40+ a month and/or being in a 12 month contract.

watty

Speedtest.net are only really useful up to 8Mbps to 10Mbps due to loading.

nickhilliard

bealtine wrote: »

As do IoffL continuously. The telcos can bring fibre to within X kms of every community in Ireland quite easily (for mobile masts for instance)

Most rural masts use microwave communications, not fibre.

bealtine wrote: »

and the local communities can then use their local knowledge to trench the fibre to their area. This is a model being pioneered in the UK and can easily be adopted here in Ireland. Any model that brings ubiquitous fibre across Ireland is worth it.

I agree that this is a good idea. However, the wayleave situation in Ireland makes it almost impossible to do this from a financial point of view - unless you have existing wayleave agreements for telecommunications services, which is something that was generally not envisaged in previous legislation.

Wayleave is hard. Don't underestimate the scale of this problem and its impact on the difficulties it causes telecommunications providers in the country.

bealtine wrote: »

Speed test metrics. These figures, to me anyway, seem more realistic and inclusive:

I respectfully disagree. Just for fun, I tested a hosted server of mine a couple of minutes ago. This box has an dedicated empty 100Mbit uplink directly into the core of one of the country's larger ISPs. The results from speedtest servers around the country ranged from 3Mbit /sec to 146 mbit/sec. The 146M link was curious because that particular box has a hard-wired 100m ethernet link. I.e. it's physically impossible to shift more than 100 mbit over that link. None of the results was even remotely accurate for this box, including the ISP's speedtest server. As a reference data point, I've measured that this box shifts 100mbit/sec reliably to other isps around ireland, and if it had faster uplinks, the box is capable of shifting 10-20 gbit/sec without getting too excited.

Some general observations on speedtest.net:

- it claimed speeds that are physically impossible
- speedtest's users are highly self-selecting
- speedtest's test times are nonrandom
- it doesn't provide representative data of actual internet usage
- the choice of ISP matters hugely
- there is no analysis of what data were collected or how the test was conducted
- it operates as a flash app in a browser window
- even high CPU usage on a client machine will cause it to report inaccurate results
- the client has no visibility into existing bandwidth usage on the links it's testing
- the speedtest servers or uplinks are not controlled by speedtest.net

Look, it's a visually pretty system which provides nice-looking one-dimensional numbers that everyone can understand. But please don't confuse its results as being data. At best, they're anecdotal curiosities from a toy.

Regarding Akamai's results, they constitute an summary of actual internet usage from a massive, non-self-selected range of Internet users, collected on a 24h basis from all service providers, globally with no exceptions, and collected exclusively by servers owned and operated by Akamai. With respect, their data are more accurate.

bealtine wrote: »

The next logical step of that program is FTTH but that seems unlikely to happen any time soon but again is an economic imperative for Ireland's future

Indeed, yes.

Nick

clohamon

The problem with Akamai is that city data is extrapolated to represent the entire country. Differences in population dispersion or the relative acuteness of the digital divide in each country are not accounted for.

The exclusion of mobile is significant if it makes up a larger proportion of connections in one country as against another. Ireland has a very high proportion of mobile connections.

watty

I'd agree though with all the speedtest.net criticism, but nevertheless it's less Skewed than Akamai which is too dominated by geographically limited UPC which on average is 10x and heading towards 30x the average for non UPC connections.

jmcc

watty wrote: »

I'd agree though with all the speedtest.net criticism, but nevertheless it's less Skewed than Akamai which is too dominated by geographically limited UPC which on average is 10x and heading towards 30x the average for non UPC connections.

They would produce different types of data, Watty,
Conceptually, Akamai would be closer to broadcast where as Speedtest is closer to P2P. The Akamai dataset (Content Delivery Network providing content to many users) would also have a far wider range than that produced by Speedtest (a far narrower demographic). That would allow Akamai to account for the skew by breaking down the statistics by network/ISP.

Regards...jmcc

watty

Indeed Akamai confirmed that the data the Minister crowed about was "misleading" as they were indeed able to identify that it was skewed by a "well known cable operator".

Ministers of course see what they want to see,

nickhilliard

watty wrote: »

Indeed Akamai confirmed that the data the Minister crowed about was "misleading" as they were indeed able to identify that it was skewed by a "well known cable operator".

I see Ireland Offline has expressed an opinion on this on their web site, alongside several other curious looking statements, but I see no indication that Akamai has actually said anything on the matter. Can you provide a reputable source for this quote from Akamai?

Regarding the general statistics being, Akamai's latest SotI report indicates that 0.7% of non-mobile internet connections in ireland are < 256k, and that for mobile connections, the average connection speed falls between 1.8Mbit/sec and 2.8Mbit/sec, depending on operator (SotI 2011Q4 report, page 39).

If you want to talk about skewing due to high speed cable, please remember that all other european countries have the same effect. Not just that, but if you check out the average peak bandwidth connection speeds for european countries, Ireland falls well down the list to position 18 (SotI 2011Q4 report, figure 33, page 34). This indicates that the countries higher up on this table will have their figures "skewed" even more.

I'd suggest you read the Akamai report. It's interesting and informative - much more so than a single table from speedtest.net.

Nick

watty

It's the Minister that needs to read it properly and not cherry pick what he wants to see.

nickhilliard

watty wrote: »

It's the Minister that needs to read it properly and not cherry pick what he wants to see.

