HOME CONNECT

his is the ninth time connect has con- methodology. One of our main concerns was to ducted its wide-ranging broadband and separate the technical performance of the tested fixed-­network test together with Ismaning- connections from the market promises that were, based network test specialist zafaco. in some cases, exaggerated or misleading – for T One might therefore be led to think that so- ­more about this, go to page 86. In line with techno- FIXED NETWORKS mething akin to routine had set in, but this would logy and market developments, we again raised a be wide off the mark. number of threshold values and thus set higher In March 2018 – well before the test period sche- technical standards. And we have a new candidate duled to take place from mid-May to mid-June – this year in the shape of Pÿur (Tele Columbus). the previous year’s winner 1&1 announced in a UNDER THE press release that it did not want to take part in our Nervous tension among the test candidates 2018 fixed-network test. It gave its reason as on­ During the test and in the period thereafter, there going renovation and integration work in its net- were some intense exchanges of opinion with the work. However, participation in the test is not at candidates – for instance due to irregularities the discretion of the candidates. With a view to ­reported by the zafaco monitoring system during editorial impartiality and comparability of results, the tests (see page 89). But we take it in good sport MICROSCOPE connect declined immediately – and included 1&1 – the nervous reactions of some network operators On behalf of connect, the zafaco network experts really put in the test. The results are on page 87. simply demonstrate how relevant our fact-based and critical test results are in the industry. And in Germany’s broadband and telephony providers to the test Improved networks, higher test standards 2018, the details of these results once again throw over a four-week period. The systems collected around 4.4 million As in previous years, zafaco and connect took ad- up a few surprises and interesting insights. The values measured over 48 locations distributed across Germany. vantage of the available time between the last net- next few pages reveal how the individual providers work test a year ago and now to enhance the test performed. HANNES RÜGHEIMER >> The purpose of this task: to provide you with facts about which providers offer the best services.

