https://atmarkit.itmedia.co.jp/ait/articles/2302/17/news007_2.html#13
https://atmarkit.itmedia.co.jp/ait/articles/2302/17/news007_2.html#13
Inside-Out
(Page 2 of 2)
February 17, 2023 05:00 PM Published.
[Takashi Taniguchi, IIJ]
Notification
View
Share
3
Previous page 1|2
We also investigated connectivity to IIJ
Next, let's run a traceroute to IIJ. The traceroute is up to "www.iij.ad.jp".
First from Japan...hmmm, not so good...not necessarily a problem with IIJ, but something to consider in the future.
$ traceroute -A -I www.iij.ad.jp
traceroute to www.iij.ad.jp (202.232.2.180), 30 hops max, 60 byte packets
1 StarlinkRouter.lan (192.168.1.1) [*] 0.624 ms 0.833 ms 1.104 ms
2 100.64.0.1 (100.64.0.1) [*] 54.028 ms 54.020 ms 54.104 ms
3 172.16.249.6 (172.16.249.6) [*] 53.998 ms 53.989 ms 54.065 ms
4 149.19.109.16 (149.19.109.16) [AS14593] 53.961 ms 53.952 ms 54.004 ms
5 * * *
6 be3093.rcr51.b060372-1.tyo01.atlas.cogentco.com (154.24.11.137) [*] 54.002 ms 31.541 ms 31.376 ms
7 be3074.ccr71.tyo01.atlas.cogentco.com (154.54.2.177) [AS174] 55.393 ms 55.360 ms 55.387 ms
8 be3696.ccr41.sjc03.atlas.cogentco.com (154.54.86.137) [AS174] 157.893 ms 157.863 ms 157.954 ms
9 be3142.ccr21.sjc01.atlas.cogentco.com (154.54.1.193) [AS174] 157.779 ms 157.757 ms 157.821 ms
10 be3176.ccr41.lax01.atlas.cogentco.com (154.54.31.189) [AS174] 157.973 ms 158.006 ms 158.102 ms
11 be3243.ccr41.lax05.atlas.cogentco.com (154.54.27.118) [AS174] 157.579 ms 157.570 ms 157.610 ms
12 38.104.85.34 (38.104.85.34) [AS174] 134.310 ms 135.534 ms 143.392 ms
13 osk004bb00.IIJ.Net (58.138.88.113) [AS2497] 151.414 ms 143.319 ms 147.114 ms
14 osk008agr02.iij.net (210.130.143.174) [AS2497] 151.173 ms 151.147 ms 155.079 ms
15 www.iij.ad.jp (202.232.2.180) [AS2497] 155.048 ms 155.075 ms 155.188 ms
I checked the connection from Starlink to IIJ in Japan.
US seems to be fine.
$ traceroute -I -A www.iij.ad.jp
traceroute to www.iij.ad.jp (202.232.2.180), 30 hops max, 60 byte packets
1 192.168.1.1 (192.168.1.1) [*] 7.040 ms 7.100 ms 7.165 ms
2 100.64.0.1 (100.64.0.1) [*] 57.382 ms 75.170 ms 75.802 ms
3 172.16.248.36 (172.16.248.36) [*] 75.350 ms 75.487 ms 76.019 ms
4 149.19.108.27 (149.19.108.27) [AS14593] 75.025 ms 75.371 ms 76.057 ms
5 76.74.57.134 (76.74.57.134) [AS6517] 75.868 ms 75.869 ms 75.868 ms
6 ae28.cr2-sjc1.ip4.gtt.net (89.149.136.149) [AS3257] 77.105 ms 56.902 ms 55.527 ms
7 as2497.cr2-sjc1.ip4.gtt.net (173.241.128.62) [*] 65.807 ms 66.413 ms 66.549 ms
8 osk004bb00.IIJ.Net (58.138.88.185) [AS2497] 179.289 ms 179.889 ms 180.189 ms
9 osk008agr02.iij.net (210.130.143.205) [AS2497] 180.181 ms 180.178 ms 180.582 ms
10 www.iij.ad.jp (202.232.2.180) [AS2497] 180.574 ms 180.570 ms 180.566 ms
I checked the connection from Starlink to IIJ in the US.
DE is via US, but this is not a problem either. It is a matter of whether it goes around Eurasia or the Americas.
