Update on Fiber Optic test

ksl · Post by **ksl** » Thu, 15 Dec 2011 6:05 pm

http://speedtest.net/result/1649824136.png download 53.4 mbps upload 25.11

curiousgeorge · Post by **curiousgeorge** » Fri, 16 Dec 2011 8:51 am

Can you do a ping test please to servers in HK/UK/NYC/LA? I am really interested in lag to overseas servers results.

ksl · Post by **ksl** » Sat, 17 Dec 2011 10:23 am

curiousgeorge · Post by **curiousgeorge** » Sat, 17 Dec 2011 4:37 pm

Very interesting comparison...most significant is the Packet Loss result from Starhub Vs your Singtel Fibre.

Although Pings are close enough to be insignificant (the feel of 'responsiveness' of our respective connections) the increased Packet Loss & Jitter will mean that I experience more "stuck" pages than you.

I would love to see results from someone else on Starhub Broadband (not fibre) to see if the packet loss is across our providor or down to the local loop (connection from PC to local switchboard).

However - if the packetloss/jitter wasn't an issue, then it does demonstrate that international speeds are not going to improve for the moajority of users.

For instance, if you are a gamer and need a good connection to overseas servers, it doesn't matter which connection you go for...fibre wouldnt actually help any.

Post by **x9200** » Sat, 17 Dec 2011 5:37 pm

Interesting. One of my web servers is hosted in Dallas... I just ran the test from my eeepc, 4 walls and 10m from the router and the Dallas server even did not show on the list. At the same time this is what I get doing the test myself to "my" server in Dallas:

100 packets transmitted, 100 received, 0% packet loss, time 103388ms
rtt min/avg/max/mdev = 226.813/254.990/1788.849/163.949 ms, pipe 2

and your results show almost 80% loss to that location. I am with starhab, very basic plan.

ksl · Post by **ksl** » Mon, 19 Dec 2011 2:47 am

x9200 wrote:Interesting. One of my web servers is hosted in Dallas... I just ran the test from my eeepc, 4 walls and 10m from the router and the Dallas server even did not show on the list. At the same time this is what I get doing the test myself to "my" server in Dallas:

100 packets transmitted, 100 received, 0% packet loss, time 103388ms
rtt min/avg/max/mdev = 226.813/254.990/1788.849/163.949 ms, pipe 2

and your results show almost 80% loss to that location. I am with starhab, very basic plan.

To Curious George, my guess is that your ISP is not emptying the Master cache as often as they should. Empty your own cache first. Then you can try to empty the DNS cache from the DOS prompt with " ipconfig/flushdns ". Then do another ping test

Post by **sundaymorningstaple** » Mon, 19 Dec 2011 6:56 am

ksl, I'm curious. Will that give you a faster result? If yes, then it's might be considered a false positive yeah?

I mean, if you base your analysis on the fastest results being made possible by all types of cleaning, then each time you go to another page, you should clean out your cache, correct? I would think you should only do speed tests in the normal operating conditions and not some lightened stripped down vehicle that is solely to be used for time trials and not general surfing.

Of course, if you are looking for bragging rights or promotional sales materials like Starhub and Singtel, then the false positives are great but not practicable, at least from my POV.

curiousgeorge · Post by **curiousgeorge** » Mon, 19 Dec 2011 8:09 am

Well I did a bit more investigating and research, thanks to guys over on HarwareZone.com.sg.

When I signed up for 50Mb/s Maxonline last year, they gave me a free D-Link DCM-604, which is a combination modem/router so I plugged it in and have been mouring a crappy connection ever since.

So, the DCM640 is a docsis 2.0 modem which means it will never get more than 32Mbps download speed, no matter how fast the connection! Currently, the only starhub intergrated modem/router that is DOCSIS 3.0 (speeds up to 100Mbps) is the Cisco DPC3925. But Starhub didn't have in stock. Weird that they sell you a fast plan and give you a modem than can't handle it!
(The used to give away Motorold SB6100 and SB6120, and they are both DOCSIS 3.0)

Yesterday I went out and spend $$$ on an Asus RT-N56U which is a true gigabit router with incredible wireless range. (Other alternative suggested was EnGenius ESR9850 but I liked the look of the N56).

So I switched the DCM640 to bridge mode (i.e. only functions as a modem) and plugged in the new router.

