Philip Khor

**Singapore tests update at the bottom of page**

** more Detailed analysis after the break **

Went to cell on Friday night. Did a poll and found out there were quite a number of DiGi supporters. With their amazing deals on phones, this strategy will pay dividends down the line. A small company undercutting the big giants in Maxis and Celcom, wise move!

Now that the iPhone is coming soon, it would be nice to see how the synthetic benchmarks fair for their network performance. I’m tethered to my Sony Ericsson phone capable of 3.5G HSPDA speeds. I have “H” signal and 5 full bars of network signal. Here is how it went.

Method: Using speedtest.net for raw speed and pingtest.net for latency and how it will perform in real life.

For those of you who haven’t used this service lately, they made changes to the algorithm to show more accurate results. Downloading files, you still get 70KB/s on my informal tests. So a 350 MB episode of TV series would theoretically take 1 h 30 min.

The speedtest is self-explanatory.

The pingtest tests your line quality, and how well it actually performs in real life. VOIP, gaming, and responsiveness is usually measured in milliseconds (a thousandths of a second). It is quite easy to interprete the results. A ping of 91ms locally means that is the amount of time to a local server; the first “hop”! Streamyx, this would be 20 ms (4 times faster). The explanation for this would be Streamyx is wired, and wireless has some overhead due to equipment inefficiencies and background noise.

Note for San Fran (345ms) and Manchester (400ms). Now that is the real benchmark of how good your ISP is. These numbers can be chopped in half if DiGi is willing to make sacrifises and investments in QUALITY bandwidth. Peering with reputable companies like Cogent, Level 3, Global Crossing and paying premium dollar for higher priority, you can see numbers improving drastically.

I will show you this effect in Singapore, where ping to the same servers *should be* twice as fast. Reason being, Singtel / Starhub is willing to pay more for higher priority premium bandwidth. Hence the better latencies.

One real world example why you want this “ping” to be fast is, when loading facebook pictures. When you click “next” when viewing an album, you EXPECT things to just load instantly. If your ISP is able to do that, then it is paying premium bandwidth.

One note: If you call up your Customer service hotline for your ISP, and tell them your speed is slow. And THEIR excuse is “The server you are connected to is slow”. You can call that bull****. All major servers are capable of very very fast speeds, about 500kB/s per user. Even philipkhor.com running on a cheap server can do that speed if your ISP supports it. So it is utter nonsense to put the blame on the “internet” or “internet congestion!”. (ahem, Streamyx, TM Nut..)

If there is one thing you can lobby for, it is latency (bring it down). Not raw speed (higher speeds).

Further reading:

http://en.wikipedia.org/wiki/Latency_%28engineering%29

http://compnetworking.about.com/od/speedtests/a/network_latency.htm

http://rescomp.stanford.edu/~cheshire/rants/Latency.html

p.s Follow me on Twitter @philipkhor for the next 3 days as I venture to Singapore!

*** Singapore test update ***

This is a test to San Francisco test server via Singnet.

Pinging sfo.speakeasy.net [64.81.79.2] with 32 bytes of data:

Reply from 64.81.79.2: bytes=32 time=238ms TTL=48
Reply from 64.81.79.2: bytes=32 time=282ms TTL=48
Reply from 64.81.79.2: bytes=32 time=245ms TTL=48
Reply from 64.81.79.2: bytes=32 time=262ms TTL=46

Approximate round trip times in milli-seconds:
Minimum = 238ms, Maximum = 282ms, Average = 256ms

We need latency of our broadband to be THIS LOW! Come on, Malaysia!

Here is some more follow up Trace Routes for your amusement. Warning: Geek Level 3 and above required to comprehend this information.

From Jaring

A. To FIOS (fiber optic) subscriber in Texas, USA
traceroute to 173.74.131.10 (173.74.131.10), 30 hops max, 40 byte packets

 1  202.190.180.194 (202.190.180.194)  1.624 ms  1.591 ms  1.580 ms

[equipment inefficiencies]

 3  ge5-0.bkj3.jaring.my (161.142.173.1)  0.675 ms  0.672 ms  0.662 ms

 4  srp1-0-0.tlg90.jaring.my (61.6.2.193)  187.270 ms  187.280 ms  187.270 ms

 5  if-5-2-0.mcore3.LAA-LosAngeles.as6453.net (216.6.61.17)  188.752 ms  188.743 ms  188.732 ms

To Cross the Pacific Ocean? 187 milliseconds! That’s faster than a blink of an eye. 2/3 the speed of light.

