Lab asks: how is SL for you?

Just Another Tequilla Sunrise, Isle of Love; Inara Pey, October 2014, on FlickrSL should be looking and feeling a lot better for many of us as a result of recent work by the Lab – how’s it going for you? The Lab asks you drop them a line in the forums or via Twitter (image: Just Another Tequilla Sunrise, Isle of Love (Flickr) – blog post)

Following-on from the grid-wide deployment of CDN support and the promotion of the HTTP pipelining viewer as the de facto release viewer, the Lab has blogged about recent improvements to Second Life, finishing with the question “how is it for you?”

The blog post, entitled, Performance, Performance, Performance, opens thus:

Has Second Life seemed a bit faster for you lately? Improving performance for all Second Life users has been an important focus for us at Linden Lab, and we’ve recently seen some great results from several projects that should make your Second Life experiences faster, smoother, and more reliable.

It goes on to make fair mention of the CDN / HTTP work, noting:

Faster Texture & Mesh Loading
The entire grid is now using a CDN service for textures and meshes. This change means that textures and meshes should load more quickly, particularly for those who login to Second Life from places that are far from our US data centers. Our testing showed dramatic improvements: average download times for textures and meshes have been reduced by more than 50% on average, and the improvement is even more dramatic outside of North America.

Quicker Viewer-Server Communications
Another way we’re enhancing Second Life performance is through our HTTP project, which improves the way your Viewer communicates with grid services. With the HTTP Project Viewer out now, the faster content download times you’ll see thanks to the CDN change get even better – we’re talking 80% faster!

For those who may have missed news on the HTTP pipelining viewer and the CDN support, you can catch-up with things via a couple of posts on this blog: SL project updates week 42/2: Monty’s HTTP update and the HTTP pipelining viewer, and HTTP pipelining viewer reaches release status as CDN support is grid-wide.

Mention is also made of the recent Group Chat updates (work is still continuing on this, and you can get updates via the Group Chat tag in this blog).

However, perhaps the most interesting aspect of the blog post is the news that there have been some significant infrastructure updates with the SL Marketplace, which appear to have slipped mention through other mediums such as the var UG meetings. Here the blog post notes:

Speeding Up the Marketplace
If you visit the Marketplace today, you should be seeing a much snappier experience than in the past. We recently deployed infrastructure upgrades for the Marketplace, and the site has since shown some of the best performance we’ve ever seen from it. Even during peak usage periods over the weekend, when in the past performance would degrade, we’re seeing response times that average 70% faster and page load times that are 30% faster than before the changes.

I’ve not used the Marketplace of late – although I have been covering the upcoming Viewer-managed Marketplace (VMM) changes that will be occurring in 2015, and which the Lab is currently gearing-up for, along with merchants and TPVs.

So have you noticed changes and improvement in your Second Life experience? If so, then why not follow the Lab’s request:

So, How’s Second Life Performing for You?
Performance improvements are generally behind-the-scenes work, and we know it’s not always as exciting as rolling out a new feature, but these changes directly impact all our Second Life experiences and our daily lives in-world. We hope you’re starting to notice the effects of these improvements – if you are, please let us know in the Forums, on Twitter, or however you prefer.

Note that is let the Lab know – not me (although general views are always welcome here, they might not be picked-up by the Lab)!

Advertisements

23 thoughts on “Lab asks: how is SL for you?

  1. I read the tech paper from HighWinds on the CDN.. and with all he caching and proxies I wonder of it will work well when there are rapidly changing assets such as presentation screen content and changing clothing and baked textures for avatars. Does this get pushed quickly to all cache copies? May be worth a test and see if its more delayed than before?

    Like

    1. In terms of avatar baking, its my uunderstanding from comments made by Lab personnel that support of server-side appearance was the first major use of the CDN. For presentations using pre-loaded textures (via scripting), don’t know if it would by better or worse than coming via the sim host / Amazon.

      Like

  2. Do you have to use the Linden viewer to get these benefits?

    On an Apple and stuck back on. FS 4.4.2 — and from office hours doesn’t sound like much is being done by the Lab to fix the issues– or they are unfixable.

