1911 replies to this topic

[Klappspaten]

GalaxyBluestar, on 04 April 2014 - 11:24 PM, said:

really wish this stuff was in a more prominant place on the forums, such info from a dev/pgi worker is MOST invaluable. great communication happening here.

I agree, I dont understand a word, but I love it.

I request Karl Berg to be appointed as employee of the month!

[Hammerhai]

Thank you. It has cleared up one or two questions I had IF I got the techspeak correctly. And apart from that it sounds like real cutting edge stuff.

Now what are you going to do with your Skyhook version Alpha in the future, I wonder ...
Me, I cannot even modify Crescent Fresh's TARGET script for the Warthog without busting something I have no clue about, cos Timex Sinclair 1000/Spectrum.

I feel old.

[Hawkeye 72]

There are massive amounts of win in this thread!

Props to Karl for taking some time here and there to drop these nuggets of fun!

[Chronojam]

Karl Berg, on 27 March 2014 - 05:44 PM, said:

Yup, we run several virtual servers.

Yup, we run multiple dedicated server instances on a single physical server. The ratio is about 24 instances per blade, but we monitor blade utilization and adjust this as needed.

And yes, our matchmaker and server architecture will have basic support for regionalized dedicated servers starting with the April 1st patch. For future requirements lets say. As for actual regional deployments that would make use of this newly added support, I can't say anything at this time.

Virtualized server infrastructure, eh? A few of us were having a chat somewhat recently about server performance and virtualization regarding MechWarrior Online, actually. What it came down to was a lot of guesswork about resource use, particularly the theoretical overhead of having an MWO/Cryengine game server instance sitting idle on a virtualized platform.

How does resource use vary between 0 (empty) and 16 players? What about when you made the +50% jump to 24 players, does it scale well?

[Suko]

Karl Berg, on 03 April 2014 - 12:59 PM, said:

Quick update on TrackIR: great success has been achieved, NaturalPoint has been extremely supportive and helpful, so many thanks to Seth and the others over at NaturalPoint! The branch is now in the hands of the Gameplay team.

I'm super excited for this. Hopefully the Oculus support isn't too far off. I would love to use my dev kit with MWO!

Edited by ShadowVFX, 07 April 2014 - 10:54 AM.

[Karl Berg]

Chronojam, on 07 April 2014 - 09:47 AM, said:

Virtualized server infrastructure, eh? A few of us were having a chat somewhat recently about server performance and virtualization regarding MechWarrior Online, actually. What it came down to was a lot of guesswork about resource use, particularly the theoretical overhead of having an MWO/Cryengine game server instance sitting idle on a virtualized platform.

How does resource use vary between 0 (empty) and 16 players? What about when you made the +50% jump to 24 players, does it scale well?

We don't virtualize the dedicated game servers. That's one of the few components we only use bare-metal for. Early on, during F&F / closed beta, we ran a bunch of virtual vs bare metal tests for player experience, and unfortunately it was quite noticeable, unless we severely under-provisioned. It sucks in terms of data-center costs, but asynchronous realtime multiplayer is hugely latency sensitive, and a good experience absolutely requires we minimize latency everywhere we can.

[Heffay]

Karl Berg, on 07 April 2014 - 11:43 AM, said:

We don't virtualize the dedicated game servers. That's one of the few components we only use bare-metal for. Early on, during F&F / closed beta, we ran a bunch of virtual vs bare metal tests for player experience, and unfortunately it was quite noticeable, unless we severely under-provisioned. It sucks in terms of data-center costs, but asynchronous realtime multiplayer is hugely latency sensitive, and a good experience absolutely requires we minimize latency everywhere we can.

Oh wow, with 24 instances per bare metal, even having a couple of dozen private matches going on can suck up the hardware pretty quick.

We use a lot of Cisco UCS devices where I work (as well as HP chassis). I don't envy your hardware and support costs!

[Chronojam]

Yeah, that really explains the concerns over private matches, and is a little surprising. In light of that, it might make sense to have a real waiting queue for private matches, although who knows what demand for them is even like these days.

[Tekadept]

Chronojam, on 07 April 2014 - 03:24 PM, said:

although who knows what demand for them is even like these days.