PROVIDERS IN THE TEST 1&1...... P. 87

EWE...... P. 90

M-NET...... P. 90

NET COLOGNE...... P. 90

O2 / TELEFONICA...... P. 88

PŸUR/TELE COLUMBUS...... P. 91

TELEKOM...... P. 87

UNITYMEDIA...... P. 88

VODAFONE...... P. 89

84 8/2018 85 HOME CONNECT

Kiel

Rostock

Oldenburg Hamburg Telekom 91,9% Bremen TEST SCOPE Meppen Hennigsdorf 1&1 86,3% The selection of tested connections reflects the distribution in the market. Lohne TELEPHONY Münster Hannover Berlin Unitymedia 75,0% We measure the providers against their own promises. Magdeburg Dorsten Bielefeld Candidates can achieve a maximum of 160 of the 500 points total Teltow O2/Telefonica 77,5% Krefeld Bochum Halle For this test, the zafaco measure­ most expensive lines. The distri­ services and when taking measu­ Düsseldorf Leipzig Dresden in the Telephony category. Hilden Vodafone 80,6% Neuss ment systems used a total of bution of connections described rements for web TV. Kassel Köln Bonn Erfurt We calculated the results in the Telephony connection to the older narrowband codec M-net 88,8% 96 test connections at 48 loca­ below therefore takes account of To do this, however, we had to Aachen Gießen category based on 347,971 evalua­ted G.711. We also tested another scenario Net Cologne 76,3% tions in Germany. Network opera­ the range of offers from providers check what level of performance Frankfurt a. M. Mainz Neu-Isenburg measurements. The end-to-end call quality by making phone calls using two channels tors active across the whole of in all bandwidth classes. a provider actually promises to its Bayreuth EWE 75,0% Mannheim of telephone connections to the provider’s while transferring data simultaneously, customers. For reasons of fair­ Saarbrücken Germany having a correspon­ Nürnberg PYUR 55,6% Heilbronn own network (maximum of 40 points), to including both uploads and downloads. dingly high reach (market share The basis: product info sheets ness, we additionally looked at Stuttgart of more than two percent) were Most of the recorded values were whether the tested connection Pfullingen Essenbach other IP networks (maximum of 80 points) Load simulations of this kind reflect how Offenburg Ulm The chart at the top shows the distribution of the Ismaning and to the German mobile networks families would typically use a landline included with 16 connections, not measured against the absolu­ types really are marketed as Eningen u. A. Augsburg points achieved in the call quality test in percen- (maximum of 40 points) was evaluated. connection, for example. tage, and the chart below shows the total points regional and specialised pro­ te maximum data rates of the such. We investigated both of Freiburg Sigmaringen München viders with four. connections but against threshold these factors based on the pro­ The test systems first attempted to As the two bar charts above show, that can be achieved in the individual categories. ­make calls in the HD-Voice mode (Codec ­Telekom, M-net and 1&1 (in that order) When selecting the access values such as “90 percent of the duction information sheets stipu­ DISTRIBUTED ACROSS OnNet products for each provider, con­ promised bandwidth”. We only lated by the Federal Network G.722), which is far superior in quality. If performed best in our test in the Tele­ Offnet GERMANY the connection could not be established, phony category. But even they had nect and zafaco aimed to closely adjusted the thresholds to match Agency. Each provider must pro­ zafaco’s 48 test locations are distri- Mobile reflect the actual distribution in the full capacity in the respective vide these data sheets to the buted throughout Germany. A total of the systems then attempted to establish some room for improvement. 0 20 40 60 80 100 the market, because not every bandwidth classes when mea­ ­public, which can be found 96 telephone and broadband connec- tions ­included in the test established customer uses the fastest and suring uploads for photo book quickly with a Google search. con­nections to each other and per- formed data and video tests. PROVIDERS AND PRODUCTS Number of lines/ PROVIDER PRODUCT Downstream max. Upstream max. Technology locations in the test TELEKOM BANDWIDTH CLASS 1 (UP TO 20 MBPS) in terms of total points. Telekom also 1&1 DSL16 16 Mbps 1 Mbps VDSL 4 of 16 Top marks for Telephony and Telekom 100% EWE DSL16 premium 16 Mbps 1,024 Mbps VDSL 1 of 4 Internet make Telekom the winner ­provides a very good service for photo M-net Surf&Fon Flat 18 18 Mbps 1 Mbps FTTX 1 of 4 of this test, despite a few small book users and gamers. However, a few 1&1 97.5% Net Cologne Doppel-Flat 18 M 18 Mbps 1 Mbps VDSL 1 of 4 shortcomings. measurements indicate that there is still Unitymedia 87.5% O2/Telefónica O2 DSL XS 10 Mbps 2,4 Mbps VDSL 4 of 16 1&1 was crowned the winner in the fixed- ­potential for improvement: Telekom came Telekom Magenta Zuhause S 16 Mbps 2,4 Mbps VDSL 2 of 16 O2/Telefónica 97.5% Vodafone Red Internet&Phone 16 DSL 16 Mbps 1 Mbps 1x ADSL, 1x VDSL 2 of 16 network tests in 2015 and 2017, but, as second to last for download times for the BANDWIDTH CLASS 2 (OVER 20 AND UP TO 100 MBPS) in 2016, Telekom did its utmost to win standardised ETSI Kepler reference pages Vodafone 95.0% 1&1 DSL50 50 Mbps 10 Mbps VDSL 4 of 16 back first place. The Bonn-based provider and in the Web TV category as a whole. M-net 80.0% EWE GIGAglas 75 75 Mbps 25 Mbps FTTX 1 of 4 gained top marks in nearly all the individual Overall, though, our congratulations go to M-net Surf&Fon Flat 50 50 Mbps 10 Mbps VDSL 1 of 4 Net Cologne 87.5% Net Cologne Doppel-Flat 50 M 50 Mbps 10 Mbps VDSL 1 of 4 tests in the categories Telephony, Broadband Bonn for a commanding victory among the O2/Telefónica O2 DSL S 25 Mbps 5 Mbps VDSL 4 of 16 Internet (along with Unitymedia and M-net) Germany-wide providers. EWE 85.0% O2/Telefónica O2 DSL M 50 Mbps 10 Mbps VDSL 4 of 16 and Web Services (along with 1&1). As a Pÿur/Tele Columbus Surf & Phone 60 + TV 60 Mbps 3 Mbps Cable 1 of 4 ­result, Telekom was the overall Germany- Pÿur 57.5% Telekom Magenta Zuhause M 50 Mbps 10 Mbps VDSL 3 of 16 Unitymedia 2play Start 30 30 Mbps 3 Mbps Cable 5 of 16 wide winner and only slightly behind M-net Verdict Telekom performs particularly well in the upload Vodafone Red Internet&Phone 32 Cable 32 Mbps 2 Mbps Cable 1 of 16 very good (447 points) to photo book providers, among other things. Vodafone Red Internet&Phone 50 DSL 50 Mbps 10 Mbps VDSL 4 of 16 BANDWIDTH CLASS 3 (100 MBPS AND MORE) 1&1 DSL 100 100 Mbps 40 Mbps FTTX 4 of 16 1&1 DSL 200 200 Mbps 50 Mbps FTTX 4 of 16 EWE DSL 100 premium 100 Mbps 40 Mbps VDSL 1 of 4 EWE GIGAglas 1000 1000 Mbps 200 Mbps FTTX 1 of 4 1&1 M-net Surf&Fon Flat 100 100 Mbps 40 Mbps VDSL 1 of 4 Very good for Telephony, Internet and Web O2, QSC or M-net, but still operates its own GERMANY-WIDE M-net Surf&Fon Flat 1000 1000 Mbps 50 Mbps FTTX 1 of 4 Telekom 97.5% Net Cologne Doppel-Flat 100 M 100 Mbps 40 Mbps VDSL 1 of 4 Services, potential for improvement with core network, “carrier interconnects” and Net Cologne Doppel-Flat 400 M 400 Mbps 40 Mbps Cable 1 of 4 Web TV – but no grounds for complaint.. – with the acquisition of Versatel in 2014 – 1&1 100% O2/Telefónica O2 DSL L 100 Mbps 40 Mbps VDSL 4 of 16 In view of how well 1&1 did, its attempt to its own fibre-optic network. This makes Unitymedia 82.5% Pÿur/Tele Columbus Surf & Phone 120 + TV 120 Mbps 6 Mbps Cable 1 of 4 withdraw from this year’s fixed-network test 1&1 a fully-fledged network operator. Pÿur/Tele Columbus Pure Speed 200 200 Mbps 8 Mbps Cable 1 of 4 O2/Telefónica 72.5% Pÿur/Tele Columbus Pure Speed 400 400 Mbps 12 Mbps Cable 1 of 4 (see page 84) is all the more surprising. The tested product mix contains 1&1’s own Telekom Magenta Zuhause L (VDSL) 100 Mbps 40 Mbps VDSL 3 of 16 Even though the results were only enough lines and rented lines. Our measurements Vodafone 90.0% Telekom Magenta Zuhause L (Glasfaser) 100 Mbps 50 Mbps FTTX 3 of 16 for the second place Germany-wide in 2018, reveal room for improvement primarily in Telekom Magenta Zuhause L + Speed XL 200 Mbps 100 Mbps FTTX 3 of 16 the Montabaur-based provider and its custo­ the area of Web TV: Here, the video response M-net 100.0% Telekom Magenta Zuhause L + Speed XXL 500 Mbps 200 Mbps FTTX 2 of 16 Unitymedia 2play Jump 150 150 Mbps 10 Mbps Cable 6 of 16 mers can be extremely satisfied with its per­ times could be better, particularly in the Net Cologne 97.5% formance in the Telephony, High-speed In­ middle bandwidth class (between 20 Unitymedia 2play Fly 400 400 Mbps 20 Mbps Cable 5 of 16 EWE 85.0% Vodafone Red Internet&Phone 100 Cable 100 Mbps 6 Mbps Cable 2 of 16 ternet and Web Services categories. 1&1 and 100 Mbps). >> Vodafone Red Internet&Phone 200 Cable 200 Mbps 12 Mbps Cable 3 of 16 does rent connection lines from other net­ Pÿur 55.0% Vodafone Red Internet&Phone 400 Cable 400 Mbps 25 Mbps Cable 2 of 16 work providers such as Telekom, Vodafone, Verdict Telephony connections to the own IP network are Vodafone Red Internet&Phone 500 Cable 500 Mbps 50 Mbps Cable 1 of 16 very good (441 points) Vodafone Red Internet&Phone 100 DSL 100 Mbps 40 Mbps VDSL 1 of 16 one area where 1&1 excels.