OH MY GOD. Even though I am still not maxing out anywhere near the 50Mbps of my plan, the difference is night and day. As a gigabit router, the speeds across my home network have significantly increased, and the handling of all that traffic has also had a very positive effect on my internet connection:

As you can see, speed has increased A LOT, ping has improved A LOT and NO DROPPED PACKETS now

I will keep my eye out for a decent DOCSIS 3.0 modem now, and will expect to increase speeds in Singapore further. But, bang-for-buck, I am happy with my connection now, especially as it is half the price of ksl's fibre connection

Two more things I learnt:
1) The routers I mentioned above will work for anyone, fibre or not. Both are gigabit routers with much better wireless than anything SH/ST/M1 will throw at you. Just use the telco eqpt as modem and use one of these puppies as the router for best results. FIve bars wireless coverage across my whole apartment (and most of River Valley)

2) Tests outside of Singapore are kinda pointless, as we are all limited by the same undersea optic cable to around 18Mbps. Anything faster in Singapore is great, but those gains won't necessarily carry over to gaming/downloads/streaming from overseas sites if your current connection is already maxed out.

Post by **x9200** » Mon, 19 Dec 2011 9:10 am

Actually I just checked what plan I am on and seems to be the same as yours (up to 50Mbs). Yes, router does matter but with this plan whether it is 1000 or 100 is secondary - I am on a modified cisco/linksys 100BASE router only and have similar speedtest results as yours. I stopped using dlink routers over 6y ago mostly for stability reasons.
This is what I get (practical results dependent on the sites I use, not the test)
- internal net (is 1000BASE on a cheap planex switch and dlink cards, MTU=9000*, 6 or 5e class cabling: 120-160Mbs
- from SG: up to 20Mbs
- from outside: up to 2-3Mbs
- upload: typically 1Mbs

*) this is very important - your hardware and software has to support so called jumbo frames

ksl · Post by **ksl** » Mon, 19 Dec 2011 6:35 pm

sundaymorningstaple wrote:ksl, I'm curious. Will that give you a faster result? If yes, then it's might be considered a false positive yeah?

I mean, if you base your analysis on the fastest results being made possible by all types of cleaning, then each time you go to another page, you should clean out your cache, correct? I would think you should only do speed tests in the normal operating conditions and not some lightened stripped down vehicle that is solely to be used for time trials and not general surfing.

Of course, if you are looking for bragging rights or promotional sales materials like Starhub and Singtel, then the false positives are great but not practicable, at least from my POV.

Yes you are quite right though curious George if he is a heavy user, will clog his pipe work before the ISP's empty their cache. I noticed that I suffered the same problem, though 12 to 14 hrs a day surfing is well above the average user usage, so when packets are dropped, it maybe a blockage only he is having or the quality and/or setup of the router he's using I had an out of date router before I upgraded to optic fiber. The scenario is the two year contract packages on broadband are normally packaged with inferior or outdated hardware.

aster · Post by **aster** » Mon, 02 Jan 2012 10:03 pm

Such high packet loss should never happen. In the EU there are huge penalties for ISPs when that happens as they are bound to provide you a connection with minimal loss, "set in stone" by the regulators.

curiousgeorge · Post by **curiousgeorge** » Tue, 03 Jan 2012 8:34 am

Well after a bit of fiddling, researching and talking to Starhub I am now doing a bit better:

Even speeds overseas have vastly improved:

That's Maxonline Ultimate 100Mbps (which is 30Mbps International Bandwidth...I think even fiber is capped at 15Mbps unless you go for their top plan).
This isn't going to work for people who need fast upload speeds of course (i.e. sending big files) but suits my needs.

The packet loss issue was with the D-Link 604 router. Go to the router's settings page and disable detection of IP Flooding. Suddenly, packet loss issues it gone

Switching it to bridge mode also had same effect.

Oh, and now running a DOCSIS 3.0 modem, which can handle the d/l speeds...I still can't understand why Starhub will sell you a 50Mbps plan and then give you a modem only capable of 32Mbps...

zzm9980 · Post by **zzm9980** » Tue, 03 Jan 2012 9:30 am

Good to see the packet loss was just local DoS protection. Those numbers worried me. From experience I know 3-5% packet loss is enough to kill video chats/voip and a lot of online games

SINGAPORE EXPATS FORUM
Singapore Expat Forum and Message Board for Expats in Singapore & Expatriates Relocating to Singapore

Update on Fiber Optic test

Update on Fiber Optic test

Who is online