[Devblog] Tranquility Tech IV!

Moving the image transfer stuff to a separate server really explains why 90% of the game works and character portraits tend to be empty black boxes a lot of the time, and why it takes DAYS for my in game profile pictures to update after changing around my cosmetics and reposing for a new profile picture.
I hope you guys get around to improving that experience, though I guess there aren’t very many of us who enjoy opening up other people’s avatar picture and putting them in the 3-D full body view to see what they have on.

Glad to see those blazing fast reboots, I can’t wait to see the day where there is no down-time, just a somewhat laggier hour as each of the clusers reboot while the others cover down on their server load to provide a complete, 24-hour seamless Eve experience!
And good luck to the software team to setting things up to allow that to work without needing a monthly full reboot, too!

I hope you guys keep an idea in the back of your minds for an in-house front end access / anti-DDOS service hosted on CCP machines, or a Cloudflare potential alternative. Cloudflare has some some shady stuff in the past with blocking IPs and IP blocks due to certain folks having political views that they oppose, effectively preventing those from all cloudflare related domains from accessing any part of the internet where they cover. I would rather not reward such practices indirectly with my subscription, though I do understand why they are used and it’s a relatively small market for high quality and high volume providers.

This happens with just one client running, no orders at all, CPU goes up from 12 to 24% and the client just freezes completely for 2-8 seconds. How it freezes depends on how long ago I used the market window before (I don’t use it often).

Everything else is fine, verfied files, game runs on SSD, no issues elsewhere including logging in or (un)docking from Jita like so many have. It all runs smoothly, just not when opening market window.

2 Likes

Nice nerdy blog, thank you.

1 Like

 

I am a bittervet.
And thank you for writing up this kind of devblog so that we can confirm we still have tech nerds who has affection to the EVE ONLINE within CCP. Please keep it up.

PS. When is session change timer going to be removed ?

3 Likes

Probably never. This is a safeguard that all the all nodes involved have finished their work in moving you; this is the timeout when the other nodes can move ahead if there is no confirmation.

4 Likes

The big question is, if I use a VPN to reduce disconnections, do I still aim to come out to London, right?

And as a side note, isn’t there any plan to make EVE more tolerant on network, and lag a bit instead of flat out kick you?

Though it would be nice to queue commands during session chance, Undock already does that, I wouldn’t mind having the same for Logout

I congratulate the technical explanation but I and some people often got connection issue around once per day … since 2023 maybe

I suppose the more important importance is player’s experience &“inter”“face” not a proudful paper

Appreciating untiring efforts!

Ahhh this makes me so sad, though I vastly prefer the timer to that small period of time without it…

Is it possible that with further iterations on hard/software the timer may be reduced? Maybe?

I shall sacrifice three herons to Bob in case it may be so.

1 Like

Thank you CCP Explorer.
It’s sad we may need to be with the session change timers for the rest of our entire life but what happens if we don’t have the timers ? I guess we are shooting ships in both sides of the gate, one is real player, and another is ghost ? :slight_smile:

Also, I still don’t think we had a session change timer back in 2016 or so when I started the game. Were there timers since the birth of the game ? :thinking:

1 Like

Have you found a way to resolve performance monitoring with the parameter sniffing enhancements?

love the irony of your CCP name by the way hahah. Would love to hear what other instance and DB level configs you have applied that you can share

2 Likes

They been in for a long time

I’m shooting in the dark, but it seems that EVEs bottleneck could be the mechanics behind offence/defense/propulsion calculation system. Each shot or fleet movement is a freaking science project behind turret / missile or propulsion mechanics…

If EVE could get fix tidi - I guarantee more online / revenue

While the backend may have seen efficiency improvements, since the introduction of the new GUI, the client chews 20-25% more CPU and GPU and is definitely more sluggish doing almost anything especially listing inventory (no more than 20 items, mainly containers).

