[Bloat] Steam's TCP analysis

[Bloat] Steam's TCP analysis

[Dave Taht]

All over the net I hear of the bloated horrors steam and windows 10
updates are inflicting on people, and several saying that inbound
shaping isn't helping. I finally got two captures of a steam download
here:

https://github.com/tohojo/sqm-scripts/issues/43#issuecomment-275281826

And aside from some potential oddities (window, timestamp) didn't see
anything terribly odd in the first trace I got there. Could someone
take a look with smarter eyeballs than I have?

-- 
Dave Täht
Let's go make home routers and wifi faster! With better software!
http://blog.cerowrt.org

Re: [Bloat] Steam's TCP analysis

[Benjamin Cronce]

[-- Attachment #1: Type: text/plain, Size: 1592 bytes --]

In the past I've seen issues with Windows Updates because the CDN was 1 ms
away. TCP wants to have 2 segments in flight, resulting in a non-responsive
TCP stream below 13Mb/s. CDNs with low RTTs cause cause issues with low
bandwidth connections. Not only does DSL tend to have a low first hop
latency, it also tends to have less bandwidth than cable, making it a prime
victim for on-site CDNs.

I just attempted to install a game(about 1GiB) from Steam and it quickly
made about 20 connections to my ISP's on-site CDN. Even if you assume a
10ms ping for someone with DSL, that's a minimum of about 1.3Mb/s per TCP
steam. Below that, TCP becomes unresponsive to congestion. 20 connections
times 1.3Mb/s is 26Mb/s of packet flooding power.

On Jan 27, 2017 10:15 AM, "Dave Taht" <dave.taht@gmail.com> wrote:

> All over the net I hear of the bloated horrors steam and windows 10
> updates are inflicting on people, and several saying that inbound
> shaping isn't helping. I finally got two captures of a steam download
> here:
>
> https://github.com/tohojo/sqm-scripts/issues/43#issuecomment-275281826
>
> And aside from some potential oddities (window, timestamp) didn't see
> anything terribly odd in the first trace I got there. Could someone
> take a look with smarter eyeballs than I have?
>
> --
> Dave Täht
> Let's go make home routers and wifi faster! With better software!
> http://blog.cerowrt.org
> _______________________________________________
> Bloat mailing list
> Bloat@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/bloat
>

[-- Attachment #2: Type: text/html, Size: 2360 bytes --]

Re: [Bloat] Steam's TCP analysis

[Benjamin Cronce]

[-- Attachment #1: Type: text/plain, Size: 1934 bytes --]

Now that I think of it, since TCP wants a minimum of two un-acked packets,
you can just reduce the rate of your ACKs to keep the sender from flooding.
Total hack of course. It's really a packet-pacing issue.

On Fri, Jan 27, 2017 at 6:11 PM, Benjamin Cronce <bcronce@gmail.com> wrote:

> In the past I've seen issues with Windows Updates because the CDN was 1 ms
> away. TCP wants to have 2 segments in flight, resulting in a non-responsive
> TCP stream below 13Mb/s. CDNs with low RTTs cause cause issues with low
> bandwidth connections. Not only does DSL tend to have a low first hop
> latency, it also tends to have less bandwidth than cable, making it a prime
> victim for on-site CDNs.
>
> I just attempted to install a game(about 1GiB) from Steam and it quickly
> made about 20 connections to my ISP's on-site CDN. Even if you assume a
> 10ms ping for someone with DSL, that's a minimum of about 1.3Mb/s per TCP
> steam. Below that, TCP becomes unresponsive to congestion. 20 connections
> times 1.3Mb/s is 26Mb/s of packet flooding power.
>
> On Jan 27, 2017 10:15 AM, "Dave Taht" <dave.taht@gmail.com> wrote:
>
>> All over the net I hear of the bloated horrors steam and windows 10
>> updates are inflicting on people, and several saying that inbound
>> shaping isn't helping. I finally got two captures of a steam download
>> here:
>>
>> https://github.com/tohojo/sqm-scripts/issues/43#issuecomment-275281826
>>
>> And aside from some potential oddities (window, timestamp) didn't see
>> anything terribly odd in the first trace I got there. Could someone
>> take a look with smarter eyeballs than I have?
>>
>> --
>> Dave Täht
>> Let's go make home routers and wifi faster! With better software!
>> http://blog.cerowrt.org
>> _______________________________________________
>> Bloat mailing list
>> Bloat@lists.bufferbloat.net
>> https://lists.bufferbloat.net/listinfo/bloat
>>
>

[-- Attachment #2: Type: text/html, Size: 2975 bytes --]

Re: [Bloat] Steam's TCP analysis

[Yoshifumi Nishida]

[-- Attachment #1: Type: text/plain, Size: 712 bytes --]

On Fri, Jan 27, 2017 at 8:15 AM, Dave Taht <dave.taht@gmail.com> wrote:

> All over the net I hear of the bloated horrors steam and windows 10
> updates are inflicting on people, and several saying that inbound
> shaping isn't helping. I finally got two captures of a steam download
> here:
>
> https://github.com/tohojo/sqm-scripts/issues/43#issuecomment-275281826
>
> And aside from some potential oddities (window, timestamp) didn't see
> anything terribly odd in the first trace I got there. Could someone
> take a look with smarter eyeballs than I have?
>

Some connections in the dump file seem to have higher reordering rate
(around 15%).
I'm not sure yet, but it looks a bit odd to me.

Thanks,
--
Yoshi

[-- Attachment #2: Type: text/html, Size: 1420 bytes --]