Quantcast

Got Homework?

Connect with other students for help. It's a free community.

  • across
    MIT Grad Student
    Online now
  • laura*
    Helped 1,000 students
    Online now
  • Hero
    College Math Guru
    Online now

Here's the question you clicked on:

55 members online
  • 0 replying
  • 0 viewing

agentx5

Unacceptable QoS loss with the website! (i.e.: lag, duplicate posts in chat, duplicate PMs, ...) This occurred at 10:20 AM Eastern Time (The zero meridian time is now 14:20 2012-07-26). Screenshots attached below (obviously I covered over my own IP address). Here you can clearly see this is an issue with this website's servers or their ISP connection to the internet, whatever the root cause may be. I would like, for the sake of all the people on this website who use it for it's purpose. To see this fixed, ASAP, please.

  • one year ago
  • one year ago

  • This Question is Closed
  1. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    Proof:

    • one year ago
  2. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    Hey @agentx5 :) It's been happening since 8pm EST. Lag has been repeating everything we say 2-8 times. It's slow. It sucks. We can't ask or answer questions without it lagging. We can't chat without it lagging. We can't look at it without it lagging. We all went to twid and even did some work on there. It didn't lag -__-

    • one year ago
  3. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    I would also STRONGLY suggest streamlining the ridiculously redundant chat system. Get rid of the multiple channels thing, make it just one channel. It's a tremendous waste of server memory and CPU processing time. Nobody hardly even uses the other channels, it's most mathematics chat (pretty much exclusively). More importantly chat is NOT an essential function to the purpose of this website. Some may disagree with me but I will cite the almost unanimously aligned point of view of the community on this topic: http://openstudy.com/updates/50096c23e4b020a2b3be82bf Let's be honest, Chat is there mostly as a means to socialize and communicate in real time and understand each other. It's important to the community surely, I'm NOT saying get rid of it! Just to be crystal clear on that. It's very beneficial to the community. I'm saying as a non-critical function of the website it needs to get streamline so it's not so wasteful. And with the way the servers have been running lately it's be atrocious and every little bit of waste we can get rid of that makes the system run better is worthy of consideration.

    • one year ago
  4. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    Lots of people use other group chats! Math is just the most common one. Math is actually (of all of them) the one that is most of topic is all. I do agree they need a way to fix the insane lag. What happens (as a mod told us,(I forgot which one)) Is that It takes our comment gets bogged and then spits it and to compensate for eating it, it gives us lots of them back.

    • one year ago
  5. zepp
    Best Response
    You've already chosen the best response.
    Medals 0

    Oh wow, 100% loss D: http://puu.sh/LSwc

    • one year ago
  6. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    Actually no @rebeccaskell94 , not lots a people. Only a few on rare occasion. Maybe it was that way once but it's illogical to keep an antiquated redundant system around simply because of nostalgia. Yahoo made that mistake and Google took their market share away and then some. Programmers: Remember the KISS policy! Also that can't be the complete story Rebecca, because if that was true you wouldn't have such latency when you go to ping or trace. So, what the Moderator told you was incorrect, or rather incomplete, sorry. I'm not saying that's not going on, but this issue looks more like what you see when there's an ISP connectivity problem, somebody is DDoS attacking (hacking) the website to troll, or the server itself is having issues (hardware issues or software). I realize that doesn't narrow it down much, but unless I had administrator access I couldn't tell you anymore, none of us regular members (or even the Mods) could. The only people who can address this are the ones behind the coding and administration of the website. Contrary to what I first thought, Moderators \(\neq\) Admins here

    • one year ago
  7. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    Well, some of the mods are admins, but not all of them. The rest of your post...went right over my head. It's beyond me lol I was just telling you what I heard. xD I'm sure you are much more knowledgeable than me on the matter.