86 8/2018 87 88 GERMANY-WIDE best resultforwebsite access. Of allthecandidates, O2/Telefónica achievedthe O2/TELEFONICA for HTTPuploadswithsimultaneousdownload. Unitymedia isoneofthetop-performingcandidates UNITYMEDIA 8/2018 HOME CONNECT O2/Telefónica O2/Telefónica Net Cologne Net Cologne Unitymedia Unitymedia for this:for Standardiseduploadanddown ­Various measurementswerecarriedout individualmeasurements.evaluated arebasedon212,556 Internet category here was to determine the behaviour of here wastodeterminethebehaviour connections.ongoing telephony The aim both directionsandalsoinparallelwith in traffic ments withsimultaneousdata bandwidths. We alsotookthemeasure were usedtodeterminetheavailable load measurementstoproviders’servers in theHigh-speedInternetcategory. pointscanbeachieved A maximumof100the500possibleoverall Vodafone Vodafone Telekom Telekom The points awarded intheHigh-speed The pointsawarded M-net M-net Pÿur EWE Pÿur EWE 1&1 1&1 HIGH-SPEED INTERNET 100.0% 100.0% 100.0% 70.0% 85.0% 85.0% 85.0% 80.0% 90.0% 80.0% 85.0% 85.0% 66.7% 66.7% 66.7% 66.7% 66.7% 100% ­ ­ for connectionsfromandtootherIPnet­ for instance,ments –for thetimes scoresfor measure­ thetelephony good scoresfor side Unitymedia. This isduenot leasttothe solid rankinginthemiddleof field along­ toimproveonlastyear,able andachieveda this year, was theMunich-basedcompany few ofourtestcriteriaweremorestringent own corenetworkbehindthis. Although a access connectionssince2014, butrunsits O2 hasbeenmarketing Telekom bitstream still remain. year’s result. Butafewweakpoints O2 was alsoabletoimprove onlast totheslowconnection This isdueprimarily ­Unitymedia andEWEsharelastplace. –where when itcomestotelephony category. However, therearestilldeficits bestintheHigh-speedInternet ­performing ­improve itsoverallscore, mostnotably ­Baden-Wuerttemberg, to able wasagain of NorthRhine-Westphalia, Hesseand provider, whichis active inthefederalstates paints analmostfamiliarpicture: The cable Compared topreviousyears, Unitymedia shortcomings intelephony. itself further, butstillshows The cableprovider hasimproved fared very wellinthiscategory.­fared very also But mostoftheothercandidates performance.marks –arespectable Unitymedia andM-netallgainedfull didinthiscategory.­candidates Telekom, scenarios. usage are distributedoverthevarioustestand points shows howthetotalachievable The lowerofthetwoadjacentbarcharts thesametime. loadedat width isfully the connectionswhenentireband The topchartshowshowwellthe

­

the pointsinsubcategoriesisshownatbottom. are showninpercentageatthetop, theweightingfor The resultsoftheuploadanddownloadmeasurements simultaneous download simultaneous upload HTTP download with HTTP download good good

servers couldalsobebetter.­servers under load. Responsetimestogaming ­upload bandwidthwasoftennotachieved 100/40 Mbpsconnections, the assured shadow: We with50/10and noticedthat where thereissomuchlight, thereisalso ofthiskind. applications for especially But the lowestbandwidthclassupto20Mbps, the lasttwo,particularly whichiscriticalin – video accessareencouraging times for andthebuffering to photobookservices works. The websiteloadingtimes, uploads result overall. tions with Vodafone, achievedagood inacquisitionnegotia- which iscurrently 100 Mbps). Despitethis, theprovider, the highestbandwidthclass(asof accessing videostreams, in particularly timeswhen width classandbuffering inthe middleandhighestband­ services thetestedwebsites,for uploadtophotobook scenarios–suchastheloadingtimes usage improvementinafewinternet also roomfor setup timesandcalllatency. Butthereis HTTP uploadwith O2/Telefónica HTTP download Net Cologne Unitymedia HTTP upload Vodafone

