dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
3555

gwalk
Premium Member
join:2005-07-27
West Mich.

gwalk to justin

Premium Member

to justin

Re: speed tests over satellite

Click for full size
I just ran one again. Ping is messed up and when trying to "share" I get a "something went wrong" error.

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

1 edit

justin

Mod

thanks,
that little error message below the Test Again button:
unable to post results - timeout

means that it timed out trying to do the POST to www.dslreports.com, of the results for the test.

The timeout is 10 whole seconds. Is that unreasonably low for POST over satellite ? how long do you wait when you post in the forum something?
edit: I imagine what is happening is at the end of the upload phase, after the test cancels, the satellite equipment is still full of stuff in transit and in buffers and that delays the next operation causing the result not to POST.
DrStrangLov
join:2012-03-28

1 edit

DrStrangLov to justin

Member

to justin
said by justin:

Does Satellite still compress uploads if they can be compressed?

Exede 5/12 users use AcceleNet servers ( »AcceleNet ), which are located in Albuquerque, Dallas, Denver, Fort Worth, Salt Lake City, Phoenix, and Seattle. All Exede users must use one of these AcceleNet servers. Wildblue users have a different setup, apples and oranges between them.

Common upload speed for most Exede users is around 3 mbps, and your upload test appears to be inline with Testmy and Xfinity ( Denver Server: 68.87.66.23 ). AT&T's Internet Speed Test, »www.att.com/speedtest/ , tends to show a higher upload speed, and higher download speeds: »Current Status , Jan 13, 2015 .

My POP is located in Denver, but my Gateway is located in Riverside, CA, so my ping times will range in 700+ ms range.

As can be seen in first post here, »Current Status , March 29, 2015 , the first test at 03:49PM shows a ping of 787ms, and but second test, 07:57PM, shows a ping of 915ms (at Ohio server). But, looking at Denver's Xfinity ping shows 732 ms during this timeframe.

Again, I'm using the Denver POP, so I was pining to an Xfinity server in Denver.

Your download speeds appears to be too high, as based upon Testmy and Xfinity's test during same primetime time frame. »Current Status , March 29, 2015
DrStrangLov

DrStrangLov to justin

Member

to justin
said by justin:

Does Satellite still compress uploads if they can be compressed?

Data Compressors

Bi-directional compression: uses data from a download to compress a subsequent upload and vice versa

»www.viasat.com/accelenet ··· chnology

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin to DrStrangLov

Mod

to DrStrangLov
said by DrStrangLov:

Your download speeds appears to be too high, as based upon Testmy and Xfinity's test during same primetime time frame. »Current Status , March 29, 2015

Can you link me with the result you feel is too high vs the "current status" or a xfinity test.
edit: never mind now I see.
justin

justin to DrStrangLov

Mod

to DrStrangLov
I'm looking at the log and after 25 seconds something peculiar starts to happen in your browser, the amount of data/second coming in keeps accelerating, leading to an ever increasing speed estimator.

I will replace the payload file, because I suspect that even though they are not compressible by gzip at all, the download compression in use has a much wider span and is picking up on repetition. Please test again soon to see if the download compression is conquered.
DrStrangLov
join:2012-03-28

DrStrangLov

Member

said by justin:

Please test again soon to see if the download compression is conquered.

See second post here: »Current Status , March 29, 2015

For the first two DSL tests, it seems upload was on high side; but when looking at last DSL test, the upload and download were way off, it would appear.

How your ping works, I don't know, but apparently, different servers are used for these tests; again, my POP is located in Denver, but since I use a dynamic IP address, software that estimates where the user lives is clueless about how satellite internet works. Google, for instance, can show me living in Montana, Nebraska, or in Kansas, overtime; different users using this same dynamic IP address.

Note - The first Xfinity test showed 26 Mbps download, which is in error.

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin

Mod

Oh man you missed the patch for your satellite, by 10 minutes, (early). Sorry. Nothing changed. But if you re-load page/re-run you'll have current code that I can look at the logs of.