    • one year ago
  8. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    Here's my bottom-line: if this doesn't get resolved soon (as in the next few day) they loose me as a customer at least for a long time. I came on today with the express purpose of getting some hard Calculus 2 concept questions answered, maybe get some of my work checked. I cannot do that, therefore this website fails at it's purpose. *shrug* I'm not mad at all here btw, but I am a bit disheartened and saddened. Things happen, but if this has been going on for more than a few hours, it's unacceptable. And yes, this website has been having lag issues increasingly as of late. This stuff today is just over the top, to the point I had to close my beloved topic I created in this section and create this one to make a legitimate complaint about the QoS (quality of service) issue. @rebeccaskell94 , I wouldn't claim to be an expert on netcode or networking (there's always somebody who knows more than you, and I am humbled next to CCNA certified professionals), but I do happen have tremendous knowledge on such things (enough to be very dangerous as they say) and I'm trying hard to not go over people's heads with this one. PS: Admins, if it's a ISP issue, I have a telecomm consulting firm I can recommend that is very experienced and can probably help out: http://ecomtek.com/

    • one year ago
  9. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    Whoa.

    • one year ago
  10. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    next few days* (arg this lag is getting to be frustrating)

    • one year ago
  11. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    I hate sounding like a bumbling fool with my written words lol >_< Derp.

    • one year ago
  12. cshalvey
    Best Response
    You've already chosen the best response.
    Medals 1

    Hey @agentx5 - sorry about the issues that have been happening, and I understand your frustration. Believe me, no one is more frustrated by outages and performance issues like these more than us. Our developers literally put in hours of work on OpenStudy, and that obviously doesn't guarantee perfect performance. What I *can* assure you is that our team is cognizant of the issue, and though you may be experiencing issues, I can assure you that OpenStudy as a whole has gotten more stable, has less downtime, and has performed better as time has gone on. And our plan is to continue that trend going forward. As to the technical issues, I won't pretend to be an expert, and will allow one of the actually smart people, or @mattfeury - whoever gets here first - to answer some of your points better. Thanks @agentx5 - I appreciate you bringing this up, and I hope that you will continue to consider OpenStudy as the best place to help out on the internets ;)

    • one year ago
  13. mattfeury
    Best Response
    You've already chosen the best response.
    Medals 2

    Howdy all, I guess I didn't make the cut for the "actually smart people", but I'll do what I can (and I won't use 'literally' in such an unnecessary fashion) ;). In regard to the technical problems, agent, most of our issues resolve around garbage collection. Don't cringe, but our server runs on the JVM, which is unfortunate but fortunate at the same time. Unfortunately, as the site's uptime increases, we retain more and more garbage, the site starts lagging, and requests may or may not get handled. I would like to affirm Colm's comment that we have put in a lot of time to get the site to a usable place. Consider the fact that you can stay on our site for "literally" days and have it stay entirely up-to-date. Top that with the fact that as we grow in popularity, we have over 200 users on the site throughout the day. I'm not trying to pass the blame here but merely give it some context. We are a very small team (only 3 developers, one of which is on vacation at the moment), and we do consider performance to be the site's #1 feature. Without a smooth experience, it doesn't matter what the site offers. That being said, we still are trying out new ideas, new features, and bug fixes. It's difficult to advance the codebase of the site without adding more clutter. We have unfortunately neglected performance while we've been building new cool stuff, but, as usual, we're about to make another pass to make things super lean again.

    • one year ago
  14. cshalvey
    Best Response
    You've already chosen the best response.
    Medals 1

    I quite literally agree with everything @mattfeury said. Which is impressive. Literally.

    • one year ago
  15. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    Are you sure you mean literally? Or, literally?