 8  0.so-5-2-0.LAX01-BB-RTR2.verizon-gni.net (152.63.112.50)  188.787 ms  188.782 ms  188.769 ms

From West Coast USA to Eastern United States = 45 ms

9  so-4-0-0-0.DFW01-BB-RTR2.verizon-gni.net (130.81.17.27)  229.781 ms  229.595 ms  229.352 ms

B. To popular social networking site
traceroute to facebook.com (69.63.187.19), 30 hops max, 40 byte packets

 1  202.190.180.194 (202.190.180.194)  11.539 ms  11.533 ms  12.804 ms

 2  te0-1-0-0.bkj95.jaring.my (61.6.10.14)  19.211 ms  19.204 ms  19.194 ms

 3  JRC-0003.GW3.KUL1.asianetcom.net (203.192.163.85)  12.689 ms  12.690 ms  12.682 ms

 4  po12-1.gw5.sin1.asianetcom.net (202.147.48.218)  17.469 ms  17.467 ms  17.754 ms

Funny, this time all Facebook traffic is routed through Singapore and Narita, Japan.

 5  po1-1-0.cr1.nrt1.asianetcom.net (202.147.48.118)  101.992 ms  102.004 ms  102.000 ms

 6  po7-0-0.gw1.sjc1.asianetcom.net (202.147.0.34)  205.709 ms  194.866 ms  194.815 ms

Data packets reach San Jose, California a bit slower at 200 ms. But still freaking fast, la :)

 7  sv1.br01.sjc1.tfbnw.net (206.223.116.166)  194.500 ms  194.496 ms  194.488 ms

 8  ae0.bb01.sjc1.tfbnw.net (74.119.76.21)  195.641 ms  195.624 ms  195.623 ms

 9  xe-7-1-0.bb01.iad2.tfbnw.net (74.119.76.171)  266.307 ms  266.036 ms  267.085 ms

This is the bottleneck, where the peering arrangement is of lower priority. Hence the 70ms additional lag.

266 ms is what Jaring can promise it’s customers. Look at my Digi example above. It is almost 50% slower at 380 ms.

10  xe-11-3-0.br01.ash1.tfbnw.net (204.15.21.147)  266.240 ms  266.250 ms  266.239 ms

C. San Francisco is a Major landing point, where Data hotels are located.

traceroute to sfo.speakeasy.net (64.81.79.2), 30 hops max, 40 byte packets

 1  202.190.180.194 (202.190.180.194)  65.938 ms  65.912 ms  65.904 ms

 2  te0-1-0-0.bkj95.jaring.my (61.6.10.14)  9.851 ms  9.844 ms  9.834 ms

 3  ge1-0.bkj2.jaring.my (61.6.162.1)  0.759 ms  0.756 ms  0.749 ms

See the jump? That, my friends, is the Pacific Ocean crossing.

 4  srp1-0-0.tlg90.jaring.my (61.6.2.193)  187.468 ms  187.470 ms  187.462 ms

 5  if-5-2-0.mcore3.LAA-LosAngeles.as6453.net (216.6.61.17)  188.762 ms  188.751 ms  188.742 ms

 6  if-4-0-0-999.mcore5.LAA-LosAngeles.as6453.net (209.58.33.21)  188.319 ms  188.115 ms  188.104 ms

 7  if-15-0-0-933.mcore3.PDI-PaloAlto.as6453.net (216.6.29.85)  199.907 ms  200.347 ms  200.337 ms

 8  ix-13-0-9-143.mcore3.PDI-PaloAlto.as6453.net (216.6.29.14)  196.914 ms  199.342 ms  211.379 ms

 9  dns.sfo1.speakeasy.net (64.81.79.2)  197.385 ms  212.171 ms  197.457 ms

How about the United Kingdom?

traceroute to www.royal.gov.uk (89.151.94.50), 30 hops max, 40 byte packets

 1  202.190.180.194 (202.190.180.194)  1.205 ms  1.174 ms  1.164 ms

 2  te0-1-0-0.bkj95.jaring.my (61.6.10.14)  8.884 ms  8.878 ms  8.868 ms

 3  ge5-0.bkj3.jaring.my (161.142.173.1)  0.823 ms  0.818 ms  0.809 ms

 4  srp1-0-0.tlg90.jaring.my (61.6.2.193)  187.358 ms  187.358 ms  187.350 ms

Apparently, going Eastwards to the America is more efficient than pulling lines through russia, iraq and central Europe. It really is.

