Life In 19x19

For what it's worth, KGS has been lagging these days for me as well...and I am on fiber.

That said, the Windows clients for Tygem and Oro are a step-and-a-half away from being malware and have their own problems. I do like that it is quite easy to find a game on those servers because they can handle more users. So I do acknowledge that being able to handle lots of users is important.

I am actually quite impressed that wms managed to code a client that has some features not found in other clients as well as features that are better implemented than they are in other clients. The platform independence is a huge plus for a Linux user like me as well. The flawless unicode support (somewhat problematic on Tygem) tells me that wms knows what he is doing. It's not just a Java thing. The Oro site has some text-based lectures that run in Java plug-ins and the text isn't proper unicode (same issue with Oro's Java client...Oro also has a native Windows client). It is one of those little things that reveal a lot about the programmer(s).

This leads me to conjecture that wms has outdated hardware or hardware that can't keep up with increasing loads.

I almost wish that Tygem would fire all their programmers and hire wms so that we could get wms' programming on Tygem server hardware.

I have had some experience working with software programmers in Korea, and I must say that, in my experience, they have a ton of bad habits.

This became a long rant but I guess I can state my most important thoughts.

1) I agree that KGS lags at times, even on the fastest Internet connection.
2) People should take your complaints seriously, and should not talk down to you. I am not quite sure if the KGS defenders are talking down to you, but they do seem to be a little defensive. Maybe they're just tired after a long day's work. Miscommunication happens
3) You should also understand that KGS needs to first weed out the obvious possibilities before they handle your complaint. The reason why tech support people first ask if your computer is plugged in is because a significant number of people forget to do that. That should not be taken as an insult.
4) wms is a probably a better programmer than everyone who works for Tygem/Oro combined, so please insult his hardware if you must, but don't insult his software!

edit: BigDoug's post from which I quoted a few sentences below has been deleted for some reason.

You may certainly be a security analyst for a large multinational company, but I don't see evidence of a deep knowledge of networking in your message

Are you seriously expecting a thesis on how KGS should handle server connections?
You're avoiding the question by insulting my knowledge.

I must admit that this statement made me smile. None of this relates to networking (expect for possibly the generic bandwidth reference). These items do impact performance, of course, but that's separate from intermittent lag in this context (although there are obviously indirect influences)

You're giving arguments against yourself. You're basically saying:

"server design, database access, game client, server hardware and bandwidth DO have an impact on performance and they DO have an influence on lag but it's a separate problem"

Sorry but that's a big insult to my intelligence and I'm not going to argue about how big the impact of bad design is on server performance. You're finding excuses for a very simple issue that many servers face. And you put that responsibility on the users while you could simply acknowledge the problem.

By saying KGS works flawlessly and it's only the users' fault if there's lag, you're only hurting yourself. Acknowledge the problems and you will get help from the community rather than complains on random forums.

We know it's a huge amount of work to optimize a database and the client/server connections. We know it costs money to buy/rent a good server. But we are not expecting or forcing you to do it. We are not insulting you and the KGS developers. They've done an amazing work.

If you're not happy with KGS, then there are other options available to you. Good luck.

It's not that I'm unhappy with KGS. I just have no choice. It doesn't work for me and, from what I gather, for a number of other people as well.

I repeat, we are not insulting KGS. We just want the comments about changing ISP and computer to stop. They are source of unnecessary flaming and solve none of the issues.

Yesterday on KGS:

joeie1: I found a flaw in KGS
joeie1: Anyone want know it?
BigDoug: people press the enter key too often ?

I liked it so much, I put it in my game info

Hello,

sefo wrote:edit: BigDoug's post from which I quoted a few sentences below has been deleted for some reason.

I actually deleted my own message about 10 minutes after I posted it. I said some condescending things in that post which I regretted almost immediately after submitting it. I therefore deleted it, but you read it beforehand. Please accept my apologies for being condescending -- it was rude and I regret it.

I therefore had an idea. Based on my understanding of your comments, I'd like to explain some of the issues which can cause lag. I'm obviously not saying that these issues are the cause of lag each time. However, I'd like to explain why saying that KGS always causes lag is an over-simplification. Note that I'm going to be as high-level as I can, so each item may be technically incorrect in some areas. I'm choosing simplicity over technical completeness.

If you're using a wireless connection, each wireless transmitter sends out a signal on a given channel. If there are nearby transmitters sending out signals on similar channels, then the connections for people using those channels can be weakened or interrupted. For example, if wireless modems for two neighbours who are playing on KGS are transmitting on similar channels, each neighbour can experience lag depending on which signal is stronger at that moment. Some of the cheaper wireless modems have limited number of channels on which to transmit, which poses a problem if you live in flats with thin walls and/or floors.

If you're using software which uploads or downloads at the same time as you're using KGS, then the performance will be affected. For example, if you're playing on KGS and your computer starts to download a security update, you'll experience lag, because the programs are competing for access.