The download compression they use is a lot better than gzip which is more of a local compression, and does not pick up on repeated chunks.
DrStrangLov
join:2012-03-28

DrStrangLov

Member

said by justin:

....\

See the third post here: »Current Status , March 29, 2015

Your upload test still seems a bit high...AT&T upload/download will be high also.

By now, Sunday night, most people have gone to bed in my beam area...after midnight for sure.

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin

Mod

I think that test looks about right?

I've got one more tweak to the upload to do (for satellite).

At the moment it is read from the remote download file, but I'll move the area it is read from to a part of the file that was never transmitted. If they have creepy upload compression that remembered it saw the same binary in the download I'd expect the upload to be off the charts and it isn't.
It is pure noise, not a jpg, not a gzipped file, but pure noisy data.
DrStrangLov
join:2012-03-28

DrStrangLov

Member

said by justin:

I've got one more tweak to the upload to do (for satellite).

See last test: »Current Status , March 29, 2015

Upload still high. Download may be OK.

I didn't realize Testmy was sending me to their CA server, so I switched to Dallas on second test. So, noticed difference between San Jose and Dallas...with my POP in Denver.

james1979
Premium Member
join:2012-10-09
Quinault, WA

james1979 to justin

Premium Member

to justin
said by justin:

The upload measurement is done as a MEDIAN ... and rather uniquely in your satellite result, vs other technologies, there is a long slow ramp up of speed at the start ...

the solution I think is to ignore more of the wind-up part, and concentrate on a median of the latter parts.

Or alternatively and in the case of satellite only, report the highest peak found as the result, rather than the median. If the highest, then it would report very close to 5.0

Which do you feel is better?

That's a good question which is why I thought a day before responding. Discarding valid data seems like a wrong idea when trying to design an honest accurate speed test. However, consider why speed tests are used with satellite ISP consumers. We like to compare HughesNet with Exede, and we like to check if our system is still working near what is advertised. That's not really any different than terrestrial speed testers, but since it is a satellite speed test, then it seems OK to me to adjust for the initial ramp up in speed and report typical sustained speeds.

As for reporting the maximum speed reached, you could just report that statistic. I don't know how far you want to take this project, but testmy.net reports statistics like that. (Unlike their speed results, testmy.net's statistics and speed graphs are usually not credible with satellite. I first maxed out at 1000 Mbps this evening. That dropped to 40 Mbps on the next test.)

A completely different suggestion inspired by a previous post would be to add an option to use SSL and just bypass the ISP's attempts at data compression. That would provide a better evaluation of what the satellite systems are capable of. In the case of HughesNet, users can (and some do) disable Web acceleration, so I would think they would be interested in that option.
DrStrangLov
join:2012-03-28

DrStrangLov to justin

Member

to justin
Around 1 AM, so beam has few users. DSL's Upload test went bonkers...should be around 3 Mbps. I'm using Dallas Testmy server, which is the fastest for those on Denver POP.

30th Mar 12:58AM (local time) »[Satellite Speed test: 14.88/25.78 760 ms]

Download 14.88 megabit/sec
Upload 25.8 megabit/sec
Ping 760ms



30th Mar 01:06AM (local time) »[Satellite Speed test: 18.42/4.23 823 ms]

Download 18.42 megabit/sec
Upload 4.23 megabit/sec
Ping 823ms

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin

Mod

Temporary ah minor problem, fixed now I think.

Regarding SSL I don't think it is necessary as the data is 100% random bytes from end to end. Perhaps you might install something that looks at your ethernet interface and shows perfect instantaneous speeds, and show me that the result really is over-reading so I can look harder?

Because if the result is the same as what is passing the interface, AND the data is incompressible, the speed reading would reflect the sync rate of the satellite path. I think.

james1979
Premium Member
join:2012-10-09
Quinault, WA

james1979

Premium Member

said by justin:

Regarding SSL I don't think it is necessary as the data is 100% random bytes from end to end.
...
the speed reading would reflect the sync rate of the satellite path. I think.