The queue would be something like this I imagine

[Hammerhai]

That does explain a lot about private matches. Bummer is that in light of the new data on premades one does wonder ...

It also points at premades either being vilified to heck and gone for nothing in CB, or having left the game.

All in all though, while I was cynical about charging for private matches, I now accept the necessity. I hope others will, as well.

At the end of it I would go so far as to say PGI is being very reasonable in listening to public demands on this.

[Roadbeer]

Speaking for myself, I fully understand that there are costs attributed to the Private Match system, and I think that requiring only 2 premium accounts is very reasonable.

It's the lack of rewards that bugs me. I'm paying for this match, let me have my cookies. Epeen isn't everything.

[Heffay]

Roadbeer, on 08 April 2014 - 09:26 AM, said:

Speaking for myself, I fully understand that there are costs attributed to the Private Match system, and I think that requiring only 2 premium accounts is very reasonable.

It's the lack of rewards that bugs me. I'm paying for this match, let me have my cookies. Epeen isn't everything.

How do you prevent people abusing that system though?

[Roadbeer]

Heffay, on 08 April 2014 - 09:48 AM, said:

How do you prevent people abusing that system though?

Sorry, it's the small L libertarian in me. Don't penalize the 99.99% for what the .01% might do.
I haven't received an answer back, but I asked Russ on Twitter if the results of Private Matches will impact Elo. I see the ability to game that a much bigger offence and damaging to the overall game than a glut of Cbills and XP would be.

[Heffay]

Roadbeer, on 08 April 2014 - 11:04 AM, said:

Sorry, it's the small L libertarian in me. Don't penalize the 99.99% for what the .01% might do.
I haven't received an answer back, but I asked Russ on Twitter if the results of Private Matches will impact Elo. I see the ability to game that a much bigger offence and damaging to the overall game than a glut of Cbills and XP would be.

Gotta plan ahead. If or GXP or other resource all of a sudden becomes a valuable commodity (unlocking tier 5 weapons modules, for example), people are going to be abusing the hell out of being able to farm them.

This isn't theoretical. This happens in every game, where people find the easiest method to accumulate resources, especially if you can do it at a rate 2-3 times faster than normal.

"LFM: GXP farming group! Get max GXP in a match every 3 minutes. 4 more slots available!"

[Rasc4l]

Heffay, on 08 April 2014 - 09:48 AM, said:

How do you prevent people abusing that system though?

On quick thought:

To prevent a couple of friends jerking the system over: No rewards unless at least 4 lances of the max 6 is present (16 players).

To prevent organized groups from jerking the system over e.g. by fielding 12 vs. 12 circle-jerk practice: No rewards in 17-24 player matches if all the players are registered under the same group or are unregistered. Meaning there must be at least one person from another group (clan) present in the match. This still allows some cheating if some cowardly groups would go for it but it should be easy to detect as PGI can watch matches from player POV. This solution might therefore have to perhaps tolerate some little amount of cheating but going with Roadbeer's libertarian thought and with the fact that the cheaters still can't steal skill I would let it slide in sight of the greater good.

[Heffay]

Rasc4l, on 08 April 2014 - 12:05 PM, said:

On quick thought:

To prevent a couple of friends jerking the system over: No rewards unless at least 4 lances of the max 6 is present (16 players).

To prevent organized groups from jerking the system over e.g. by fielding 12 vs. 12 circle-jerk practice: No rewards in 17-24 player matches if all the players are registered under the same group or are unregistered. Meaning there must be at least one person from another group (clan) present in the match. This still allows some cheating if some cowardly groups would go for it but it should be easy to detect as PGI can watch matches from player POV. This solution might therefore have to perhaps tolerate some little amount of cheating but going with Roadbeer's libertarian thought and with the fact that the cheaters still can't steal skill I would let it slide in sight of the greater good.

People would just organize on the Comstar or NGNG teamspeak servers. In fact, that's how most of them would be run anyway. The odds of an organized unit doing it strictly in house is pretty low. But the number of private matches full of people farming would be significant.

Elo shouldn't change for exactly the same reasons. Make a throwaway account, then help farm your friends up to 2500.