    • one year ago
  16. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    All righty, I'll jump in for funzies. Here's a couple of points to bear in mind: - We do see-saw a bit on performance. There are moments when things get worse for a short time as we address new performance issues that come up as we get more traffic. The reason is because, as Matt pointed out, there are three of us. We can pretty much either build new stuff or work on performance, but rarely both at the same time. So we deal with performance issues as they come up. - There is no ruder or more frustrating thing when bringing up issues with a site than acting like you know what the site's internal architecture is like. While I've posted the occasional high-level overview of the site's architecture, and it's possible you've read some of them, please don't make feature recommendations based on this vague understanding. There are possible things that could be done to improve throughput related to chats, but 95-99% of throughput issues on this site are garbage collection problems that produce 30+ second server pauses and require a server restart to start with a clean slate. The proper approach to improving performance is twofold: (a) split server processes so that any given process can use a smaller heap, and therefore even bad GC pauses can be shorter, and (b) reduce longer-lived objects so that the generational collector can grab them before a full GC has to happen. I've been working on (a) all week, and (b) will follow as we'll have smaller heaps that we can analyze more easily to see where garbage is being generated. - It's not an ISP issue. - Chats may or may not be consolidated in the future. It's something we've discussed. But we don't change features that we consider important because performance is bad, we improve performance so features we consider important can work better. We consider chat important because it builds the community as a place that people can idle in addition to learn. We think that's important, because we think it increases the likelihood that people will stick around even if they're tired of helping for a while, and therefore integrate into the greater community at large. This is a place to learn, but learning doesn't have to suck, and breaks are good. The per-group chats make things a little more manageable from a user perspective, even when math is the most active chat by far, and they do provide some guidance as to starting points for discussions. They do not have an extraordinarily large performance impact. Certainly not one that you've noticed. - We are not your slaves, and we don't appreciate demands. We're providing a service, for free. We know that this doesn't absolve us of a responsibility to make the service rock, and we think our track record on making OpenStudy rock is pretty good. We know the performance issues are frustrating; we use the site, too. We hear you when you complain, and we work to make the experience as good as possible within our time and financial constraints. TL;DR: This has nothing to do with networking, everything to do with large (10-15GB) heap garbage collection, we're working on it, and please don't make demands. PS: Your screenshots are, unfortunately, quite meaningless. You ran tracerts and pings, both of which rely on ICMP packets. We run on Amazon's EC2 platform, and ICMP packets are blocked by default, so these two strategies to determine server connectivity will fail every time. We use more effective strategies internally to be notified of when there are significant connectivity issues. This morning's issues were not connectivity-related, they were responsiveness-related.

    • one year ago
  17. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    First let me say I'm very impressed with the responses from all 3 of the purple people (unofficial term) who posted above. Very good replies both @mattfeury & @cshalvey and I agree strongly with you both. Second, let me talk to I think you took personal offense to the way I worded this, and perhaps you were frustrated by the situation as well (reading over my text I wrote now and trying to put myself in your shoes, I can see how you interpreted what I wrote as something personally against your hard work). This is not the case, I'll try to explain. I come from a background (this is my bias) of continuously seeing developers in the game industry who have gotten into a work culture that is lazy (not placing blame on anyone specifically, but it's true in repeated cases that are often VERY noticeable by the customer) and/or constrained by their corporate hierarchy's bureaucratic policies (i.e.: Electronic Arts, Perfect World Entertainment, etc.) I have the utmost respect for independent developer teams and I can point to many, MANY instances of incredible feats done by these smaller groups of passion-driven individuals. Whether it be Minecraft (which also uses the JVM btw) or even this website, itself you all are miracle workers! I meant it. Bravo! That said it's also sometimes hard when you're on the developer/creator side of the problem to see what the users who are creating yourself are experiencing. It becomes very personal. I remember the first weapon code scripts and animations for them that I did were hammered with criticism and I took it personally. What took me awhile to understand was that what people were saying about what I made wasn't to be taken as whining or demanding but rather as, what Stephen Covey says in his 7 Habits of Highly Effective People, as an opportunity to sharpen the saw. If I did come of "demanding" in my written tone, this was not my intention. You are nobody's slave, and you are in charge of your own domain. Likewise you're also in charge of the success or failure of your domain, and it is in this that much is to be gained from customer feedback. It's a real art sometimes to knowing how to analyze feedback, and a lot of logic. My personal favorite tool for this is the root-cause tree diagram and giving "weighting" to each thing that is said. You can then take a look at this diagram and assess what can reasonably be done to take corrective measure and what things people need to just not whine about so much. It's almost as if there's a balance between trying to do everything the customers are wanting and doing the things you can in the time allowed. The key (and it's waaaay easier said than done) is to find what's best for the community as a whole. A community you can take pride in because it was YOUR team's code that created it! Again that, is one of the amazing miracles of being the developer of something that brings people together. :-) Aside: and 3 people is crazy for something this massive, I can only think of the initial works of Minecraft and the game Natural Selection (now their own small company called Unknown World Entertainment) as parallel feats of developer passion. Truly impressive, and no I didn't know there were only 3 of you, nor that you weren't paid D-: IMHO, you deserve a bonus ladies & gents! Now to address your specific points @shadowfiend ... (yes, I still have more to say but I need to break this post up into two pieces I think in case y'all have anything you want to write while I'm writing the next post)