Telekom (420 points) (422 points) M-net PŸUR EWE 1&1 Verdict Verdict

0 102030405060708090100

100,0% 100,0% 100,0%

82,0% 95,0% 95,0% 76,0% 90,0% 95,0%

VODAFONE in themiddlebandwidthclass(over20to photobookmeasurements upload timesfor Telephony andthesecond-fastest category goodconnectionsetuptimesinthe with very DSL lines. Intheresults, Vodafone shined linesandseven basedonninecable pany ­progress. This year, weassessedthecom­ ­Vodafone ADSL/VDSL connectionsisin Deutschlandnetworkandthe dent Kabel indepen­ betweentheformerly Integration cable connections. all, there were problemswiththeir company.Düsseldorf-based Above It isamixedbagthisyear forthe TRANSPARENCYMONITORING THROUGH German cities, comprises thisplatform units, whicharedistributedacross48 platform. Inadditiontothemeasurement benchmarking self-developed kyago a very limitedperiod.a very ­provide userswithfasteruploads for of 200Mbps. This functionaimsto rate connections withadownstreamdata volumeofmax.data on 15megabytes a increases theuploadbandwidthfor free networkcapacity), “Upload Boost” router,ditions aremet(supportedby these connections: When certain con­ ­unknown optionofDOCSIS3.0on somewhat thepreviously activated ­research revealed, Vodafone hadrecently just priortothisyear’s test. As our on afew Vodafone connections cable when zafaconoticedafewirregularities provider’s hotline. service contactingthe regularcustomer–by any connections, zafacoaddressesthis–like cause isthoughttobeoneofthetested detectedimmediately.therefore Ifthe constantly.this data problemsare Any level,and application butalsoanalyses network ­records themeasurementsat with businessintelligenceplatform. system warehouse andamanagement a numberofcentralservers, adata In Vodafone’s case, thereweresomeirregularitiesduringthetestperiodthatcouldbringdisadvantages forusers. The benchmarkingplatformfromzafaco constantlymonitors thetestconnectionswhilemeasurementsaretaken. For itsnetworktests, zafacousesthe This wasalsotheproceduretaken During thetests, notonly kyago

good

setup. Vodafone madethenewfea only ofthetest order toensurethestability test connectionsaftertheendof April in nochangeswouldbemadetothe that withallcandidates since wehadagreed connections (whichzafacousestotrans sideeffects:­negative Incasesof VPN UploadBoosthasafew the factthat test transparency. is problematic Equally request (https://bit.ly/2rLqUQ4). connect’sture publicinapressreleaseat ourtestascritical, before shortly mentation Both weandzafacoregardthe our opinionwasoverhasty. menting thisfunction, amovewhichin Vodafone imple by diditselfno favours thetimewewenttopress.by Eitherway, ­rectified, clarified hadbeendefinitively question ofwhentheerrormightbe the Fritzbox6490. Neitherthis, northe root causecouldlieinthefirmwareon a VPN withotherinternetservices. userswhoalsouse for than advantages bringsmoredisadvantages therefore perconnection) individually disabled The boost(whichcannotbeenabled/ ofupto20percent. lossrates ­package ofuptotwoseconds and latencies even without VPN, therewere farhigher connection theaffected the internetfor fer measurementresults)andpingsfrom (see boxbelow). connections load boostfunctiononcable introducedup­ also problemswithanewly TV categories. Inthetestperiod, therewere improvement inthe and Web Services Web Mbps. alsohasroomfor And thecompany connectionswith400/25and500/50 cable mised bandwidthwasoftennotachievedon run initsownnetwork. Inaddition, thepro­ download measurementstothetestserver theuploadand poorresultsfor by ­offset under 100Mbps). However, thesewere This goes against ourguidelineson This goesagainst Vodafone the that stated recently

(392 points) Verdict imple

­

­

­ ­

­platform detectsfaults andirregularities. The managementsystemonzafaco’s kyago on theaffectedVodafone cableconnections. far higherpackagetimeshavebeendetected Since activation oftheUpload Boostfunction, upload couldbeimproved. Vodafone’s HTTPdownloadswithsimultaneous Boost wasactivated. connections alsorosesharplywhenUpload Package lossratesontheaffectedVodafone O2/Telefónica Net Cologne Unitymedia Vodafone Telekom M-net Pÿur EWE 1&1 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 57.1% 45.7%