$ traceroute -I -A www.iij.ad.jp
traceroute to www.iij.ad.jp (202.232.2.180), 30 hops max, 60 byte packets
1 StarlinkRouter.lan (192.168.1.1) [*] 0.787 ms 0.892 ms 1.095 ms
2 100.64.0.1 (100.64.0.1) [*] 49.182 ms 53.034 ms 52.057 ms
3 172.16.248.40 (172.16.248.40) [*] 51.907 ms 52.050 ms 52.234 ms
4 149.19.108.137 (149.19.108.137) [AS14593] 51.753 ms 51.955 ms 52.099 ms
5 ffm-b5-link.ip.twelve99.net (62.115.37.20) [AS1299] 51.982 ms 52.180 ms 52.073 ms
6 ffm-bb1-link.ip.twelve99.net (62.115.124.116) [AS1299] 51.069 ms 46.994 ms 58.790 ms
7 prs-bb1-link.ip.twelve99.net (62.115.123.13) [AS1299] 66.711 ms 55.301 ms *
8 ldn-bb1-link.ip.twelve99.net (62.115.135.24) [AS1299] 71.006 ms 71.125 ms 71.274 ms
9 ldn-b1-link.ip.twelve99.net (62.115.121.27) [AS1299] 70.824 ms 70.952 ms 71.095 ms
10 202.232.1.61 (202.232.1.61) [AS2497] 70.648 ms 70.766 ms 70.902 ms
11 lax002bb12.IIJ.Net (58.138.83.186) [AS2497] 202.563 ms 202.361 ms 198.955 ms
12 osk004bb00.IIJ.Net (58.138.88.113) [AS2497] 314.642 ms 314.728 ms 307.116 ms
13 osk008agr02.IIJ.Net (210.130.142.26) [AS2497] 312.090 ms 312.277 ms 311.955 ms
14 www.iij.ad.jp (202.232.2.180) [AS2497] 310.908 ms 310.843 ms 311.026 ms
I checked the connection from Starlink to IIJ in Germany.
Obtaining Statistics
Starlink also collects information on the time the satellite link is lost. The collected information can be retrieved via Starlink's smartphone application. The same screen can also be viewed by accessing Dish's IP address "192.168.100.1" using a browser.
http://192.168.100.1/にアクセスした画面 with a web browser
http://192.168.100.1/にアクセスした画面 with a web browser
When looking at this, short stoppages of 0.1 seconds to a few seconds are rather common. The reasons are categorized into three types.
NO SIGNAL RECEIVED (no signal)
NETWORK ISSUE (switching satellites)
POSSIBLY OBSTRUCTED (obstacle detection)
These momentary interruptions are not fatal, but they can be absorbed in operation.
Management screen on smartphone app
Management screen on a smartphone application
Top left: UPTIME records momentary interruptions. LATENCY records delays, and USAGE records the speed at which communications occur.
Top center: When a momentary disconnection is recorded, the time, duration, and reason for the disconnection are recorded. Upper right: Particularly long momentary disconnections (2 seconds or longer) can be extracted and checked.
Lower left: Communication disconnections of 5 seconds or longer can be checked.
Middle left: Visibility of the surrounding area that can be observed from Dish can be checked.
Japanese localization of the application
The Starlink application is now available in Japanese (confirmed on December 6, 2022). I think I understand a little better just because the reason for the shutdown is in Japanese. If there is no suspension, the message "Keep it up, Kanbare" will be displayed.
Japanese-language Starlink smartphone application screen
Japanese-language Starlink smartphone app screen
Left: The top screen of the application. If you are logged in, you can remotely check the status of the Starlink you are managing.
Center: Instantaneous disconnections of 0.1 second or longer are recorded.
Right: You can check the momentary disconnection. On this day, there was no momentary breakdown of more than 2 seconds.
About the installation environment of STARLINK
We ask for an installation site within the range where our colleagues can work without difficulty. In Japan, we rented an area where a large parking lot under IIJ management was available. In the U.S. and Germany, we installed the equipment on the balcony of the apartment where the employees live.
The visibility report confirms that Japan is close to the best environment. Below, the light blue area represents the "Clear View," or the area free of obstructions (that would interfere with communication), and the light red area represents the "Obstructions," or the area where obstructions were detected.