    • one year ago
  18. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    (Haha, fair. It did come off as a bit demanding, but I didn't take personal offense to it. Just wanted to be clear. And we fully agree about customer feedback—that's why this group is here, and I really do think our track record on that has been pretty stellar, too. I promise we're not lazy ;) Oh, and for the record, we *are* paid. We just don't charge our users anything :P)

    • one year ago
  19. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    Actually first to respond to your post-script @shadowfiend: I was at first going to try to use the hping tool to get around the IMCP blockade with different protocols (TCP, UDP, ICMP and RAW-IP protocols for example) that many servers will respond to, but I didn't want to play games with your firewalls nor did I want to risk leaving a log of me having been there which could be possible misconstrued as me somehow attempting to hack the website. I simply used the basics and posted the images (after blanking out my IP address, which I know you all can see). More importantly I didn't want to give any bad knowledge to the OTHER users when I posted this topic/question publicly. Make sense? ;-) I don't want to divulge more than this, but hex codes can be fun and I know more than I should. Enough said. You make a fair point about the limitations of the proof provided to apply to this case, thus I am incorrect by incompleteness myself.

    • one year ago
  20. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    In short, I showed respect & restraint for your sovereign domain, curiosity wasn't worth it. The golden rule, essentially :-D PS: Curiosity is also said to kill cats.

    • one year ago
  21. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    In that case...be more curious!

    • one year ago
  22. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    "There is no ruder or more frustrating thing when bringing up issues with a site than acting like you know what the site's internal architecture is like." 1. Actually I can think of several thing ruder, but I can agree with your point of view that it's frustrating. It's equally frustration for users, and from being a server operator in many cases I can tell you that customer impression IS important. Very important. So it's a double-sided sword, in a sense. "While I've posted the occasional high-level overview of the site's architecture, and it's possible you've read some of them, please don't make feature recommendations based on this vague understanding." 2. Yes & no, it's a bit more complicated than that... Suggestions are needed, not all are feasible for time or money reasons. In short, depends on prioritization and objective goals. (also a case of easier said than done eh?) "There are possible things that could be done to improve throughput related to chats, but 95-99% of throughput issues on this site are garbage collection problems that produce 30+ second server pauses and require a server restart to start with a clean slate." 3. Whoa... that's too long... Yeah JVM is terrible for this in large-scale applications, and a monstrous pain to debug the Java code and optimize it. That's why most large Minecraft servers for example use a 3rd party server-side modification of the source code (i.e.: Bucket), because there are so many chunks to process when some player is traveling far out or into new territory (see: http://www.minecraftwiki.net/wiki/Chunks). Often times you will have some MASSIVE lag on the server, (in game terms massive lag is >0.5 seconds of freezing up repeatedly and/or instances of >2 seconds lag-out freezes) Other times it can be from things like a lot of data is processing, for example look what happens when somebody blows up a mountain of TNT ( http://www.youtube.com/watch?v=pP2E4JpOBzo ). There is where, from my experience, I felt it could be potentially a physical systems resources issue if not ISP (and you said it wasn't ). I realize Minecraft may be a crude comparison, but it does have certain parallels in this case even if the JVM is being used in VERY VERY different ways. Same concrete foundation layout, different buildings entirely, if you will. This foundation (garbage collection routines causing lag and needing optimization for instance, is definitely in common with both OpenStudy servers and Minecraft multiplayer servers) Heck I don't know, maybe email Notch and see if he's got any helpful tips? He's a big fan of Java, and he might be able to give your some ideas your team hasn't thought of yet. If you can sell him on the noble mission of this website he might even give you his professional advice free of charge :-D Ya never know with people & the power of persuasion.