89 GERMANY-WIDE HOME CONNECT M-NET The Munich city network operator Speed Internet checks to the test server WEB SERVICES & WEB TV Telekom 85.0% wins overall on points for the in their own network, best DNS resolution 140 of the 500 points total can be achieved in the Web Services category – covering access to web pages, DNS 1&1 90.0% third time. times, fastest response times for gaming and gaming servers and photo book uploads. A further 100 points are available in the Web TV category. Unitymedia 95.0% The Munich-based company offers fast servers, very good web TV results and the ­connection in its coverage area (Munich second-best telephony results. Uploads and downloads are not every­ Kepler reference page) of five German O2/Telefónica 80.0% and Ulm metropolitan areas, many areas However, the test results also revealed thing. A user perceives the quality of an and five international web hosting pro­ Telekom 91,4% Vodafone 70.0 % of Bavaria, Main-Kinzig district/Hesse) and that M-net’s slower connections (below internet connection in specific applica­ viders were accessed for each browser. 1&1 91,4% achieved the highest number of points in 20 Mbps) have their limitations. As an tions – such as access to websites, In addition, we recorded the loading Unitymedia 85,7% M-net 100.0% the test for the third time in a row. Above all, ­example, the times for uploads to photo ­response times when using gaming ser­ times of different, frequently used web­ O2/Telefónica 86,4% Net Cologne 85.0% customers who are able to get fast, fibre- book services are slow. Our recommen­ vers, uploads to photo book services or sites as well as duration and error rates Vodafone 85,0% optic access from M-net will be pleased dation: If you are interested in M-net, access to videos via content providers. of uploads to photo book providers Cewe, EWE 75.0% M-net 88,6% by the top performance. In this year’s test, choose a high bandwidth. More than 3.5 million of the measure­ ip.labs, Fujifilm and Orwo. Pÿur 80.0% this means: the best results – together with ments we evaluated therefore relate to In the Web TV section, we carried out Net Cologne 87,9% EWE 85,0% M-net leads the field for access to the standardised Telekom and Unitymedia – for the High- Verdict the quality of web services and a further “over-the-top” measurements to video ETSI Kepler reference web page. very good (450 points) 358,384 to the quality of web TV. content providers, such as Amazon, Pÿur 62,9% We paid particular attention to the DNS ­Facebook, Netflix, YouTube, Zattoo and The results for Web Services are shown at the resolution (i.e. the translation of URLs the ZDF Mediathek. Where possible, top in percentage, and the results of the test into IP addresses), because this has a ­these measurements were made with can­didates in the Web TV category at the bottom. significant impact on the response time adaptive streaming. when accessing web pages. To exclude The two bar charts below show the Telekom 72.0% caching mechanisms e.g. in the router weighting of points in the two categories. 1&1 80.0% NET COLOGNE as far as possible and force measure­ The individual results are shown in the Unitymedia 82.0% Despite a somewhat mixed showing, (together with M-net and Telekom) offers the ments to the DNS in the individual chart on the right: The frontrunners for O2/Telefónica 85.0% Telekom 55.0% the Cologne-based regional provider fastest response times for gaming servers. provider’s network, 40 DNS requests web services are Telekom, 1&1, M-net Vodafone 68.0% 1&1 65.0% achieved a good result overall. In addition, the provider comes joint first with from a list of 1000 top German websites and Net Cologne. Net Cologne, EWE, O2/ M-net 84.0% Unitymedia 67.5% As in the previous year, the network provider EWE in the Web TV category. What stood in (according to the recognised “Alexa list“) Telefónica, M-net and Unitymedia provi­ covering the Cologne/Bonn region was able the way of an even better overall result was were selected alternately every hour. The ded particularly impressive services in Net Cologne 88.0% O2/Telefónica 75.0% to improve again in 2018. The company, the high call connection times from and to list of top 1000 URLs was recalculated the Web TV category – while interestin­ EWE 88.0% Vodafone 32.5% which is run by utility company Rhein­energie, other IP networks and the German mobile once a week and sorted in a random gly, big providers like Telekom and Voda­ Pÿur 77.0% the Cologne/Bonn Sparkasse (Savings Bank) networks, amongst others. The results for ­order each day. fone fell behind in this area. M-net 70.0% and the Kölner Verkehrs-Betriebe AG (Colog­ HTTP uploads and uploads to photo book To determine the response times for REGIONAL/SPECIALISED Left: Weighting of points in the individual tests in the Net Cologne 80.0% ne Transport Service Limited), counts on a services could also be improved on – parti­ the gaming servers that are highly im­ DNS Web Services category. Right: Weighting of points mix of fibre optics, coax cable and (V)DSL, cularly for connections in the lowest band­ portant for gamers, zafaco measured the ETSI Kepler EWE 70.0% reference page for the measurements in the Web TV category. which is also reflected in the combination of width class (up to 20 Mbps). times to five different key gaming ser­ Web pages Response and Pÿur 55.0% our test connections (see page 86). On ave­ vers. To test performance for website Gaming servers buffering times Picture quality Net Cologne leads the field when it comes to re- rage, based on these results, Net Cologne Verdict ­access, standardised test pages (ETSI Photo book and error rates 0 20 40 60 80 100 sponse and buffering times for web TV services. very good (428 points) 0 20 40 60 80 100

EWE PŸUR/TELE COLUMBUS Insgesamt erzielt EWE ein gutes anderen IP-Netzen sind spitze. Wegen ­langer Pÿur participated in our test for the ­telephony measurements, Pÿur had the Telekom 83,3% Telekom 97,5% ­Ergebnis. Probleme gibt es bei der Verbindungsaufbauzeiten sowohl im eigenen first time in 2018, but couldn’t quite worst results because HD Voice is not 1&1 90,0% Telefonie und bei manchen Uploads. Netz als auch zu anderen IP- und Mobilfunk­ 1&1 96,3% keep up with the other candidates. ­supported; the provider came second-to-last Unitymedia 91,7% Der Name EWE steht für Energieversorgung netzen lassen die Gesamtergebnisse in der Unitymedia 80,0% The Berlin-based provider Tele Columbus for data measurements. The response times Weser-Ems – der Netzbetreiber bietet zwi­ Sprachdisziplin dennoch zu wünschen übrig. was established as the result of a merger for gaming servers, photo book uploads and O2/Telefónica 91,7% schen Ems und Elbe, in Brandenburg und Bei den Testanschlüssen mit 16/1 sowie O2/Telefónica 97,5% ­between several regional cable providers. Web TV also had some room for improvement. Vodafone 90,0% auf Rügen VDSL- und zum Teil auch Glasfa­ 75/25 Mbps fallen zudem ­Uploads häufig Vodafone 80,0% ­Under the recently introduced “Pÿur” brand, On the test connections with 60/3 and serleitungen an. Die an den Testanschlüssen unter die zugesicherte Geschwindigkeit. Das it covers around 3.6 million households, 400/12 Mbps, the download bandwidth M-net 91,7% ermittelten Durchschnittswerte zeigen ein wirkt sich etwa beim Hochladen zu Foto­ M-net 87,5% mainly in the federal states of North Rhine- was often below the promised data rate, and Net Cologne 91,7% etwas gemischtes Bild. Bei Web-TV ist EWE buchdiensten in der kleinsten Bandbreiten­ Net Cologne 73,8% Westphalia, Lower Saxony, Thuringia, Saxony, DNS resolution times and loading times for klasse aus. Auch für EWE gilt daher: ­sofern EWE 98,3% zusammen mit NetCologne bester Anbieter EWE 82,5% Saxony-Anhalt, Berlin and Brandenburg. web pages were also problematic. >> im Testfeld, auch die DNS-Auflösungszeiten möglich, höhere Bandbreite buchen. In the first year of taking part in our fixed-net­ Pÿur 91,7% und die Sprachlaufzeiten im eigenen Netz Pÿur 58,8% work test, the cable provider unfortunately Bei Web-TV-Bildqualität und -Fehlerraten erzielt sowie für ankommende Verbindungen aus Verdict Pÿur brings up the rear for calls from and to other showed average performance. For the Verdict EWE die besten Ergebnisse aller Kandidaten. good (422 points) IP networks as well as for other criteria. satisfactory (336 points)