 5  if-5-2-0.mcore3.LAA-LosAngeles.as6453.net (216.6.61.17)  189.018 ms  189.010 ms  188.999 ms

 6  Vlan10.icore1.LAA-LosAngeles.as6453.net (209.58.85.90)  188.992 ms  190.161 ms  190.199 ms

 7  4.68.63.65 (4.68.63.65)  193.883 ms  194.649 ms  194.708 ms

 8  ae-73-70.ebr3.LosAngeles1.Level3.net (4.69.144.116)  207.164 ms ae-63-60.ebr3.LosAngeles1.Level3.net (4.69.144.52)  205.598 ms ae-73-70.ebr3.LosAngeles1.Level3.net (4.69.144.116)  207.145 ms

 9  ae-2-2.ebr3.SanJose1.Level3.net (4.69.132.9)  202.397 ms  206.533 ms  204.383 ms

10  ae-73-73.csw2.SanJose1.Level3.net (4.69.134.230)  199.205 ms ae-83-83.csw3.SanJose1.Level3.net (4.69.134.234)  195.447 ms ae-93-93.csw4.SanJose1.Level3.net (4.69.134.238)  205.232 ms

11  ae-64-64.ebr4.SanJose1.Level3.net (4.69.134.241)  195.452 ms ae-74-74.ebr4.SanJose1.Level3.net (4.69.134.245)  196.592 ms ae-84-84.ebr4.SanJose1.Level3.net (4.69.134.249)  195.430 ms

This hop is the most facinating. It is the best indicator of cross country latency. This case, it shows 80 ms from West coast to East Coast. It is busy rush hour now, so traffic is slightly slower.

12  ae-2-2.ebr4.NewYork1.Level3.net (4.69.135.186)  264.405 ms  276.635 ms  276.612 ms

13  ae-84-84.csw3.NewYork1.Level3.net (4.69.134.122)  262.776 ms ae-94-94.csw4.NewYork1.Level3.net (4.69.134.126)  270.467 ms ae-64-64.csw1.NewYork1.Level3.net (4.69.134.114)  275.530 ms

14  ae-81-81.ebr1.NewYork1.Level3.net (4.69.134.73)  265.242 ms ae-91-91.ebr1.NewYork1.Level3.net (4.69.134.77)  264.831 ms ae-61-61.ebr1.NewYork1.Level3.net (4.69.134.65)  265.214 ms

This two show the crossing of the Atlantic Ocean, about 70 ms. Not too shabby.

15  ae-42-42.ebr2.London1.Level3.net (4.69.137.69)  333.109 ms ae-44-44.ebr2.London1.Level3.net (4.69.137.77)  334.754 ms  335.724 ms

16  ae-25-52.car5.London1.Level3.net (4.69.139.102)  334.733 ms  333.674 ms  344.556 ms

17  195.50.112.18 (195.50.112.18)  338.619 ms  338.583 ms  338.552 ms

18  * ge-0-3.cadogan2.rtr.dedipower.net (81.29.95.234)  340.242 ms *

Cogent is a low cost bandwidth provider. Porn sites use them cuz it can ship Terabytes of Data for 1/2 the cost of more established Tier 1 networks like Level 3. The ping is slower, but Rapid Share and file sharing sites are more than happy to co-locate their servers with them.

Trace done using Cogent.

A. Tracing the route to philipkhor.com (202.190.176.100)

  1 gi1-47.224.mpd01.lax01.atlas.cogentco.com (66.250.4.5) 4 msec 4 msec 4 msec

They’re using REALLY new equipment since they’ve only been established for under 5 years. So first few hops are awesome.

  5 ix-5-2-0.mcore3.LAA-LosAngeles.as6453.net (216.6.61.18) 4 msec 4 msec 0 msec

This tells the WHOLE story. Within 6 hops you’re in Malaysia, specifically, Jaring Data Center in Bukit Jalil! Amazing, huh? <200 ms is being generous already. I’m sure the link is very heavy usage now cuz everyone loves Cogent here in Asia. Jaring pun ada direk line!