A firewall can introduce lag because it's making decisions on whether to allow or not allow incoming or outgoing traffic. This is actually quite common. Note that firewalls can exist both on your computer and on your ISP's routers and switches.

The modem which transmits to the ISP depends on various equipment (e.g., the physical line which carries the signal to the switch, how many of the ISP's customers are using the switch at that very moment). My aunt's internet performs poorly when it rains, so she would experience lag at that time if she played on KGS. In our house, we experience lag when there is a major event nearby, because our ISP diverts bandwidth to cope with the increased traffic from their customers at the event. If your using shared physical lines and your neighbour starts to download a large file (e.g., a movie), then your performance is going to be affected. (I can't remember if this is true for both cable and ADSL.)

If you don't live the US, then the signal from your computer must be transmitted to the US. Any problems with the routers en-route or with the transmission lines will impact the quality of the signal, thereby causing degraded performance (seen as lag on KGS). For example, here in Australia, they use undersea cables which are occasionally damaged by submarines and fishing nets. Network performance is degraded until the cable is replaced or repaired. Note that there are similar routing issues within the US (e.g., the recent storms in the US degraded network performance and even cause complete shut-down in some cases).

These are just some of the many issues which can cause lag. If your computer was directly plugged into the KGS server, then it would be much easier to diagnose lag. However, because of the many pieces which make up the internet's infrastructure, poor performance of any of them impacts the network performance. When I worked for the vendor, we used very expensive diagnostic equipment and engineers to diagnose networking issues and they were rarely easily solved.

From your perspective, you experience lag on your personal computer. From a KGS admin perspective, we regularly receive queries from people about poor performance and our personal connection is working just fine. It's not unusual for someone to complain about 10,000 ms response time (using server stats) at the same time that I'm experiencing 150 ms response time. In this instance, something is causing that poor response time, but it's unlikely that KGS is the cause, because I'm not experiencing the lag and I'm not receiving complaints from other people.

For these reasons, when someone complains about lag, the response of KGS admins is to try to identify the cause of the problem. Sometimes we can help and sometimes we can't. However, we generally start with questions about firewalls and what applications you're running, since it's not possible for us to diagnose issues with firewalls, routers or undersea cables.

The lag on KGS that began at 00:00 UTC on July 1st was caused by the leap second. Many Unix systems were affected, see e.g. http://www.theregister.co.uk/2012/07/02 ... _airlines/

It was fixed, as far as KGS was concerned, by a reboot at around 03:00 UTC on July 1st.

sefo wrote:It's not complex. Generally when you use a 10 years old, custom made game server that is updated once in a while by a single developer, you must rewrite it from scratch to catch up with your server increased activity, make it more secure, faster...etc.

It's not complex, and you're wrong in a pretty obvious way. KGS in fact was rewritten from scratch about 5 years ago. Since then, the server it is on has grown faster by a factor of more than 10, while the average user count has gone up by not quite a factor of 2. So why do you expect the server to be the problem?

As Nick posted, there was huge lag on July 1-ish because of the leap second bug that hit my system. In general, since the server moved from Texas to the midwest, there has been more lag; the new ISP's networking just isn't as well connected or as robust. I'm not completely happy with that but moving again is a lot of work and money, so I have to sit tight for a while.

sefo wrote:The lag on KGS is not caused by a bad connection on the client side. Nor is it a network or ISP problem.

sefo wrote:My diagnosis is simple and doesn't require any technical knowledge:

...

The problem probably comes from the server design, database access, game client, server hardware and bandwidth.

And from my non-technical diagnosis, I highly doubt it has anything to do with KGS itself -- but more likely some combination of local OS, wireless, and JVM memory issues.

I have multiple machines of different OS types on the same wireless network each logged into KGS at the same time. -- one experiences severe lag, another doesn't.


And from my technical diagnosis, I come to the same conclusions.

Changing wireless router (again) mostly removed the issues from my Macs.

Let's blame Java -- which wms is actively fixing, by the way (by rewriting in HTML5 instead of Java)

wms wrote:As Nick posted, there was huge lag on July 1-ish because of the leap second bug that hit my system. In general, since the server moved from Texas to the midwest, there has been more lag; the new ISP's networking just isn't as well connected or as robust. I'm not completely happy with that but moving again is a lot of work and money, so I have to sit tight for a while.

Well, that just about explains things.

We should have a KGS "better ISP" kickstarter or something. People are donating to Kaya.gs, so why not K.gs?

I'd like to point out, that lag itself is not the problem. The problem is the consequences of the lag - namely losing games due to a factor that has nothing to do with go and is utterly out of your control.

As far as the blame game goes, BigDoug's post about the long chain of possible sources of lag makes a fair amount of sense, but it doesn't explain why people have such lag problems on KGS and not on other servers.

From my perspective though, it's irrelevant where the lag comes from. Whether it's caused by submarines or sleepy servers, I think that KGS should do something about the real problem, and make sure that losing because of lag is not possible.