    • one year ago
  23. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    @rebeccaskell94 , no, I happen to like the people here. :-) There's a time and a place for such measures and this is neither the time nor place. To relate it in common terms, it would be as inappropriate as me walking into a formal business meeting and pulling down pants to examine the underwear people are wearing. It's rude.

    • one year ago
  24. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    Given that splitting out the application resources into multiple processes is a viable approach at this point, we're going to pursue it and use the smaller heaps to do better analysis of our own memory usage before wasting anyone else's time. There's plenty of GC tuning that could be done (and plenty we've already done), but at this point it seems pretty clear it's boiling down to (a) generating too much garbage that's being retained too long and (b) not being able to analyze a 10-15GB heap snapshot with the tools we have available. Splitting out will also help us deal with a bunch of potential issues down the road, so it's a good investment for a variety of reasons.

    • one year ago
  25. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    I was referring to cats...

    • one year ago
  26. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    I will say, I have never seen an architectural recommendation given with minimum knowledge that has been helpful. In conversation, yes. In posts like this, no. There just isn't enough context to properly interpret the possible issues when you haven't engaged with the people who are running the system, and people spend time refuting recommendations instead of dealing with the true problem.

    • one year ago
  27. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    We're also in regular touch with some pretty top notch folks who have been dealing with the JVM for a while, and their conclusion has been the same as ours for all intents and purposes :)

    • one year ago
  28. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    "The proper approach to improving performance is twofold: (a) split server processes so that any given process can use a smaller heap, and therefore even bad GC pauses can be shorter, and (b) reduce longer-lived objects so that the generational collector can grab them before a full GC has to happen. " 4. For part (a) it sounds similar to what Google does with search queries and their custom-Linux-based server farms. I've often wondered how they do that exactly, but never had the time or specific interest to really look into that low-level programming stuff. (I tend to be concerned with the higher-level source code objects, and know the low level networking aspects only because of experience and the need to know combined with my curiosity pushing me further). And like I said earlier, there's always somebody who knows more than you. For part (b) that definitely sounds like an optimization algorithm, and as crazy as it sounds the guys (and gals) who work on the Minecraft code might just have a tip y'all might want to try. "I've been working on (a) all week, and (b) will follow as we'll have smaller heaps that we can analyze more easily to see where garbage is being generated." (a) +1/day Kudos for you :-D (b) Have you considered: * old threads/questions? * repeat questions? * people who spam in the chat? * messages of "first question" posted to new chats? * user account database (what do you use? xml? implicit? sql? I'm a bit curious now) * refresh queries? * search engine web-bots (i.e. Google's webcrawler)? * nested commands? * recursive procedure calls that may be unnecessary? I guess part of the question is where is the restraint? Bandwidth, memory, processing, or a combination of the above. (You've probably already reviewed this but it doesn't hurt to ask I guess)

    • one year ago
  29. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    I feel this is definitely hitting the nail on the head with this one, bravo indeed! "(a) generating too much garbage that's being retained too long and (b) not being able to analyze a 10-15GB heap snapshot with the tools we have available. " Part of the problem is in why... (and yes a fresh reboot does solve it for a time, but it keeps filling up) ^_^ I wonder, is it possible to use a system that doesn't use Java? o_O Or is that simply out-of-scope of feasibility.