90 8/2018 91 MEASUREMENT RESULTS TEST RESULTS O2/ Net Pÿur/Tele PROVIDER Telekom 1&1 Unitymedia O2/ Vodafone M-net Net EWE Pÿur/Tele PROVIDER Telekom 1&1 Unity­media Telefónica Vodafone M-net Cologne EWE Columbus ­ Telefónica Cologne Columbus TELEPHONY MEASUREMENTS Connections to own IP network / other IP networks (max. 40/80) 39/78 40/77 33/64 29/78 36/64 40/70 39/59 34/66 22/47 MEASUREMENTS FOR CONNECTIONS WITHIN OWN IP NETWORKS Connections to mobile network (max. 40) 30 21 23 17 29 32 24 20 20 Number of samples 17,725 17,814 17,702 17,748 17,718 4,398 4,464 4,452 4,446 SCORE FOR TELEPHONY (MAX. 160) 147 138 120 124 129 142 122 120 89 Call setup time (s)/of which > 1 s. (%)/unsuccessful (%) 0.36/0.03/0.00 0.29/0.00/0.01 0.63/0.27/0.01 0.54/0.07/2.65 0.40/0.26/0.14 0.27/0.00/0.02 0.46/0.16/0.00 0.76/2.03/0.25 0.89/19.56/0.09 HTTP downloads with/without simultaneous upload (max. 35/35) 35/35 35/35 35/35 35/35 35/16 35/35 35/35 35/35 32/20 Call quality (POLQA MOS)/of which < 4.0 (%) 4.47/0.01 4.46/0.56 4.47/0.67 4.46/0.28 4.45/1.41 4.46/0.50 4.47/0.35 4.47/0.19 3.61/100.00 HTTP uploads with/without simultaneous download (max. 15/15) 15/15 15/10 15/15 10/10 15/10 15/15 15/10 15/10 15/15 Call latency (ms)/of which > 100 ms (%) 52/0.01 46/0.00 88/7.88 49/0.01 70/2.45 39/0.00 53/0.06 41/0.01 86/4.46 SCORE FOR DATA (MAX. 100) 100 95 100 90 76 100 95 95 82 MEASUREMENTS FOR CONNECTIONS TO OTHER IP NETWORKS DNS resolution (max. 20) 18 18 17 18 16 19 18 18 11 Number of samples A side/B side 19,384/19,432 19,502/19,595 19,410/19,287 19,477/19,449 19,242/19,406 4,261/4,137 4,311/4,287 4,323/4,324 4,287/4,280 ETSI Kepler reference pages (max. 20) 17 18 19 16 14 20 17 15 16 A-side call setup times (s)/of which > 2 s (%) 0.63/0.05 0.64/0.02 0.88/0.08 0.66/0.03 0.61/0.19 0.61/0.00 0.71/0.42 1.23/1.02 1.38/11.40 Web pages top categories (max. 40) 34 34 32 36 32 34 34 34 28 Gaming servers (max. 20) 19 19 17 12 19 19 19 17 10 B-side call setup times (s)/of which > 2 s (%) 0.60/0.11 0.60/0.03 0.91/0.12 0.72/0.08 0.59/0.13 0.76/0.36 0.87/9.75 1.14/1.72 1.14/0.56 Photo book services (max. 40) 40 39 35 39 38 32 35 35 23 Unsuccessful calls A side (%)/B side (%) 0.09/0.06 0.08/0.14 0.08/0.60 0.17/0.09 0.82/0.28 0.14/0.12 0.37/0.49 0.49/0.42 0.12/0.35 SCORE FOR WEB SERVICES (MAX.140) 128 128 120 121 119 124 123 119 88 A-side call quality (POLQA MOS)/of which < 4.0 (%) 4.44/3.37 4.44/3.63 4.44/3.53 4.44/3.45 4.44/3.74 4.37/10.88 4.38/10.51 4.38/10.16 3.58/100.00 Response and buffering times/picture quality & error rates (max. 40/60) 22/50 26/54 27/55 30/55 13/55 29/55 33/55 30/58 22/55 B-side call quality (POLQA MOS)/of which < 4.0 (%) 4.44/3.38 4.43/3.65 4.44/3.64 4.44/3.51 4.43/3.76 4.37/11.23 4.37/10.10 4.38/9.86 3.59/100.00 Performance Web TV services in bandwidth class 1/2/3* good/good/satisf. good/satisf./good –/very good/good very good/very good/good satisf./satisf./satisf. very good/good/good very good/very good/very good very good/very good/very good –/good/good A-side call latency (ms)/of which > 100 ms (%) 65/0.83 63/0.35 81/7.62 63/0.39 72/5.29 61/0.21 68/4.51 64/0.22 100/36.38 SCORE FOR WEB TV (MAX.100) 72 80 82 85 68 84 88 88 77 B-side call latency (ms)/of which > 100 ms (%) 64/0.30 65/1.01 81/7.31 63/0.40 72/5.87 67/10.83 74/12.13 64/1.99 86/15.02 447 441 422 420 392 450 428 422 336 MEASUREMENTS FOR CONNECTIONS TO THE MOBILE NETWORK VERDICT max. 500 very good very good good good good very good very good good satisfactory Number of samples 23,214 23,990 24,613 24,138 21,629 6,097 6,191 6,168 6,210 * Bandwidth class 1: up to 20 Mbps Call setup time (s)/of which > 6 s (%)/unsuccessful (%) 3.15/2.08/0.59 3.67/2.82/1.72 4.20/5.91/0.13 4.03/6.86/1.85 3.11/2.11/0.68 3.53/1.73/0.20 4.17/13.87/0.18 4.37/6.78/1.51 4.72/16.75/0.61 Bandwidth class 2: over 20 to under 100 Mbps Bandwidth class 3: 100 Mbps and more 7(6BEST7 6,(*(5IN TEST 7(6BEST7 6,(*(5IN TEST Call quality (POLQA MOS)/of which < 3.0 (%) 3.34/12.57 3.18/27.23 3.15/22.61 3.19/24.15 3.34/12.62 3.30/13.99 3.14/28.49 3.20/21.98 3.03/28.52 GERMANY-WIDEBUNDESWEIT REGIONAL/SPARTEREGIONAL/SPECIALISED Call latency (ms)/of which > 300 ms (%) 254/19.37 255/21.52 287/43.95 261/25.66 261/24.98 249/20.34 264/30.84 