Bandwidth on Preferences

Can anyone explain exactly how this works please. Can you increase your performance on sl if you increase the bandwidth here? Unsure what it does. Thanks

I've been told that infact having too much set on the preferences lowers performance. I didn't quite understand it all though. Basically, what I have been told is that you should set it to roughly 50% of what your bandwidth is or the max, whichever is lower. So if you have a 2mbps connection, set it 1mbps. If you only have a 1mbps connection, set it to 500kbps. If you have a 20mbps connection, set it to the make, which is 1.5mbps.

Not sure of all the technical stuff, but I think it is a simillar principle to setting the GPU memory lower to reduce the memory leak.

My isp is running at approx. 5.3mb so you say I should set the preferences at 2.5mb (2500kbps)

1.5mbs is the max preference goes upto. So base on what I have been told, you should set it to max.

Again, this is what I was told, but I can't remember all the technical reasons behind it I'm afraid.

Thanks Conan

When an ISP quotes you a performance rate, they're always telling you the best number you can theoretically see - if they say you've got a 1Mb/second rate, you're almost never going to actually see things download that fast.

Also, when you get close to your actual limit, the speed tends to vary more. Say you're actually getting up to 768Kb. If you try downloading something big at that rate, you'll probably wobble around between 650-768 instead of getting a steady 768. If, on the same line, you tried to download something at 500Kb, you'd probably be able to stay pretty close to that number.

I think the deal with setting the SL rate at less than the ISP-advertised rate is to keep you from totally maxing out your line (allowing for AIM, email, web browsing, etc at the same time) and to keep SL from getting confused about variations in the line speed - better to be a bit slower and steady than faster and wobbly.

The bandwidth setting is a way of presetting the maximum bandwidth you want to allocate to the client, bearing in mind you should allow for seperate data streams for such things as music, movies and voice, which are not included or controlled within the client settings.

If you arrive on a sim having TP'd in, or logged in there, push the bandwith up high for a minute or two will allow textures etc to download, before resetting back to around 1/2-2/3rds of your ACTUAL bandwidth (not as sold as bandwidth), to use voice, media etc. If you find that you are experiencing packet loss, try reducing the bandwith down in small increments to find a happy balance for the prevailing conditions.
Setting the clients bandwidth allocation is not a fixed setting, the client will throttle the bandwidth as and when needed, although it is slow to respond.

Having the setting too high, can result in packet loss, due to the fact that the controller only sets the download, and not the upload bandwidth.. when you have too wide a download against the upload data rates, packet loss can occur.
It's not the prettiest solution due to the very dynamics of the internet connection, but at least it is something.

Now you have me confused, my speed test still says I'm running at 5.3mb so the sl speed you say would be best at a lower rate than the max. so say around 500 - 900kbps would you say that is safer.

...so, if you are losing packets, *reduce* your bandwidth little by little until you are not losing more than a few percent at most (and none is ideal .

This is the case where reducing bandwidth can improve overall throughput.

If it's your ISPs speed test, it's probably testing the speed between you and somebody very, very close to you on the network. SL is a likely bit farther away..

There is no confusion.... If you check your bandwidth every minute of everyday for a week... you will rarely find the same connection speeds for more than a breif time. Take into account packet loss and time latencies.. will give you a clearer picture. If you also test from a single source, this is not indicative of RL uses.

I use a programme that automatically checks, ping, latency and bandwidth upload and download speeds across 4 continents, using a large database of available servers... by the end of the test run, which I have set at several different times of the day and night.. I am able to see patterns. The low ebb of my 8mb ASDL connection runs in line with 'busy/peak' periods relevant to the continents timezones, the highs, being relevant to the slack times (when most are asleep). Combined with the servers and nodes relevant to my own loaction, creates a connection chart that resembles a heart monitor output chart, only more erratic and varied.

Well, I have a connection with 20mb download (according the ISP).

I did serveral tests on speedtest.nl. My actual download speed varies between 14mb and 16mb. So I should set on around 7mb?

Morwen.

i see an opportunity to mention the bandwidth slider hack!

when you teleport to a sim and the loading seems to slow or stall, open up that preferences tab and wiggle the bandwidth slider from left to right really quickly.

then watch with awe as everything around starts rezzing super quickly.

No... you are only testing to a 'local' server, in the same country. Bear in mind my post about realtime usages over several contintents, and the effects of latency and ping times. To fully test your system, and gain realistic 'actuals' you have to to take readings which include Ping, Latency, Download and Upload speeds into account.

Also, SL is hosted in USA, and you will be using the same pipelines through the major transatlantic backbone that the rest of Europe. I also use TraceRoute to check individual servers throughout the world, giving latency and ping values, especially our various streaming servers throughout Eurpoe and the USA.

cant say for positive regarding sl, but in hl or quake jacking up the downstream speed too high results in packet loss.

I'd still say set it to 1500kbps in your case. If your ISP max was 1500, then yes 500-900 would be appropriate.

Again, the client's preferences only go upto 1.5mbs - so set it to that.

I just had to repeat this.... it's not always about bandwidth...