  6 srp3-0.bkj3.jaring.my (61.6.2.194) 196 msec 196 msec 196 msec

  7 ge0-2-1-0.bkj95.jaring.my (61.6.162.195) 200 msec 204 msec 208 msec

  8 te3-0-0.bkj2.jaring.my (61.6.10.22) 200 msec 196 msec 196 msec

B. Tracing the route to google.com (74.125.19.104)

  1 gi1-47.224.mpd01.lax01.atlas.cogentco.com (66.250.4.5) 4 msec 4 msec 0 msec

  2 te7-8.mpd01.lax05.atlas.cogentco.com (154.54.30.190) 0 msec 0 msec 4 msec

  3 google.lax05.atlas.cogentco.com (154.54.13.106) 0 msec 0 msec 4 msec

  4 216.239.46.182 0 msec

Very very fast indeed. Faster than a lot of LANs!

  7 google.com (74.125.19.104) 12 msec 12 msec 16 msec

C. Tracing the route to msn.com (207.68.172.246)

  1 gi1-47.224.mpd01.lax01.atlas.cogentco.com (66.250.4.5) 4 msec 4 msec 0 msec

  2 te7-3.mpd01.sjc01.atlas.cogentco.com (154.54.2.130) 16 msec 16 msec 16 msec

Merely equipment inefficiencies. MSN traffic = fast.

  5 ge-7-3-0-57.sjc-64cb-1b.ntwk.msn.net (207.46.47.249) 12 msec 16 msec 16 msec

  6 ge-7-2-0-0.pao-64cb-1b.ntwk.msn.net (207.46.47.40) 16 msec 20 msec 16 msec

D.Tracing the route to facebook.com (69.63.187.19)

  1 gi1-47.224.mpd01.lax01.atlas.cogentco.com (66.250.4.5) 0 msec 0 msec 4 msec

  2 te8-3.mpd01.lax05.atlas.cogentco.com (154.54.0.214) 0 msec 0 msec 4 msec

  3 Tenge6-3-10000M.ar3.LAX2.gblx.net (64.215.195.153) 0 msec 4 msec 0 msec

  4 FACEBOOK-INC.TenGigabitEthernet6-2.ar1.PAO2.gblx.net (67.17.162.38) 16 msec 12 msec 16 msec

  5 ae0.bb01.pao1.tfbnw.net (74.119.76.132) 40 msec 16 msec 12 msec

nuff said

This last one is to:
202.190.176.100 [philipkhor.com]

Hop

(ms)

(ms)

(ms)

IP Address

Host name

1

20

19

8

72.249.128.105

-

2

17

11

13

64.129.174.181

64-129-174-181.static.twtelecom.net

3

10

20

34

4.69.145.52

ae-63-60.ebr3.dallas1.level3.net

4

16

51

7

206.82.142.10

if-1-0-0-32.core2.dtx-dallas.as6453.net

5

103

41

44

216.6.84.69

if-10-0-0-1703.mcore3.laa-losangeles.as6453.net

6

51

59

65

216.6.61.18

ix-5-2-0.mcore3.laa-losangeles.as6453.net

7

164

166

261

61.6.2.194

srp3-0.bkj3.jaring.my

8

237

249

243

202.147.32.89

po13-0.gw2.sin1.asianetcom.net

9

260

256

233

61.6.10.22

te3-0-0.bkj2.jaring.my

10

243

237

231

202.190.176.100

silver.centralmalaysia.com

Jaring is really really good. They’re smart. If you target home broadband market, you got stiff competition from wireless broadband providers. They become the data center. Where companies and businesses run their servers on. So if you wanna ship big chunks of content worldwide, Jaring datacenter ftw. It is relatively cheap and trouble free too. TM’s datacenter is huge, good local connectivity. But internationally, suffer a bit due to the hungry Bit Torrent dudes here in MY.

Comments? E-mail me pip@thepippo.com

Related posts:

1. I’m on the 27″ iMac What the heck. What the heck. What the heck. The 27″ iMac on the new Taylor’s lakeside labs? That’s freaking insane. The business IT guys are here looking now. But...
2. Digi Telecommunication HSDPA (3.5G) One word: awsm. 1. Latency Tracing route to philipkhor.com [202.190.176.100] over a maximum of 30 hops: 1   212 ms   189 ms    97 ms  172.18.44.14 2   127 ms    86 ms   659...
3. Give credit where it is due 5 things. 1. At 6am, I woke up and checked The Internetz for my usual morning fix. Alas, my site was down. Called up Easynet Interactive and they gently informed...