254/18.07 288/42.16 MEASUREMENTS FOR DATA HTTP DOWNLOAD Number of samples 8,935 8,941 8,910 8,943 8,940 2,215 2,239 2,236 2,138 HTTP download throughput < 90% of bandwidth (%) 0.01 0.04 0.09 0.20 0.00 0.32 0.31 0.40 1.08 CONNECT AND ZAFACO TEST METHODS Faulty HTTP download (%) 0.01 0.00 0.01 0.01 0.13 0.00 0.04 0.00 0.00 During their rigorous fixed network test, connect therefore be documented and accessible at CONCLUSION HTTP DOWNLOAD WITH SIMULTANEOUS UPLOAD enjoy the support of their long-standing and www.avm.de/fritz-labor. The same applies Hannes Rügheimer, Number of samples 8,703 8,728 8,704 8,713 8,711 2,186 2,213 2,219 2,105 ­trusted partner, zafaco GmbH, who are based in for devices from other manufacturers, for example connect author HTTP download throughput < 90% of bandwidth (%) 0.00 0.07 0.00 0.02 16.20 0.05 0.00 0.09 6.27 Ismaning, Germany. The company operates test for Telekom’s Speedport router (https://bit.ly/ Although the editorial team had to Faulty HTTP download (%) 0.01 0.01 0.02 0.01 0.23 0.00 0.00 0.00 0.05 HTTP UPLOAD connections at a total of 48 locations. Find out 2KcSdxX) conduct some unpleasant exchanges Number of samples 8,935 8,940 8,910 8,943 8,938 2,215 2,238 2,236 2,138 more about how the broadband lines provided by POLQA values (Perceptual Objective Listening of opinion with some nervous parti- HTTP download throughput < 90% of bandwidth (%) 0.00 0.02 0.13 6.50 0.90 0.27 0.36 0.40 0.19 the network operators are distributed over these Quality Assessment) were determined to assess cipants from prior to, during and Faulty HTTP download (%) 0.01 0.00 0.01 0.01 0.13 0.00 0.00 0.00 0.00 locations on page 86. From 14th May to 10th June call quality, PEVQ-S values (Perceptual Evaluation ­after the tests, the overall outcome is HTTP UPLOAD WITH SIMULTANEOUS DOWNLOAD 2018, an estimated 4.4 million measurements of Streaming Video Quality) to assess web TV positive: The quality of fixed-network Number of samples 8,889 8,910 8,884 8,897 8,893 2,214 2,232 2,232 2,133 were performed by automated test systems at ­video quality. Test criteria and scores can be found services has improved vastly, thanks HTTP download throughput < 90% of bandwidth (%) 0.02 12.56 0.00 43.81 3.31 0.00 47.18 43.59 0.00 ­these locations. The results were subsequently in the boxes on the previous pages. Detailed no doubt in part to our critical tests. Faulty HTTP download (%) 0.02 0.01 0.01 0.02 0.35 0.00 0.00 0.00 0.05 assessed for call quality, data rates, error rates, ­documentation of our comprehensive test proce­ Telekom won the Germany-wide title MEASUREMENTS FOR WEB SERVICES response times and video quality, to mention just dure is available to download at www.zafaco.de/ back from last year’s winner 1&1 – DNS a few parameters. To ensure essential mainte­ de/whitepaper. the result of visibly greater efforts. Number of samples 355,870 356,187 354,764 356,264 356,145 88,181 89,148 89,149 88,989 nance work did not lead to unfair disadvantages, And even though 1&1 seemed to DNS resolution (ms)/of which > 20 ms (%)/faulty (%) 9/1.66/0.02 8/1.09/0.15 12/2.16/0.02 10/0.23/0.12 13/3.15/0.17 5/0.10/0.03 9/1.42/0.01 7/1.18/0.00 24/36.43/0.10 zafaco also included an overnight “maintenance doubt how it was going to perform ETSI KEPLER REFERENCE PAGE break” and did not measure any values in the before going into the test, the Monta­ Anzahl Proben Number of samples 88,275 88,725 88,422 88,662 88,680 21,972 22,229 22,216 22,171 HTTP load time (s)/of which > 5.5 s (%)/faulty (%) 1.77/4.99/0.19 1.61/3.01/0.15 1.61/1.65/0.11 1.57/8.95/0.16 1.84/6.45/0.21 1.35/0.46/0.06 1.58/6.81/0.06 1.59/9.96/0.08 1.71/8.21/0.08 ­period from 2 am to 6 am. baur-based company achieved a 10 WEB PAGES Automated telephony, data transmission and great result. Among the regional and Number of samples 87,915 88,371 88,112 88,322 88,460 21,894 22,153 22,160 22,138 web TV measurements were performed around specialised providers (network ope- Web page load time (s)/of which > 16 s (%)/faulty (%) 3.59/0.47/0.32 3.54/0.49/0.35 3.83/0.35/0.32 3.48/0.51/0.35 3.80/0.39/0.33 3.56/0.16/0.39 3.73/0.25/0.36 3.60/0.19/0.38 3.86/0.34/0.42 the clock. The providers were allowed to set up the