Additionally the disconnects for both the chat server and the game server (stuck in warp before the system finally disconnects) continue with attempts to re-login meeting with the message that the server closed the port (port 26000) for some minutes which suggests backend/cloudfare issues. This happens with both my fibre ISPs each with a different network routing path (one based in Melbourne and the other in Sydney Australia).

On top of that there are regular issues with Cloudfare being unable to contact your “secure” server for shop deals etc - my complaint to customer support went nowhere because the person dealing with it just did not understand that when Cloudfare produces an error message saying it cannot contact your secure server, the problem is NOT with my routers but with either YOU or Cloudfare or both.

So, yes, from the endusers’ point of view I’m not sure I have seen any improvement.

1 Like

If we want to talk fundamental issues: the one is that entire solarsystem is run on a single node. Changing that would require significantly improving scene warp in/out performance, but solving that, it will be possible to spread the solarsystem load across multiple nodes, one per scene. So that each encounter of the fight will be handled by a separate node.

Run EVE with logger and capture a number of instances of this occurence, then create a bug report and attach the logs.

it depends on workload. AMD is still struggling with the same issue that plagued them in the 2000s and made AMD servers all but unviable through the late 2010s in large scale enterprise environments

intercpu compatibility and communication. AMDs only real competitors in the ultra high end server market are Intel and IBM and that wont change any time soon.

the top end Amd Epyc 7702 based system we have been trialing at my work place has been a nice change of pace over the standard Xeon Platinum 8280’s we have been using for the current servers but they lack the raw total compute power and multi-workload ability we have come to depend on.

the difference? our primary server has 32 Intel Xeon Platinum 8280’s for 896 cores at 2.7ghz with a boost upto 4.0ghz. the newer AMD Epyc server has 2 AMD Epyc 7702’s with 256cores at 2ghz base and boosting upto 3.35ghz, both retain the same A6000 gpus from nvidia. the issue? as a direct result of AMD’s longstanding issues around cpu interconnection the AMD server has a little less then 7% of the processing power and raw total performance of the xeon based system stacked ontop of issues around software and hardware support. the Intel systems retain near total widespread software and hardware support to the extent that we have tools that have not been updated since 2007 that function as new in 2023. we have tools from 2023 that work on systems as far back as 2005. ive used a range of software on tape drives and floppys at work in the last year and all of it has worked as intended, because we have been using the same high end server line for as long as we have had servers deployed.

The difference is largely seen in total cost. the AMD test system totalled less then 1/10the cost. but has 7% of the performance and the much newer tech with all the cutting edge bells and whistles has all the issues, problems and flaws that come with brand new cutting edge hardware, much of our current software is coded to work on and with the current intel server solution regardless of the hardware deployed as long as it retains the hardware interconnection solution provided by HPE. things work out of the box because they were made for a solution that has been in active use in enterprise since the 90s in some form or another.

i have spent many many many hours getting even basic software to work. we have actually given up on the blade based database management solution we have been using in some form or another since 2010 because it simply can not be made to run on the amd hardware without reinventing the wheel.

there are things like the server level cross blade rendering solution that will never work because the AMD system will HCF every time the server tries to render down raw 8k video projects despite having the same video processing power and despite adobes attempts they have yet to come to us with an answer, just potential timelines on a solution that may have some similar level of performance. the best they have offered is that we swap to a entirely CPU based rendering pipeline as that would far better suit the amd systems and there unique design choices.

3 Likes

We’re planning a SQL only blog for a later date where we can share more info on the various configs we have going on.

The param sniffing stuff, have not gotten deep into that yet, but that’s basically where the db team is at now - sql 2022 testing and excitement - it’s got a ton of great features that’ll be useful not only to us dba’s but our data engineering team as well.

3 Likes

Something to try:

Create a new profile in the launcher, and try that with your account. (https://support.eveonline.com/hc/en-us/articles/5885278408860-Profile-Management-Tool)

I ran into a similar problem, and this sorted it. (yes, you’ll need to bring over your overview and chat settings at some point too.)

3 Likes