Yes, I understand that speed tests use random data. I posted about my studies with the scp program earlier. With Exede-12, it is faster to bypass their attempts at compressing compressed data by bumping up to an encrypted network layer. As downloading OS's, games, image collections of beach scenes, uncopyrighted media, etc. seem to be very popular, it seems like that might be a desirable preference option.

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin

Mod

Sorry I didn't mean to state the obvious. It actually was not until recently random bytes, it was random huge chunks, repeated. gzip etc does not waste time looking for 30mb chunks that repeat but perhaps the satellite did.

Anyway - the problem at hand is that you (sorry - DrStrangLov) thinks the download is too fast?
but are you also suggesting that the attempts to compress compressed data makes it slower?

james1979
Premium Member
join:2012-10-09
Quinault, WA

james1979 to justin

Premium Member

to justin
I encountered a ping problem with the speed test on an Android phone and Exede-12. I selected WiFi and received the results in the red box:




I assume this is because the speed test is not aware of the WiFi going a satellite ISP ... ?

I don't have an idea about the failed upload: »[WiFi Speed test: 17.58/0.00 219 ms]
james1979

james1979 to justin

Premium Member

to justin
said by justin:

but are you also suggesting that the attempts to compress compressed data makes it slower?

Yes, I strongly believe that attempts to compress only compressed data with satellite "Web acceleration" will be slower than satellite transmission of compressed data without compression as that is what happened. I will not share my model of understanding on the matter as it is incomplete. (This is where we need dbirdman, but still, their technologies are proprietary.)

I really only suggested SSL as an option for the future to make your speed test more unique.

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin to james1979

Mod

to james1979
said by james1979:

I encountered a ping problem with the speed test on an Android phone and Exede-12. I selected WiFi and received the results in the red box:

I assume this is because the speed test is not aware of the WiFi going a satellite ISP ... ?

Correct.
I suppose it should know from the DNS resolution of the IP that it is still satellite.

The only difference apart from guesstimates about speeds and up vs down ratio is the derivation of something close to ICMP ping without doing any ICMP ping.
DrStrangLov
join:2012-03-28

DrStrangLov to justin

Member

to justin
said by justin:

Temporary ah minor problem, fixed now I think.

Exede users, historically since 2012, could obtain around 3 Mbps upload, and 20 Mbps download. On ethernet interface, I have download stats when I was downloading Win-8 beta via Exede-5, but do note, the Maximum transfer rate is skewed, most likely due to how Exede's modem "dumps" in batches: »[Exede] File Download: Windows 8

Picture: »/r0/do ··· load.jpg

When marsh_0x (deceased) was downloading Win-8 beta via Exede-12:

Picture: »/r0/do ··· down.png

Today's speed test...uploading still appears high:

30th Mar 08:50AM (local time) »[Satellite Speed test: 18.95/4.13 774 ms]

Download 18.95 megabit/sec
Upload 4.13 megabit/sec
Ping 774ms

DrStrangLov

DrStrangLov to james1979

Member

to james1979
said by james1979:

I encountered a ping problem with the speed test on an Android phone and Exede-12.

Exede-5 test too: Test done via "Unsure" button, on desktop computer.

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin to DrStrangLov

Mod

to DrStrangLov
ok but is it not possible to grab a view at the interface while the test is runnning? I believe there are a lot of free utilities for this but not being on windows cant point you at a specific one. Another user poster one that looks good. I think if you search "data purist speedtest dslreports" google will find the post.
DrStrangLov
join:2012-03-28

DrStrangLov

Member

said by justin:

ok but is it not possible to grab a view at the interface while the test is runnning?

Best I can do...notice the upload's maximum transfer rate, which above photo shot was taken from previous run. It's appears their compression, along with WiMaxx coding crunches down the upload file size.