rators with a Germany-wide market GAMING SERVER reference server for some data measurements in share of less than two percent), Number of samples 44,187 44,447 44,275 44,413 44,446 11,002 11,137 11,136 11,120 their own network to rule out influences from out­ M-net was victorious for the third Average ping time (ms)/of which > 20 ms (%)/faulty (%) 10/0.00/0.17 11/0.24/0.18 14/4.01/0.15 14/12.47/0.23 13/0.73/0.18 10/0.27/0.15 10/0.01/0.17 13/0.40/0.23 23/81.62/0.21 side their own networks. The end-to-end call time in a row. Net Cologne also PHOTO BOOK SERVICES (BANDWIDTH CLASSES BW: BW1 = UP TO 20 MBPS; BW2 = OVER 20 MBPS TO UNDER 100 MBPS; BW3 = 100 MBPS AND MORE) ­quality was determined within the network and via achieved a very good result. Good Number of samples BW1/BW2/BW3 4,316/6,475/23,712 8,641/8,621/17,251 –*/10,782/23,584 8,619/17,264/8,629 4,312/10,796/19,382 2,092/2,158/4,302 2,157/2,160/4,312 2,163/2,149/4,329 –*/2,153/6,450 connections to other IP networks and to the middle rankings were achieved by HTTP upload time (s) BW1/BW2/BW3 17.89/5.09/1.77 18.47/5.65/2.16 –*/13.18/4.14 16.06/5.97/2.85 19.43/4.61/3.14 36.86/4.76/2.69 34.91/4.99/2.25 36.26/2.46/2.20 –*/15.47/6.70 ­German mobile networks. Unitymedia and O2/Telefónica in the HTTP upload time > 22 s BW1/ 14 s BW2/ 6 s BW3 (%) 0.14/0.11/0.12 0.19/0.06/0.30 –*/0.47/0.53 0.05/0.06/0.20 0.44/0.06/0.12 100.00/0.00/4.65 100.00/0.14/0.05 100.00/0.00/0.30 –*/100.00/41.62 The devices officially offered by the respective comparison of Germany-wide provi- Faulty HTTP upload BW1/BW2/BW3 (%) 0.02/0.06/0.00 0.05/0.01/0.00 –*/0.00/0.00 0.00/0.02/0.01 0.00/0.00/0.03 0.05/0.00/0.02 0.05/0.00/0.00 0.00/0.00/0.00 –*/0.00/0.02 providers with the latest firmware were used on ders and EWE in the regional compa- WEB TV (BANDWIDTH CLASSES BW: BW1 = UP TO 20 MBPS; BW2 = OVER 20 MBPS TO UNDER 100 MBPS; BW3 = 100 MBPS AND MORE) the tested connections. If permitted by the network rison. Despite good performance Number of samples BW1/BW2/BW3 7,489/10,907/41,508 14,866/14,621/29,800 –*/18,877/41,432 14,461/29,324/14,937 7,426/18,680/33,979 3,700/3,778/7,382 3,774/3,750/7,627 3,779/3,810/7,277 –*/3,838/11,362 Video response time BW1/BW2/BW3 (ms) 35/36/33 35/39/26 –*/24/26 32/33/27 40/42/43 22/26/28 31/22/23 28/20/26 –*/33/41 operator, we also activated the auto update func­ overall, Vodafone struggled with a Video response time > 160 ms BW1/BW2/BW3 (%) 2.97/4.65/4.90 2.79/6.18/2.68 –*/0.85/1.13 2.60/2.93/1.75 5.52/5.16/5.84 2.46/2.49/3.08 0.74/0.72/0.66 2.46/1.78/2.79 –*/2.03/2.77 tion on the used routers to ensure the devices few problems which in connect’s Buffering time at start BW1/BW2/BW3 (s) 2.00/1.37/1.38 2.02/1.41/1.28 –*/1.43/1.37 1.72/1.44/1.37 2.28/1.67/1.48 1.99/1.55/1.24 2.09/1.53/1.35 2.09/1.35/1.22 –*/1.57/1.49 were using the most recent firmware version. view are of their own making. New Buffering time at start > 3.4 s (BW1)/ 2.5 s (BW2)/ 2.2 s (BW3) (%) 1.68/2.22/7.36 1.91/2.69/3.38 –*/4.14/8.33 0.53/1.75/5.63 11.52/6.79/11.01 2.27/3.10/3.72 0.98/2.56/3.40 1.61/1.31/3.23 –*/1.46/6.20 ­Nevertheless, it is connect and zafaco’s principle provider Pÿur/Tele Columbus, on the Picture quality (PEVQ-S MOS) BW1/BW2/BW3 3.96/3.96/3.96 3.96/3.95/3.96 –*/3.97/3.98 3.97/3.96/3.96 3.93/3.95/3.97 3.99/3.97/3.98 3.98/3.99/3.98 3.98/3.97/4.00 –*/3.96/3.96 to only test configurations available to the general Network test experts: zafaco managing directors other hand, was not able to perform MOS degradation > 2.0 BW1/BW2/BW3 (%) 0.05/0.02/0.16 0.09/0.15/0.38 –*/0.01/0.01 0.03/0.13/0.23 0.09/0.16/0.04 0.05/0.00/0.00 0.00/0.03/0.01 0.00/0.00/0.00 –*/0.08/0.06 public. For the Fritzbox models used in the test, the Christoph Sudhues (left) and Bernd Oliver Schöttler quite as well as its rivals. Faulty web TV access BW1/BW2/BW3 (%) 0.13/0.26/0.37 0.13/0.25/0.02 –*/0.07/0.05 0.08/0.01/0.01 0.39/0.16/0.09 0.16/0.08/0.11 0.05/0.00/0.04 0.03/0.00/0.03 –*/0.00/0.02 firmware used by the network operator must (right) in front of one of the test systems. Measurements rounded for clarity. * At the time when the test connections were chosen, the network operator did not offer a product in this bandwidth class.

92 8/2018 93