What I would like to see done, is for KGS to implement some form of lag detection. Would it perhaps be possible for the client and the server to compare timestamps?

daal wrote:I'd like to point out, that lag itself is not the problem. The problem is the consequences of the lag - namely losing games due to a factor that has nothing to do with go and is utterly out of your control.

As far as the blame game goes, BigDoug's post about the long chain of possible sources of lag makes a fair amount of sense, but it doesn't explain why people have such lag problems on KGS and not on other servers.

From my perspective though, it's irrelevant where the lag comes from. Whether it's caused by submarines or sleepy servers, I think that KGS should do something about the real problem, and make sure that losing because of lag is not possible.

Thanks for putting that in fewer words and simpler sentences. I hope they finally understand the problem now. It's exhausting to try and explain in long polite posts what is the problem.

To wms, bigdoug and others: I lose because kgs kicks me out of the game +70% of the time (extreme lag or plain disconnection) IT ONLY HAPPENS ON KGS. Is it clear enough formatted like this?

daal wrote:What I would like to see done, is for KGS to implement some form of lag detection. Would it perhaps be possible for the client and the server to compare timestamps?

That's not possible and probably subject to abuse. KGS already lets you exit the game and resume it later when *server/client problems* (I insist!) have less impact on the clients' connections.
Other servers just give you a few minutes to come back to the game or less you lose. It's because they don't have connection problems.

As xed_over pointed, I also believe it's an implementation problem. Java needs to go.
If the solution is on the work, you could just have said so and explain that the problem should go away when it's done.

sefo wrote:Thanks for putting that in fewer words and simpler sentences. I hope they finally understand the problem now. It's exhausting to try and explain in long polite posts what is the problem.

To wms, bigdoug and others: I lose because kgs kicks me out of the game +70% of the time (extreme lag or plain disconnection) IT ONLY HAPPENS ON KGS. Is it clear enough formatted like this?

You understand that your network communication to KGS takes a different path than to other servers and services, right? Do a tracert to KGS, and another to another server, and see that they aren't taking the same path.

It takes my connection, from my router to KGS, 9 hops. Only the first few are in common with the path taken to Microsoft. After that they diverge. If any hop in the path between me and KGS is faulty or slow, the entire path is slow.

The point is: just because it only happens with KGS does not make it KGS' fault.

AmyTS wrote:

sefo wrote:Thanks for putting that in fewer words and simpler sentences. I hope they finally understand the problem now. It's exhausting to try and explain in long polite posts what is the problem.

To wms, bigdoug and others: I lose because kgs kicks me out of the game +70% of the time (extreme lag or plain disconnection) IT ONLY HAPPENS ON KGS. Is it clear enough formatted like this?

You understand that your network communication to KGS takes a different path than to other servers and services, right? Do a tracert to KGS, and another to another server, and see that they aren't taking the same path.

It takes my connection, from my router to KGS, 9 hops. Only the first few are in common with the path taken to Microsoft. After that they diverge. If any hop in the path between me and KGS is faulty or slow, the entire path is slow.

The point is: just because it only happens with KGS does not make it KGS' fault.

I seem to be detecting a lag in communication.

Is the likelihood of faulty hops on the way to other go servers any smaller? No. If it's not KGS's fault, why don't we experience lag on other go servers?

Perhaps this is purely subjective, as I play more games on KGS than on other servers, but quite frankly, I have literally never suffered because of lag anywhere besides on KGS.

daal wrote:I seem to be detecting a lag in communication.

Is the likelihood of faulty hops on the way to other go servers any smaller? No. If it's not KGS's fault, why don't we experience lag on other go servers?

Perhaps this is purely subjective, as I play more games on KGS than on other servers, but quite frankly, I have literally never suffered because of lag anywhere besides on KGS.

And I rarely experience lag on KGS. It's almost as if you're actively refusing to understand that different routes are taken to each service on the Internet, and KGS does not own the switches and wires between you and their server(s). You obviously didn't bother to do a tracert like I suggested.

AmyTS wrote:You obviously didn't bother to do a tracert like I suggested.

The point is not that that daal should track down his cause of lag with tracert or whatever, but that KGS should implement a better lag handling protocol. Even with horrendous lag, games should not be be lost because of it. If Tygem et al can figure out a way to avoid lag losses, then why not KGS? It's a tricky problem, I'll grant you, and I can't think of a way of solving it, but clearly, its possible, as no one to my knowledge has ever complained of a lag related loss on Tygem or Oro.

Edit: one possible way - the client application measures the time between the opponent move appearing on screen and the move being chosen. This time gap is submitted to the server, which decides whether the user loses on time. It also sends down an updated time left value to the client.

Indeed. Time is kept on the server and sent to the client with each move. One problem with your proposed solution is a hacked client would allow people to have more time each move than they should have by telling the server they were lagging when they really weren't. Say for instance I see my opponents move, examine the position for 60s, then I move and my client tells the server I experienced 5s of lag when there was none. People could cheat the clock and it would be very hard to detect, let alone prove.