    • one year ago
  30. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    The current system is optimized to use weak references to all database information, unless we're passing it around in messages. So just about every situation where we look up actual data is generally subject to GC and is looked up again as needed if the GC collects it. Chances are the GC issues are due to inadvertent retentions, be it because of a Scala closure closing over a non-weakened reference or something of the sort. Search bots are a possible but unlikely source of pain. Most of the stuff they hit is optimized to run through stateless paths that don't trigger server-side sessions.

    • one year ago
  31. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    "I will say, I have never seen an architectural recommendation given with minimum knowledge that has been helpful. In conversation, yes. In posts like this, no. There just isn't enough context to properly interpret the possible issues when you haven't engaged with the people who are running the system, and people spend time refuting recommendations instead of dealing with the true problem." Good point. "be it because of a Scala closure closing over a non-weakened reference or something of the sort." I'm kind of scaring myself that I actually understood that... >_<

    • one year ago
  32. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    The Lift framework we're using is pretty much second to none for building a system like ours, and the JVM is second to none in terms of VMs we could be running on. I don't see a situation where an equivalent code base on a different VM would be doing any better, except insofar as we would probably have been forced to split out into separate processes much sooner. So it's the JVM's fault we've gotten so far without having to deal with this problem, which is one of its pluses, not its minuses :)

    • one year ago
  33. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    You're scaring me too.

    • one year ago
  34. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    Fun fact: we once tried building our widget code on Node.js. It crashed and burned at maybe a quarter of the load we were running on the JVM (probably in large bart because the JVM threads, so we don't have to worry about firing up separate processes to scale on the same machine; either way, that was a fraction of the page serves we're seeing now for the widget). Being completely stateless, our widget code (still on the JVM) runs on a smaller instance than the main OpenStudy site and is still serving 3 million pageviews a month without basically any work from us in the past 6-9 months.

    • one year ago
  35. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    Oh man that sounds painful... >_< Like a car that's lit on fire and then rolled off a mountain side... You really got me wondering how your server(s) handle this on their side... A few things can be gathered intuitively like the LaTeX/TeX code activation on the client's side and the Draw tool being partly on the server side (and I did see that an eraser feature was something you all were considering the feedback if I recall correctly, which is excellent for geometry stuff). "So it's the JVM's fault we've gotten so far without having to deal with this problem, which is one of its pluses, not its minuses :)" Haha! Great point. I suppose that is a positive. The purpose was accomplished sufficiently for the task up to this point. Although I must bring us back to topic though and point out that the customers are finding it to be aggravating when it lags as severely as it happened before. Computers only care about processing the data, doesn't matter what it is, they don't care. People on the other hand are bothered by lag, even if only at a subconscious level. Now some lag is simply too short for people to notice, or only happens rarely it's just "meh" once in awhile. I guess what I'm saying is your team has really really got your work cut out if it's garbage collection optimization. One thing we tried with a 1000-player Minecraft server that had issues like this was a once-every-few-hours auto-reboot, or rather it was a partial reboot of the subroutines that got so bloated over time. I don't suppose having more RAM would help eh? :-D Maybe create a Donate to Reduce Lag fund? :-P lol Btw, to be clear, we're talking about the heap being bloated to the point the site is crawling on it's hands & knees right? And is it the live or the dead objects cause most of the issues? Can you tell with the tools you have?

    • one year ago
  36. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    I know I'm talking in general here, I'm just curious as to what is making the hardware want to cry in pain. ^_^

    • one year ago
  37. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    It's heap bloat, yes, but specifically heap bloat with objects that aren't collected in young generation collections. That's what triggers the long full GCs. Full GCs tend to reclaim ~4-5GB of space, which is ~1/3 of the heap size, so we can assume there are plenty of dead objects, they just don't die fast enough. Unfortunately tuning the GC properly for that is non-trivial, as there are tradeoffs involved in waiting longer before promoting objects to the tenured generation. More RAM would help, but only by delaying the inevitable. We're aware of the severity of the issues, which is why we're working on it right now.

