Ξ welcome to cryptostorm's member forums ~ you don't have to be a cryptostorm member to post here Ξ
Ξ any OpenVPN configs found on the forum are likely outdated. For the latest, visit here or GitHub Ξ
Ξ If you're looking for tutorials/guides, check out the new https://cryptostorm.is/#section6 Ξ

Silverlight/Netflix performance on cryptostorm

Looking for a bit more than customer support, and want to learn more about what cryptostorm is , what we've been announcing lately, and how the cryptostorm network makes the magic? This is a great place to start, so make yourself at home!
User avatar

Topic Author
Tealc
ForumHelper
Posts: 238
Joined: Tue Jan 28, 2014 12:38 am

Silverlight/Netflix performance on cryptostorm

Post by Tealc » Sun Oct 19, 2014 11:30 pm

{split from general Windows perf-tuning thread, for clarity ~admin}

I do think that this is really important!

I have a lot of "customers" that use the Portuguese node to lift some geo location restrictions to watch local national TV trough a service called http://www.meogo.pt , they use the same system has Netflix, the Silverlight plugin, nevertheless and despite the speed (90 MB/s, awesome!), I get a lot worse quality conditions on CStorm them in my old VPN, IPvanish.
At this point I'm doing month to month subscriptions on IPvanish to accommodate 70% my costumers, I would like, a lot, to change them ALL to CS, but that would implicated a great improvement in the TV quality and a big, big, improvement in uptime (what's up with the "evil" node?).

I already talk about this with PJ, and he told me that this as nothing to do with the difference in the cipher between IPvanish (2048) & CS (4096), but in my small understanding I think that more the cipher is bigger more it's compressed, so less quality in image related services, or I'm wrong?

So for advice, I actually can't give one, but I can say that looking at some of the more windows based VPN's like IPvanish that would give some solution's, but them again I could be wrong.

User avatar

cryptostorm_admin
ForumHelper
Posts: 56
Joined: Tue Jan 01, 2013 5:43 pm
Contact:

Silverlight TCP-based MBR metrics

Post by cryptostorm_admin » Thu Oct 23, 2014 8:47 am

Tealc, this Silverlight issue has come up in the past, albeit under different circumstances. It's unlikely to be cipher-related, but it's premature to say for sure until we get some objective metrics.

The best way for us to knock this issue out is to be able to replicate it from machines where we can do packet-level analytics. It is possible to do this second-order, via data captures from customers... but that is always a challenge.

Is this exclusively a Windows-based issue? Is it exclusively related to Silverlight-based services? Do tests across these windows sessions for non-Silverlight applications produce suboptimal performance, as well? If it's exclusively Silverlight, we know that Silverlight has some known oddnesses in how it handles socket instantiation, and in general is sort of twitchy and proprietary and whatnot. But it's also fairly easy to massage, in terms of ensuring its judgement on MBR metrics isn't wildly off-base.

But before we dig too deep into that, let's be sure this is a Silverlight issue and not something more Windows-general.

Thank you,

~ cryptostorm_admin

User avatar

Topic Author
Tealc
ForumHelper
Posts: 238
Joined: Tue Jan 28, 2014 12:38 am

Re: Optimising Windows-based cstorm sessions: community advi

Post by Tealc » Fri Oct 24, 2014 9:51 pm

parityboy wrote:@Tealc

Is that 90Mbytes/sec, as in Gigiabit speeds? Who is your provider?
Yeah I typed it wrong, it's 90Mbit/s and not 90MBps LOL

My provider is UPC Cablecom (Switzerland) I got a real speed of 250 Mbit/s / 15 Mbit/s connected directly to my router by Ethernet cable, I have to say that this is quite awesome but I would prefer 50Mbit/s / 50Mbit/s :-D

But nevertheless the speeds I've got in Brisa or in Frankfurt are quite impressive for a VPN, I'm very impressed :clap:
Last edited by Tealc on Fri Oct 24, 2014 10:29 pm, edited 1 time in total.

User avatar

Topic Author
Tealc
ForumHelper
Posts: 238
Joined: Tue Jan 28, 2014 12:38 am

Re: Silverlight TCP-based MBR metrics

Post by Tealc » Fri Oct 24, 2014 10:10 pm

cryptostorm_admin wrote:Tealc, this Silverlight issue has come up in the past, albeit under different circumstances. It's unlikely to be cipher-related, but it's premature to say for sure until we get some objective metrics.

....

Thank you,

~ cryptostorm_admin
Hi there and thank you for your reply, actually PJ send me an email talking about the problem with the provider and the recent hardware problems, it just sucks that this only happend to a exit node that I really need :-D I'm also keeping a eye in the node before sending my "customers" from IPvanish to CS.
Concerning the Silverlight plugin, I've done extensive tests in different Windows computers with different hardware and even different windows versions.
In all the different computers the video quality of CS is a LOT worse them with IPvanish but I did notice that in the Windows 7 X64 machine CS was a little better them in the others Windows machines, but I can't know why :shock:
I think that this is cipher-related because I've seen this in another VPN "service" I've used, I belied that IPvanish uses the less secure cipher and with that the metadata is less compressed making the already compressed video signal in Silverlight better some how.
But this is just me saying, I'm not a expert in any of this, I'm a Telecommunications Engineer and my knowledge in this is very limited

User avatar

parityboy
Site Admin
Posts: 1215
Joined: Wed Feb 05, 2014 3:47 am

Re: Optimising Windows-based cstorm sessions: community advi

Post by parityboy » Sat Oct 25, 2014 4:55 am

@Tealc

When you say "video quality" do you mean the actual video quality, or do you mean buffering and dropouts? Have you checked the CPU load on the hardware that's actually handling the VPN connection, whether PC or router, while the video streams are being played?