So true..Be aware (As is the case I found out the hard way) some ISPs also pose bandwidth sharing policies, like the one I have with COUGH>>HUGHES<<COUGH, called an equal usage act, trying to balance across their subscribers...I did not know this until the next day after I first joined SL (..and spent nearly 5 hours straight In-World...)my upload / download had slowed to a whopping 3 kb/s...Thruogh some searching on the Hughes customer support I found where they showed my usage and they had enforced a MANDATORY slowdown for 24 hours as a result of exceeding that limit and they claim that continued abuse of the limit will impose stricter limitations...

Another thing I did to help, since some places In_World may be sparsely populated and others that you may TP to may be simply overwhelming, graphically, for your already taxed system:

Use the key command (? - Someone told me the keys) to open DEBUG (All it does is add 2 columns to your tabs across the top of your display, DEBUG and SERVER..From the DEBUG menu, you can select graphics specific items and click through them to enable or disable features, like trees, LOS (Line of sight) distance, clouds, ripples in water, even completely down to pure wireframe models if you want...

Sometimes, by tweaking these options, you can get a far smoother running system when you're In-World and the chunking, lagging feeling can be alleviated to an extent...It deifinitely helped me...

Quik

lol, thank you (:

Ctrl+Alt+D brings up the Client and Server menus... whilst that is sticky (stays there even when you loggout) the checks for disabling certaining types of rendering, isn't. You can launch the Debug menu and trawl through the settings (at your peril) and make setting changes there, although again, not all are sticky. I believe someone suggested editing the xml settings files on your system... but unless you know what you are doing.... lol..

With regard to the connection speeds dropping, this can also be caused by interleaving of your connection, this is basically, taking that you have a contention ratio of say 50:1 users all using your 10mbps connection placed on a server node, some nodes will be heavier used than others, so now the interleave the nodes to gain a better average connection. In theory this may work, unless your part of town is a light user area, you will find your 'slack' is now being shared with other heavier users, bringing your average down to match theirs albiet that theirs is now slightly higher...

AWM,

Thanks for the info; very helpful! Which (hopefully free) Windows XP program can you recommend for finding out bandwidth, speed, and such? Or is it already lurking on my computer in a really obvious place?

Oryx

Oryx, you don't need to install any software to do a speed test of your internet connection as long as you have a working web browser like Internet Explorer or Firefox, fire one of those up and go to http://www.dslreports.com/speedtest and click on the Flash 8 plugin based speed test (since you most likely already have Flash installed it is usually easier then using the Java based test).

It is a good idea to make sure you are not running any other applications while performing the test, especially applications that use the internet.

It's true, you don't need to install some software to test your connection.. a simple search of yahoo or google will reveal a prolifercation of test servers. However, unless you make a folder in Bookmarks, and run each test from a scattering of all the continents, you are only getting a very small slice of a good average. Before installing the software I now use, I did exactly that. Apart from the time it took, most don't give you ping and latency results, or in some cases, your upload speed.

Running one test after the other in a manual way, would give me such variances it was hard to judge which results were good. Using the links, one after the other several times to two different servers in the same country.. first one stated 6mb, second said 256kbps, jumping right back to the first, still said around 6mb, and the second stating 750kbps.. which one is right? Even when I was experiencing connection slow downs, some sites would report a good connection.
The main differences was the size of the test files used by some website testing. Some being 50mb, others being 256kb...

By using the programme I do, it uses the same parameters for each test, drawing its range of servers across the world into a database which is updated everytime I run the programme. It tests a minimum of 3 different servers for ping, latency, upload and download each. Whats more, it not only saves the results into a portfolio, but I can forward them directly to my ISP.

When I used this programme over a month, sending results 5 times a day (you are allowed to forward the results up to 10 times per hour under current laws without it being classed as spam), my ISP actually did something and I had them de-interleave my connection at the exchange node and I now get much more consistant connection speeds. It would be very rare anyone getting action from their ISP, by taking a website test and forwarding the results.

Read this as an example

[url="http://www.broadband-help.com/articles/networking/broadband-speed-test/

The software I use (I am not employed, or on any commission with this company) can be found here:

[url="http://www.broadbandspeedtest.net/download/

They offer a crippled free to try basic package, or for a one off fee (£10 or $15) you get the whole works which is updated eachtime you run the programme.

One word of warning... if your connection has a cap for downloads (99% of ISP's do in the UK, even when stated 'Unlimited', but is overridden by the 'Fair Usage' terms), running this programme to the extend I have, you may find yourself being emailed by your ISP. Contact your ISP and get a Support Number from them, and an email address you can forward the results to. Noting to them, that you are running these tests and that should it take you over your capped limit, it will only be for the duration of supplying them with this information. My ISP were quite happy about this, as it provided them with good and consistent information which they could use to assess the issue.

OK, I tried Gabby's test as mentioned above (thanks!) and I ran it 5 times to the server in San Francisco, and then 5 times to the server in Texas. The San Francisco one seemed pretty stable, with an average download speed of 12376 KB, upload of 536 KB, and latency of 28.6. I'm assuming this is good? Then when I did the Texas one, the results were all over the spectrum, from a minimum DL speed of 2500 KB all the way up to 12000 KB.... same with UL and latency. Why is the San Fran one stable, while the Texas one all crazy n stuff?