    • one year ago
  38. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    And the consequences/side-effects for shortening the lifetime of the objects too much with the current system would likely be? You know my first suspect earlier was the short-term variables, which is why I made the educated guess about chat. But now I can see it's more complicated than just that, it's an entire site code problem, and one which will require much pizza and headache medicine on late nights to solve. PS: You're probably now one of my favorite developers, genuinely. It is sadly uncommon to find developers so open and willing to discuss design functionality elements with a "regular" user. :-) Those that do have my undying loyalty to their works of art.

    • one year ago
  39. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    Oh, shortening the lifetime is the obvious goal, it's just not as straightforward to figure out what is causing longer object lifetimes. That's what we need proper heap analysis for, but the tools generally start dying around 10-12 GB heaps, and ours runs at 15. And it's not as useful to run the analysis on the smaller version of the heap, because the retention situation tends to be a bit different in those cases than in the production situation.

    • one year ago
  40. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    Basically it's likely that there's someone retaining objects here or there that we don't realize is retaining them. Once we figure out the culprits, by analyzing a heap snapshot for objects that are being retained, we can figure out how we want to reduce the retention (usually we use a RestoringWeakReference, which is a utility class we've rolled that encapsulates a WeakReference and a function to rehydrate the reference if it gets lost and someone tries to use it: https://gist.github.com/2580081 ).

    • one year ago
  41. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    Open source is in my blood, talking about what we're building is something I'm always willing to do :)

    • one year ago
  42. cshalvey
    Best Response
    You've already chosen the best response.
    Medals 1

    Favorite developer, eh @agentx5? I presume I remain as favorite Community Manager? I'm not sensitive or anything, but I'd be totally crushed if I wasn't...

    • one year ago
  43. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    Colm is always gonna be my favorite >.> c; xD

    • one year ago
  44. lgbasallote
    Best Response
    You've already chosen the best response.
    Medals 0

    side question...what's QoS?

    • one year ago
  45. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    Aww @cshalvey ^_^ lol @shadowfiend 15 GB heap?! That's as much as that guy in the Minecraft TNT video said it took to lock up his personal supercomputer >_< Yeah well if you need any additional RAM sticks, let me know what type and maybe we can see if we can get a community effort on this to fill in while you go after the core of the issue. Extra RAM is never a bad thing, I think :-3 And it's nice to have a believe in Open Source on the team, part of what makes some of the really good software that ticks out there (i.e.: Wikimedia & Firefox) stay ahead of the curve. Maybe you could talk some sense into these guys then eh? http://crypticstudios.com/ & http://easystudios.se/ :-P They seem to have gotten into the thing about making products you have to "gamble" money on to make a profit, instead of the spirit of exploration. In fact they are so close source the communities have to do the actual research on the game's mechanics. An example is this wiki run by the lady "Lohr the Wikimistress" who made almost all of this on her own, no help at all from the developers, and she relies on volunteer nerds to keep it up to date: http://www.champions-online-wiki.com/wiki/Main_Page Now do you see why I had the bias I did when I first made this thread? >_< It should also be noted that most of these programs they make (the games) make you digitally agree to legal agreements that basically allow them to change anything you have, even stuff you paid real money for, and nerf the hell out of it or even delete it because they felt like it and you are owed no explanation or compensation (and then they wonder why their communities are enraged). For me, both fun and learning go hand and hand (although this website is much more about learning than just having a good time, which is perfect for it's place in the world of the internet, fills a needed much-needed niche), and when the spirit of purpose is lost it makes me sad. I sincerely hope this website continues to proper and MY purpose, the spirit of this concern I have created here in this thread, it to do my part to help that prosperity continue and even grow into the future. :-)

    • one year ago
  46. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    Quality of Service.

