Repetitive sounds and chat lines

[Arduenn Schwartzman]

Before you start thinking it might be my internet connection:

1. I power-cycled my modem/router and it still happened.
2. I empied cache/rebooted and it still happened.
3. My connection speed, 1500 kbps, complied with http://wiki.phoenixviewer.com/fs_speedtest
4. Two other people (Finland, Germany), in different countries than me (Netherlands), experienced the same with the sounds.

So here are the symptoms:

1. Sometimes, not every time, sounds played from scripts, from double clicking sounds in the inv and even from gestures, do so in a repetitive way. Most of the time double, sometimes tenfold in a very rapid sequence, during which the sounds even overlap.
2. Viewer GUI ('Buttons') sounds sometimes even repeat themselves too.
3. On one occasion, a scripted object produced chat lines double as well. The chat content was retreived as a result from an HTTP request to an external web server. Not sure if this is related to the sound phenomenon.

The sound thing has been going on for a few weeks now.

Anyone got any idea if this is server-related?

[Arduenn Schwartzman]

(Maybe it's an NSA fiber-optics beam splitter feedback thing?)

[Janelle Darkstone]

(Shhh.  We don't discuss the NSA fiber-optic beam splitter feedback thing in an unsecured public forum.)

[Ayesha Askham]

Arduenn

This phenomenon has been in SL for at least 3 years now.  It is (so far as I know) a result of the sound-file buffer of the viewer overloading.  I believe it began with Snowglobe-based viewers and was unfortunately ported over into the original V2 code.

Firestorm devs attempted to solve this issue a few times, but only with limited success until Firestorm v4.  Whether the current Linden viewer has the issue or not I do not know.

I wonder if viewers such as Cool VL and Singularity also have the issue.  It certainly occurs whether or not you use XP or Win 7.

As to ameliorating the effect , one thing that might help is to reduce the range at which you perceive sounds in Prefs, otherwise just don't visit sims that run high numbers of sound files.  It starts in-world, but as you note it affects UI sounds as well.

[Theresa Tennyson]

---

Arduenn Schwartzman wrote:

Before you start thinking it might be my internet connection:

1. I power-cycled my modem/router and it still happened.
2. I empied cache/rebooted and it still happened.
3. My connection speed, 1500 kbps, complied with 
   http://wiki.phoenixviewer.com/fs_speedtest
4. Two other people (Finland, Germany), in different countries than me (Netherlands), experienced the same with the sounds.

So here are the symptoms:

1. Sometimes, not every time, sounds played from scripts, from double clicking sounds in the inv and even from gestures, do so in a repetitive way. Most of the time double, sometimes tenfold in a very rapid sequence, during which the sounds even overlap.
2. Viewer GUI ('Buttons') sounds sometimes even repeat themselves too.
3. On one occasion, a scripted object produced chat lines double as well. The chat content was retreived as a result from an HTTP request to an external web server. Not sure if this is related to the sound phenomenon.

The sound thing has been going on for a few weeks now.

Anyone got any idea if this is server-related?

---

Repeated anythings are often a result of packets being not lost but delayed. Your viewer sends a request and doesn't get a reply when it's expected so it sends the request again, thinking the request was completely lost. Meanwhile the originally requested packet finally arrives as does the re-requested packet and you get repetition.

I live in the US and have never had this happen with sounds in the last three years. I did have it happen occasionaly with chat text, but only during the time I was living in an extended-stay hotel and was relying on a shared wireless connection.

For several weeks some people from Europe have been having a variety of connection issues and I believe they all are on ISP's that contract with the "above.net" backbone company for the path from Europe to the Linden Lab servers - the problem seems to be within the above.net system somewhere in the US. This sounds like a similar problem and if so there's little that can be done on either the user end or the Lab end.

[Ayesha Askham]

Theresa

I agree with your analysis up to a point, packet loss can indeed cause these issues, especially repeated lines of chat both local and IM.

The sound buffer issue is not simply packet-loss related, indeed I have encountered it while seeing zero packet loss.

The issues with excessive ping times to certain European and overseas countries from the USA is by no means solely limited to the users' ISPs.  While it always appears that ZayoCC (Abovenet) is at the root of the trouble, the variations between LL simulator hosts implies another factor which is within the Linden intranet. Whether there are ANY Transatlantic connections NOT operated by ZayoCC, I simply do not know.  It might also be germane to wonder whether communication between Linden Lab datacentres is mediated by ZayoCC.

ddTemin asks a related question on another thread in this forum, and it is likely that throttling of data is causing that phenomenon.

There is no doubt (it is clearly and easily measurable) that Linden Lab is sending far more data down connections that used to be the case from the servers to the user and this is resulting in throttling at several stages within our connections.

To say that it is beyond Linden Lab's ability to ameliorate this is just not correct and some efforts on their part are essential if the LL userbase is not to shrink still further. 

SL cannot survive if it becomes restricted to the USA.  It is worth my adding that some parts of the USA are now seeing ping-time response issues with certain functions in SL.

[Theresa Tennyson]

---

Ayesha Askham wrote:

The issues with excessive ping times to certain European and overseas countries from the USA is by no means solely limited to the users' ISPs.  While it always appears that ZayoCC (Abovenet) is at the root of the trouble, the variations between LL simulator hosts implies another factor which is within the Linden intranet. Whether there are ANY Transatlantic connections NOT operated by ZayoCC, I simply do not know.  It might also be germane to wonder whether communication between Linden Lab datacentres is mediated by ZayoCC.

 

---

After reading your description of problems with ping times and "good" and "bad" servers, I ran tracerts to the various simulators you reported. The "good" and "bad" servers were on exactly the same path and had pings within a millisecond of being identical for me. Granted, my ISP doesn't use above.net as a backbone provider but if the problem was with the LL infrastructure it should have shown up as a difference.

Many people are having problems with not the SPEED but the NUMBER of connections SL uses when using HTTP connections - there's an ongoing project to reduce the number of HTTP connections the viewer opens but in the meanwhile you can turn HTTP textures off, reduce your MaxMeshConcurrentRequests, etc.

ETA - There are a number of transatlantic paths not run by ZayoCC. What path your ISP uses is largely a function of their contracts with other providers. For instance, my connections generally go on Level3, even across the Atlantic.

[Ayesha Askham]

Theresa

Again, your points are valid as far as they go.  The routes to the simservers I have measured WERE different when I tested them, so what that shows I'm not sure.:smileyfrustrated:

ETA: Of course it helps to thinks while posting...I never did find out if the sims I restarted my region to were on different paths...I assume that's what you were referring to, Theresa.  It actually worries me more that you found them to be the same...perhaps it was just a bad host.  The ones I was referring to were sims I visited when the high Ping issue started for me a few weeks back.

The issue with the number of http connections IS a problem for a number of router designs, and that issue is well documented, though I don't have the url to hand.

Turning http texture get off causes a severe performance hit for many (it does for me) so it is a case of "out of the frying pan into the fire" really.  Reducing the MaxMesh thingy is a real "fail" for me, it causes too many other issues.:smileysurprised:

The routes across the pond may well be different depending on where you start from, all I know is that all the routes I have had data for from Western Europe use Abovenet.

[Arduenn Schwartzman]

Thanks for all your replies, folks. I'm going to try if things can be soved by reducing http connections and the amount of ambient sounds in the region. I'll keep you posted here.