User avatar

Topic Author
Tealc
ForumHelper
Posts: 238
Joined: Tue Jan 28, 2014 12:38 am

Re: Optimising Windows-based cstorm sessions: community advi

Post by Tealc » Tue Oct 28, 2014 10:49 pm

parityboy wrote:@Tealc

When you say "video quality" do you mean the actual video quality, or do you mean buffering and dropouts? Have you checked the CPU load on the hardware that's actually handling the VPN connection, whether PC or router, while the video streams are being played?
Yes I do mean the actual video quality, actually I have no buffering or dropouts, that's very nice. Concerning the CPU load on the hardware it shows exactly the same load on all computers when using IPvanish or CS, btw I also use OpenVPN with IPvanish, no proprietary software, and I've tried the widget of CS and still the same problem with the video quality.

I have to say that this isn't the first time that I have a problem with the "video quality" while trying others VPN's and I found that only a few did very well with the quality thing.
One thing I found to be funny is that while seeing Netflix in the US server there is no buffering (maybe a little at the beginning) no dropouts and the quality is HD!

User avatar

parityboy
Site Admin
Posts: 1215
Joined: Wed Feb 05, 2014 3:47 am

Re: Optimising Windows-based cstorm sessions: community advi

Post by parityboy » Tue Oct 28, 2014 10:58 pm

@Tealc

hhhmmmm....usually when I see video quality worsen when watching a stream, it's usually due to some rate-adaptive algorithm which delivers less pixels as the bandwidth drops, but your bandwidth seems ok. I'm wondering if what you're actually experiencing is packet loss at some point on the connection.

User avatar

Topic Author
Tealc
ForumHelper
Posts: 238
Joined: Tue Jan 28, 2014 12:38 am

Re: Optimising Windows-based cstorm sessions: community advi

Post by Tealc » Tue Oct 28, 2014 11:24 pm

parityboy wrote:@Tealc

hhhmmmm....usually when I see video quality worsen when watching a stream, it's usually due to some rate-adaptive algorithm which delivers less pixels as the bandwidth drops, but your bandwidth seems ok. I'm wondering if what you're actually experiencing is packet loss at some point on the connection.
I did thought of that some time ago, before telling support, and I did a "ping www.telepac.pt -t" while connect to CS and trying to see the meogo.pt and there where no packet loss, and I tried it again right now and the same no packet loss.
But your point is valid, and I have another one for you, if I use my HTC ONE M8 and my wife's M7 with the MEOGO.pt android app while connected with OpenVPN to Brisa, there's no problem with the quality of the image.... strange right?

User avatar

Pattern_Juggled
Posts: 611
Joined: Sun Dec 16, 2012 6:34 am
Contact:

pinning down causal roots via a/b testing

Post by Pattern_Juggled » Wed Oct 29, 2014 4:40 am

How about a test video streaming session that's non-Silverlight, across the exact same cstorm-wrapped network topology, to see if that exhibits any issues? H.323 or whatever... does someone have a resource for doing such streaming-video performance tests off a standardised backend, so that it's not being mediated by self-adjusting algorithms?

We could stick a video file up on a machine somewhere and hard-map it to a public-facing URL, so it's just sitting there waiting to stream - minimal extraneous moving parts. Use it basically like the wget's we do of test files, across nodes... just thinking out loud, here. That might help narrow down what we're seeing here. (for what it's worth, I think this is a Silverlight tweaking question, not a network-level issue in general, and can be solved if we pin it down more precisely)

Are there any Silverlight-specific perf-tuning tools out there already? It's a widely used protocol, and surely lots of developers have wrestled with making sure it imputes proper session capacity... when developers do such things, they often leave a trail of well-designed, useful toolsets behind for others to leverage in the future. Those would likely solve most of the easy problems, in terms of basic issue diagnostics.

Cheers,

  • ~ pj

User avatar

Topic Author
Tealc
ForumHelper
Posts: 238
Joined: Tue Jan 28, 2014 12:38 am

Re: Silverlight/Netflix performance on cryptostorm

Post by Tealc » Thu Nov 06, 2014 12:31 am

BUMP....

Concerning the image quality, I've seen that IPvanish is now having the same problem that CS, so I've set up a Rasp PI in my parents home network (that's the same network that the program I'm using to see the portuguese TV) and put a OpenVPN to test things out, and I've found this:

- With RC4-MD5, RC4-SHA, AES128-SHA, AES256-SHA, it works fine and the image is HD
- With DES-CBC-SHA, DES-CBC3-SHA it's a lot worse

Both server and client :-D

One funny thing is that IPvanish is still using the same AES-256-CBC, so I've come to the conclusion that this is the ISP (www.meo.pt) way of fucking things up for the VPN's and I'm afraid we will never find a solution :-( besides having a server in www.telepac.pt DC but I belied that's TOO EXPENSIVE for trying.
So I'm looking forward for the new Portuguese machine, since this one will be in another DC, maybe the ISP haven't gotten to block speed and quality on this one.

User avatar

Topic Author
Tealc
ForumHelper
Posts: 238
Joined: Tue Jan 28, 2014 12:38 am

Re: Silverlight/Netflix performance on cryptostorm

Post by Tealc » Thu Nov 06, 2014 12:50 am

Comparing a speed test I get from Switzerland with my 4G stable 4 bars network:

Image
(with CS)

Image
(with IPVanish)

So in terms of speed CS is actually a lot better them IPvanish, same DC :-D

Post Reply