[Rasc4l]

Heffay, on 08 April 2014 - 12:18 PM, said:

People would just organize on the Comstar or NGNG teamspeak servers. In fact, that's how most of them would be run anyway. The odds of an organized unit doing it strictly in house is pretty low. But the number of private matches full of people farming would be significant.

You're right, it would be bad. I'm just trying to find out some general restrictions to make it possible that you still earn the rewards from 12 vs. 12 private matches between organized groups. I would hate it if one needed to go so far as to having to register each clan match at PGI to get the cash. I'm already easily paying 120 000 per match as a light player (UAV+arty+air) in organized drops and even when we win, I might still lose cash.

[Heffay]

Rasc4l, on 08 April 2014 - 12:23 PM, said:

You're right, it would be bad. I'm just trying to find out some general restrictions to make it possible that you still earn the rewards from 12 vs. 12 private matches between organized groups. I would hate it if one needed to go so far as to having to register each clan match at PGI to get the cash. I'm already easily paying 120 000 per match as a light player (UAV+arty+air) in organized drops and even when we win, I might still lose cash.

Maybe 1/4 rewards. Figure out what the maximum rewards you could get, then divide that result by the desired rate. Throw in another fudge factor to encourage people to join the public queue, and then maybe you have a deal. No Elo adjustments.

[Roadbeer]

Rasc4l, on 08 April 2014 - 12:23 PM, said:

You're right, it would be bad. I'm just trying to find out some general restrictions to make it possible that you still earn the rewards from 12 vs. 12 private matches between organized groups. I would hate it if one needed to go so far as to having to register each clan match at PGI to get the cash. I'm already easily paying 120 000 per match as a light player (UAV+arty+air) in organized drops and even when we win, I might still lose cash.

Well, Russ stated on Twitter that rewards in the 12v12 Free Private matches were possible, but I'm looking at that as "Hey, you get to select your opponent instead of roll your dice in the 12v queue (which they anticipate doing away with) so, here's a cookie.

Thing is that the 12v isn't entirely representative of the smaller large group queue, who just want to play with their friends, are paying for that privilege (with Ptime, and MC down the road) and are still penalized with Consumable prices AND not being rewarded.

A segment of the population who have grown tired of having their success/failure determined by the Derpherd, yet, if they want to advance in any form in this game, are forced to do so.

Sorry, I didn't mean to hijack this thread, I'm having this conversation elsewhere on the Forum, feel free to engage me on the topic there.

[GalaxyBluestar]

Karl Berg, on 07 April 2014 - 11:43 AM, said:

We don't virtualize the dedicated game servers. That's one of the few components we only use bare-metal for. Early on, during F&F / closed beta, we ran a bunch of virtual vs bare metal tests for player experience, and unfortunately it was quite noticeable, unless we severely under-provisioned. It sucks in terms of data-center costs, but asynchronous realtime multiplayer is hugely latency sensitive, and a good experience absolutely requires we minimize latency everywhere we can.

Chronojam, on 07 April 2014 - 03:24 PM, said:

Yeah, that really explains the concerns over private matches, and is a little surprising. In light of that, it might make sense to have a real waiting queue for private matches, although who knows what demand for them is even like these days.

this is why i nominate this as the most important and highly prized thread on this forum or in the history of PGI communication.

CJ a staunch hater who has conducted his own library on the failing misgivings and miscomunications which has lead to missunderstandings rumour mongering and general hatred that has put pgi on the ropes is now finding solice with information from a direct source, who is allowed to speak his mind and explain often enough to give details which leads to a satisfing outcome for producer and consumer alike. Promote this thread to the command chair and let it be a QA for employes to take their 10 minute break from the focus of a task just to chill out with details.

this is what should be happening on a daily basis much better than twitface material.

so pgi fox once hinted at 21K unique logins for an weekend's duration of the current queue setup. is it possibly we could have a best day worst day figures on that front? or any plans with anticipating the numbers distribution as far as public play -> faction play -> private matches are concerned so far as playerbase distribution? would love to hear the viability of the three queues split considering 1pv only queue didn't happen and that was just 1 queue split over the 4 game modes that would've been available. now we're talking about skirmish too so 3modes and 1 queue {pubbie queue} with an additional 3 modes and 1 queue {faction play} and whatever for private matches. a split of 7 instead of 4 now, how will MWO cope!?