Visibility Report Results from Starlink's Smartphone App (Japan)
Visibility Report Results from Starlink's Smartphone App (Japan)
Germany has more jagged edges compared to Japan, but they do not seem to be recognized as obstacles. Roofs of buildings and trees seem to be seen as jagged, and it is as if we are looking at the sky with some kind of fisheye lens.
Results of visibility report by Starlink smartphone application (Germany)
Visibility Report Results from Starlink's Smartphone App (Germany)
In the U.S., they are installed on top of the outdoor unit of an air conditioner to get as much visibility as possible, but it is partially blocked by an obstacle. They have a building next to their apartment.
Results of a visibility report by Starlink's smartphone application (USA)
Results of a visibility report from Starlink's smartphone app (U.S.)
Because of this situation, it is expected that the order of advantage in terms of quality will be "Japan -> Germany -> USA". Also, Japan and Germany have G1 antennas, while the U.S. has G2 antennas, so there may be a difference there as well.
How to measure line quality
I wondered what kind of tool I should use to measure line quality, but I decided to use "SmokePing" because I thought it would be better if I could measure latency instead of bandwidth-consuming communication speed.
SmokePing is a tool that uses ping to accumulate latency information and create a fuzzy-looking graph. When this fuzziness becomes more noticeable or turns black, you can intuitively sense that the quality has deteriorated. SmokePing has many other uses, so if you are interested, please search and check it out.
Where were the measurements taken?
At first, we thought that the IP address "100.64.0.1" of the network equipment in SpaceX and the ground station gateway would be good.
However, when we actually measured it, we encountered packet loss and other phenomena that prevented us from measuring properly. Although we did not follow it deeply, it seems that ICMP for network devices is not an appropriate target for measurement, since it is sometimes restricted by QoS in consideration of device stability.
Therefore, we decided to use the IP address "8.8.8.8.8" of Google's public DNS service, which we have used many times in the past.
What will we be measuring?
The following are the results of a one-hour measurement on Starlink in Japan, where each measurement consists of 20 ICMP echoes (i.e., pings) to "8.8.8.8". The average delay and variation are plotted on the vertical axis, color-coded according to the number of packet losses in the 20 pings.
Results of 1 hour measurement by SmokePing
One-hour results from SmokePing
This is the ping response of Google's public DNS servers, and assuming that the quality of Google's servers is constant, I thought that this would be an indirect measurement of the quality of the line to the servers. This measurement shows that the performance is between 30ms and 40ms, which is in line with Starlink's specifications.
About the measurement results
Next, during the year-end and New Year holidays (10 days from December 29, 2022 to January 7, 2023), the following are the results of measurements taken against "8.8.8.8" from each location.
Japan (Tokyo), but appears to be the most stable. The worst latency value is also kept at around 60ms. There are some areas where the color is a little uneven, but this may be due to the timing when Speedtest was programmed to be performed intermittently.
Measurement result by SmokePing (Japan)
Measurement result by SmokePing (Japan)
Next is the United States (San Jose). As shown in the visibility report, the measurement results show that the environment is quite unfavorable. Packet loss is also observed, especially around January 5, 2023, and appears to be getting worse.
SmokePing measurement results (U.S.)
SmokePing results (U.S.)
Germany (Dusseldorf) is stable. However, the worst latency seems to be slightly worse than in Tokyo.
Measurement result by SmokePing (Germany)
SmokePing measurement results (Germany)
Consideration of fluctuation by long-term measurement
Some measurements have been taken since the summer of 2022. The measurements were taken in San Jose, U.S.A., for about four months. There was a period of time when measurements stopped. This data should be considered as a reference only, since the measurement conditions are not the same.
This may be a convenient interpretation, but it appears that quality has repeatedly deteriorated and improved. I thought that the quality might be maintained by the infrastructure enhancement by adding satellites every week in the U.S., where the customer increase is expected to be the largest.
Results of 4-month long term measurement
Results of 4-month long term measurement
As the satellite constellation grows, the overall bandwidth available to the constellation grows as well, with SpaceX launching a rocket every week and deploying 55 Starlink satellites. I can't imagine any operator being able to keep up with this momentum.
I hear that in 2023, a new satellite will be launched that will support direct communication with smartphones and will have greatly enhanced performance. It also seems that satellites will be connected to each other by laser links without the need for a ground station. It will be interesting to see how this technology can be used with equipment that costs about the same as a smartphone, and with communication fees that are a little higher than those of mobile carriers, but without any strange restrictions.
0 コメント:
コメントを投稿