Indeed. Now, can you provide a reputable reference for your statement:

watty wrote:

Akamai confirmed that the data the Minister crowed about was "misleading" as they were indeed able to identify that it was skewed by a "well known cable operator".

Nick

bealtine

nickhilliard wrote: »

Indeed. Now, can you provide a reputable reference for your statement:

Nick

The dataset that Akamai use is secret and not open to scrutiny so cannot be "fact" or sanity checked. Which cities were used to represent Ireland? To the best of my knowledge Akamai only measure the connection speeds in cities where they register more than 50,000 unique IP addresses so they only measure the performance of major towns/cities not rural Ireland. Yet no Irish town/city is in the top 100 globally yet Ireland itself is...

The admission was in private correspondence between Akamai and IrelandOffline and thus was shared with the committee and again is not available in public without their permission. You are always welcome to email the author of the report and clarify. I have his email address

nickhilliard

bealtine wrote: »

The dataset that Akamai use is secret and not open to scrutiny so cannot be "fact" or sanity checked.

Not quite sure what you're implying here. Akamai have no axe to grind regarding Ireland and it's not unreasonable for them not to want to release 1 trillion log entries per day of commercially sensitive information.

bealtine wrote: »

Which cities were used to represent Ireland? To the best of my knowledge Akamai only measure the connection speeds in cities where they register more than 50,000 unique IP addresses so they only measure the performance of major towns/cities not rural Ireland. Yet no Irish town/city is in the top 100 globally yet Ireland itself is...

No. What they said - publicly, in the report, page 12 section 3.2 - was that the city view table data were taken from urban centres with > 50k ip addresses, excluding academic institutions and mobile broadband providers. You're correct to say that no irish town / city is in the top 100 cities in the City View tables in the report.

The Global Average Connection Speed results - where Ireland is ranked as #7 - include data from the entirety of each country, both urban and rural but excluding mobile which is handled separately. This is stated on page 12, top of page:

and includes all countries/regions that had more than 25,000 unique IP addresses make requests for content to Akamai during the quarter.

This appears to directly contradict IoffL's assertion that "The Akamai methodology only accepts figures from fixed line connections in three Irish cities; Dublin, Cork and Limerick".

Regarding "skew", any operator which provides substantially different speed connectivity speeds to the arithmetic average will skew the overall results. The point is that this happens in every area with high speed operators, not just Ireland / Dublin. It is also reasonable to expect relatively high standard deviations from any statistical analysis of a data set which attempts to average out raw numbers over a sample range of which spans two orders of magnitude.

Nick

bealtine

Here is what the author of the report said:
"Akamai has data for many more than 900 cities, but the data set I get for the report only includes cities where we saw more than 50,000 unique IP addresses (geolocated to that city) make requests to Akamai during the quarter.

While there were no Irish cities within the top 100, it appears that both Dublin & Cork would have ranked in the top 200, with Limerick within the top 600 -- this is for the average connection speed metric."

The Akamai data is therefore skewed to city data and is not representative of Ireland as a whole and seems to only measure Dublin/Cork/Limerick but not Skibereen. This is where the "assertion" came from.

No dataset is perfect, the metrics chosen will give a specific set of results. The exact same criticisms of the speedtest results can be leveled at the Akamai dataset as is acknowledged by the author.

https://blogs.akamai.com/2011/11/the-future-internet.html

Nick we're going round in circles here, Akamai clearly stated their data is only for cities and I accept the author's word on the subject.
I really suggest you take the matter up with the author.

nickhilliard

bealtine wrote: »

Here is what the author of the report said:
"Akamai has data for many more than 900 cities, but the data set I get for the report only includes cities where we saw more than 50,000 unique IP addresses (geolocated to that city) make requests to Akamai during the quarter.

While there were no Irish cities within the top 100, it appears that both Dublin & Cork would have ranked in the top 200, with Limerick within the top 600 -- this is for the average connection speed metric."

Your interpretation of what he said is:

bealtine wrote: »

The Akamai data is therefore skewed to city data and is not representative of Ireland as a whole and seems to only measure Dublin/Cork/Limerick but not Skibereen. This is where the "assertion" came from.

I'd suggest that you clarify with the author whether he's referring here to the data used for building the city reports (which the SotI paper clearly states as coming from cities with > 50k ip addresses), or whether he's referring to the data used for the global view reports (which the SotI paper clearly states as coming from countrywide / regional populations of > 25k ip addresses).

It looks pretty clear to me that he's talking exclusively about the City view reports here and not the global view reports. If he's talking about the data used for the global view reports, then he is directly contradicting the stated methodology in the SotI.

More to the point, why would Akamai ignore data from a country as a whole (which they collect, because they collect everything), and only take sample data from the country's major urban centres, then make a public statement in one document that the data is from the combined rural + urban population, but privately say that they extrapolated the data from the urban regions to the entire country, and somehow this magically caused Ireland rating to make it up the global league tables from somewhere between 100 and 200, to position #7 globally?

I suggest you clarify this with Akamai because IoffL has publicly declared your interpretation as a matter of fact, but the interpretation make no sense whatever and flatly contradicts other formal public statements that Akamai have made on the issue.

Nick