    Like

    1. CDN texture and mesh fetching works across the board with all viewers. The HTTP Pipelining updates are viewer-specific, and the code will be finding its way into TPVs over time (Alchemy already has it, for example).

      Mac issues – this is a tough one, largely because the Lab don’t appear to be seeing some of the issues others are, although I believe some TPV devs are digging into things. Something like the typing input lag aren’t actually seen as an SL issue, as its been reported it can occur elsewhere. The Lab is also of the opinion that updating to OS X 10.9 tends to resolve the issue for many people – although whether this can be said to be for “all” people is open to debate. The Firestorm team are continuing to poke at LL at TPV meetings on the matter as well.

      Like

      1. I’m running the latest OS X 10.9 and I get the keyboard input lag issue regularly. I also do _not_ see this issue with other games that can experience graphics lag. We’ll see how 10.10 fares soon…

        Like

        1. Yup, hence why I said the Lab’s position was open to debate :).

          Be aware that Yosemite apparently uses older nVidia drivers (if you use nVidia), and there have been reported issues with keyobard shortcuts – see my week 43 project update for info.

          Like

          1. Quite open to debate indeed.

            I’ve posted this elsewhere, but the chat lag issue seems to be completely independent on viewer, hardware, and operating system. It’s deep embedded in the LL code for CHUI.

            What happens (and this is why it cannot be replicated easily) is that, the newer the hardware, the smaller the lag is. There is a threshold below which humans simply cannot notice any lag (something around 200ms).

            Now, because I happened to have two 2007 Macs where the chat lag issue was very visible — measured in minutes, not milliseconds — I’m aware that it truly exists. When I got a new Macbook Pro, first running Mavericks, and now Yosemite, the hardware is so fast that I can very well understand that ‘nobody’ at the Lab (or, for that matter, most TPV developers) cannot notice any chat lag.

            But if you have suffered from it for months, you know the tell-tale signs. You become better at spotting issues, if you know what you’re looking for. And yes, way faster hardware will turn ‘minutes’ into ‘milliseconds’, making chat lag barely imperceptible: but it’s still there, while FS 4.4.2 doesn’t have it.

            There is even a way to replicate this on super-fast hardware: if you’re on a laptop, just let the battery drain until the end. When it drops below 10% or so, Mac OS X will start to slow down the CPU in order to conserve battery. And then there comes the chat lag, in all its glory. Of course everything gets slowed down (while on FS 4.4.2, chat lag is independent from 3D rendering lag on a slow machine) so it’s harder to spot.

            My advise to anyone who is working on fixing this and cannot replicate it is to grab a very old computer, run FS 4.4.2, go to a chat-intensive area, and measure what happens in chat. Then switch to a newer (i.e. 2014) viewer. The difference has to be visible! Switch to Mavericks, Yosemite, and super-fast hardware, and, while the chat lag stops bothering you, you’ll still know what you’re looking for — it will still be there, hovering around the 200 ms threshold. Sometimes it goes a little above that and it’s immediately perceptible; most of the time, it might not be, and thus the claim that it’s not chat lag at all but merely overall rendering lag, etc. — but it isn’t. It appears even on low-polygon areas with few avatars, even if you turn settings down, just have impostors, etc.

            Like

            1. Interesting that Apple reverted to older nVidia drives. What was the purpose of that? I tried to find some reference confirming this (outside the SL blogosphere), but couldn’t find it. Might we SL residents be particularly sharp in detecting those things? 🙂

              Like

              1. The warning came from Cinder Roxley, who has considerable experience with Apple (and is doing much to dig into various Mac-related viewer issues).

                Like

            2. I’m not sure if you’ve been following my group chat updates, but the focus of the work has been on the back-end of the chat service, where there are significant issues related to message handling versus the handling of chat-associated updates (tracking people logging in or out of SL, joining leaving chat sessions, circlating these updates to others in the group chat / group, etc).