30th Mar 06:21PM (local time) »[Satellite Speed test: 15.94/2.95 1068 ms]

Download 15.94 megabit/sec
Upload 2.95 megabit/sec
Ping 1068ms

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin

Mod

So that tool says the current rate was 2.656 mbit/second
and the peak rate was 4.22 mbit/second (perhaps while it filled a hardware buffer?)

and the test recorded 2.95 mbit/second?

on the face of it, that looks within the bounds of rightness?

Note that I am discounting any compression possibilities because the byte stream on upload is both 100% random, and not seen in the download phase. Compressing 100% random bytes is akin to inventing a perpetual motion machine.
wh7qq
join:2012-06-12

wh7qq to justin

Member

to justin
Click for full size
Exede, Hawaii-Anchorage-Seattle:

"Incomplete result:
Browser stalled too many times during test. result unreliable"

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin

Mod

said by wh7qq:

"Incomplete result:
Browser stalled too many times during test. result unreliable"

Can you paste the link to the result (follow the green button), not a screen shot?
If you restart your browser does it still stall out?

Anyway I need to look at the log to see why, it is something on your side.
DrStrangLov
join:2012-03-28

DrStrangLov to justin

Member

to justin
said by justin:

...tool says the current rate was 2.656 mbit/second
and the peak rate was 4.22 mbit/second (perhaps while it filled a hardware buffer?)

With NetWorx program, once the start button is clicked, then "average" is based upon total time. So, this time, I tried to estimate when upload would happen/stop and then I clicked on start/stop button, with results below.

I have no idea how large the modem's buffer is, but Maximum transfer rate is skewed to modem's buffer.

People would have still been using internet on the Exede-5 beam I'm using...one more test needed. I also locked server down to Ohio server.

Maximum Transfer Rate - 5.84 Mbps
Current Transfer Rate - 4.1 Mbps

30th Mar 10:35PM (local time) »[Satellite Speed test: 5.65/2.74 798 ms]

Download 5.65 megabit/sec
Upload 2.74 megabit/sec
Ping 798ms
DrStrangLov

DrStrangLov to wh7qq

Member

to wh7qq
said by wh7qq:

Exede, Hawaii-Anchorage-Seattle:

"Incomplete result:

Pick a server to use....in that dialog, second item from right, just click on it.

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin to DrStrangLov

Mod

to DrStrangLov
said by DrStrangLov:

Maximum Transfer Rate - 5.84 Mbps
Current Transfer Rate - 4.1 Mbps

30th Mar 10:35PM (local time) »[Satellite Speed test: 5.65/2.74 798 ms]

But now "average" says 3.2 megabit/second in NetWorx, and the result from the test is 2.74
I've forgotten if you're thinking the results are a bit low or a bit high for upload.

Looking at the log, the entire upload was done where the graph dips off its 3.2 peak, It dipped because the file was gone, but not yet acknowledge, and for that 2.9mb random file, it took 8.6 seconds end-to-end (minus latency) which converts to 2.73 megabit.

The peak speed the file was disappearing at was 3.2mbit, but then there was a delay for acknowledgement, and the final speed was 2.7

I can change the "Satellite" test to report the peak speed of 3.2, if you think that is more representing the sync rate with the satellite?
DrStrangLov
join:2012-03-28

DrStrangLov

Member

said by justin:

But now "average" says 3.2 megabit/second in NetWorx, and the result from the test is 2.74

Networx's results are based upon when I started and stopped its program. ViaSat's modem has a buffer (and compression), so Networx's speed results can be skewed, including the average value.

ViaSat (Exede) uses an »AcceleNet server, which once it receives the data, its large internet pipe(s) can send very quickly packets to their destination.

DSL's download speeds seam reasonable (OK) for Exede users. But, DSL's upload results may/are skewed higher than Testmy results.

This result: Upload 4.13 megabit/sec »[Satellite Speed test: 18.95/4.13 774 ms]

I've never seen one Exede user going this high. Testmy showed after this test:



It may be the »AcceleNet server "stored" and then sent a bundle of packets to test server.

Thus, downloads tend to reflect other test servers' speeds, but ViaSat's »AcceleNet server may influence upload speeds, due to its "hold and send" nature.