    • one year ago
  47. cshalvey
    Best Response
    You've already chosen the best response.
    Medals 1

    QoS = Quality of Service I would assume, correct? And... beaten to the punch by shadowfiend... pwned

    • one year ago
  48. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    I kinda feel like I might look smart if I stay on this post.

    • one year ago
  49. karatechopper
    Best Response
    You've already chosen the best response.
    Medals 0

    Lol same with becca:)

    • one year ago
  50. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    lol yep he beat both you and I to the punch @cshalvey in answering @lgbasallote's question :-D

    • one year ago
  51. lgbasallote
    Best Response
    You've already chosen the best response.
    Medals 0

    ohh thanks.. @rebeccaskell94 you'll feel smarter here http://mail.openstudy.com

    • one year ago
  52. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    I'm not planning on taking so long for the split-process stuff that we need more RAM @agentx5 , no worries ;)

    • one year ago
  53. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    I'm not clicking that link...

    • one year ago
  54. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    ^_^ But RAM is tasty! (jk, I get it)

    • one year ago
  55. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    In a perfect world we'll be spinning up new instances instead of more RAM soon.

    • one year ago
  56. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    Is that a pun on the Perfect World developer group? (i.e.: The Chinese company that owns Cryptic Studio that I linked above) Or am I just reading into that too much? lol

    • one year ago
  57. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    PS: If people are wondering where InsightBB is, it's Insight Broadbrand, my ISP, located in Louisville KY. (not that you couldn't find this with a few minutes of searching on the internet, there's millions of people around me, you're not going to find me) :-3

    • one year ago
  58. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    Lol. Definitely reading into it too much :p

    • one year ago
  59. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    Yeah it's been a long day between Calculus 2, doing some consulting work, and this topic... I think I'm starting to have neurons in my brain have their own heap size overloading issues :-D

    • one year ago
  60. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    Hahaha. That's what breaks are for :p

    • one year ago
  61. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    Sounds brutal.

    • one year ago
  62. agentx5
    Best Response
    You've already chosen the best response.
    Medals 3

    Breaks? What are those mortal? :-3 (I'm kidding, I'm human too and yes I should have taken one today besides a 10-minute lunch)

    • one year ago
  63. Agent_Sniffles
    Best Response
    You've already chosen the best response.
    Medals 2

    @rebeccaskell94 You're right..lol I feel smarter being on here. But, I think I know a lot more by reading all this. o_O And no. No killing cats, unless you wanna eat them. ;D

    • one year ago
  64. Agent_Sniffles
    Best Response
    You've already chosen the best response.
    Medals 2

    Baahah! That's racist... xD

    • one year ago
  65. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    Sniffles that was rude. You shouldn't say stuff like that about the poor kittens

    • one year ago
  66. Agent_Sniffles
    Best Response
    You've already chosen the best response.
    Medals 2

    Bahahah! I'm gonna get you back Becca!

    • one year ago
  67. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    hahahhahahha lmfao omg rofl xD I'm literally laughing omg

    • one year ago
  68. shadowfiend
    Best Response
    You've already chosen the best response.
    Medals 6

    Take it to chat folks.

    • one year ago
  69. rebeccaskell94
    Best Response
    You've already chosen the best response.
    Medals 1

    Yes sir.

    • one year ago
  70. JackJones
    Best Response
    You've already chosen the best response.
    Medals 0

    @_@ utterly confused

    • one year ago
    • Attachments:

See more questions >>>

Your question is ready. Sign up for free to start getting answers.

spraguer (Moderator)
5 → View Detailed Profile

is replying to Can someone tell me what button the professor is hitting...

23

  • Teamwork 19 Teammate
  • Problem Solving 19 Hero
  • You have blocked this person.
  • ✔ You're a fan Checking fan status...

Thanks for being so helpful in mathematics. If you are getting quality help, make sure you spread the word about OpenStudy.

This is the testimonial you wrote.
You haven't written a testimonial for Owlfred.