              These updates had been vying heavily with actual messages for back-end processing time, leading to issues with lag and messages failing to get through, etc., and Simon Linden has been committing considerable effort to reducing this impact. Anecdotal evidence from very large groups – such as the Firestorm Support English (25,231 members with very busy chat) – suggest the changes have led to a noticeable improvement, with fewer issues being encountered.

              Further work is currently underway, looking at another element of the chat service which may lead to lag: how the system goes about looking-up where you are as you move about in-world in order to ensure messages are correctly routed to you. This becomes especially significant if you’re moving between regions a lot, either via teleporting or via vehicle with multiple region crossings.

              How much of a visible improvement this latter work might make to things is unclear. However, the focus on the back-end processing in both of these cases should hopefully yield benefits to everyone, regardless of client-side hardware (which is obviously harder for the Lab to take into account when trying to optimise things).

              Liked by 1 person

            3. I am on latest 10.9 and a 2010 iMac using 4.4.2. and I have chat lag — especially bad in IMs.

              In a sim running well two avatars and I get the lines about three back– and other party is responding to me about three replies back.

              Also while typing I will gave to wait for a while sometimes for what I typed to appear in the box before I can finish and hit enter.

              My connection is a cable fiber and ping 70 ish to Dallas servers using speedtest. I run at 32 draw and low settings. Which stinks– used to be 100 draw in 2009.

              Seems so odd that talking is harder now with less people in world than when the place was stuffed with noobs teleporting and probably having more IMs going all at once.

              Like

              1. “My connection is a cable fiber and ping 70 ish to Dallas servers using speedtest. I run at 32 draw and low settings. Which stinks– used to be 100 draw in 2009.”

                In terms of speedtests to the simulators, you might want to try a test to Phoenix, AZ – that’s now the location for all the simulator servers.

                “Seems so odd that talking is harder now”

                Are you sure, or is that purely perception? Group chat issues have been around since at least 2007, and the famous is group chat still broken URL, which came about as a result of the poor performance with group chat, has been running since July 2009. As such, I’d suggest that it is not any worse; it’s just that there hasn’t been any noticeable improvement, which makes it feel worse. (although there does seem to have been some improvement oted as a result of the recent work, at least in some of the larger groups where ghroup chat is commonplace).

                Like

                1. I’m sure it’s worse. Two or three years ago I’d experience chat lag in the form of messages I sent taking a long time to show up, but I never had a case where simple typing echo was visibly and regularly lagging behind my typing (and I’m not _that_ fast a typist). This is a relatively new thing, and while yes, it is somewhat better with newer hardware, it still is enough to render me virtually mute at times. LL blaming it on MacOS doesn’t sit right with me. There’s some thread that’s being tasked with doing too much in my opinion, and that’s down to how LL has designed their event loop(s).

                  Like

                2. Sorry I was unclear — group chat is what it is — seems better lately if server up for the group .

                  But IMs lag inside the one to one window– that is what destroys relationships. That lag is newish almost a year now.

                  So discouraged Apple issues are sidelined — I want a new computer to roll in the new glossies like a puppy in autumn leaves — but SL mac issues make it not worth it right now.

                  Like

  3. Not being very active on sl this week, didnt do any trip around mainland so i cant speak about the state of crossing sims.
    Didn’t notice any change in how things rezz on where ‘ve been login on Nautilius, neither on our private homestead.
    Did notice a better rezzing time on the few music venues i went, but they where not to much populated. Still i do think there the rezz times did improve.
    Strange thing happen swhen you zoom out and turns away from a avatar or object, when you zoom back, all seems to be in need to rezz again, But this behavior was already happening.
    So for the lil use we have on sl this week we cant really say we are ntocing any advantage, but on the few private sims that host music venues.
    Using latest Ukando viewer.

    Like

    1. “Strange thing happen swhen you zoom out and turns away from a avatar or object, when you zoom back, all seems to be in need to rezz again, But this behavior was already happening.”

      This is a result of the interest list updates, and can be particularly noticed when camming around, or in cases where you look away from a moving object and then back towards it (when it may suddenly seem to “jump”, as the intervening updates on its motion may not have been sent to the viewer).

      Like

Have any thoughts?

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.