To solve the issue of too many in one system for the server

Sure jumping straight into the battle was silly, but not jumping at all would have been just as silly.

Yeah totally. Not jumping at all wouldnā€™t have achieved anything.

There were a series of mistakes, like all major losses involve. Not having move ops finished. Not staging in their Fortizar progressively in the days prior. Jumping straight into fighter bombers.

Then, jumping into a system that was already at 5000 people. It seems a number of the older heads successfully used the logout/login trick to get themself back out before they even got in, which would have also added to server load.

Lotā€™s of things could have been different and Iā€™m sure if this could be done over, PAPI would approach it differently.

1 Like

In the cases of these large battles, CCP could limit the use of cans, fighters, drones etc.

The issue is destiny canā€™t be scaled. Also the locks from DB have an exponential delay (synchronization CPU usage is in square of access rate) but maybe thatā€™s an addressed issue.

HW changes can make the server handle more activity, but since the activity is more than linear with the amount of players (exponential but for high number?) then the only solution is to make HW upgrades and program changes in order for Eve to be handled on a SINGLE server.

When the Eve physics engine can handle the whole number of connected users( 30k ?), then the game can allow people to make as big of a fight as they want. However this requires a lot of changes, including possibly a second level of time dilation to reduce the ticks of the server to 1/10th, and/or the disable of AOE effects(in a non deterministic way, so as to not allow this to be abused), and/or the simplification of some mechanisms to reduce the server load.

hypervisor assumes virtual machine, which is bad for CPU intensive calls and especially on hardwares specifically made for high performances.

You have to prove there is an issue. The only issue I can see is that the current state allows people to prevent a fight by amassing enough defenders, therefore making structures perfectly safe if the defender can prepare before the attacker comes in. Itā€™s something thatā€™s mandatory as long as the servers canā€™t handle the full number of connected players in one single grid.

ā€œshouldā€. nope. There is no reason to arbitrary remove the strength of a side at an arbitrary limit.

Well it specifies virtualisation (not just VMs) and isnā€™t so much an assumption. Once a hypervisor is in use, virtualization is being used. With current generation hypervisors, CPU intensive operations are not negatively impacted in a significant way. I will happily post public papers tomorrow if you want.

And there is no need to assume CCP use virtualisation, the devblog for TQ Tech III clearly shows ESXi hosts:

Yes, although itā€™s not handled on a single server now. Itā€™s single sharded (ie. the database is a single shard), but it runs on multiple servers.

But yes totally. Continued hardware developments and changes in the code base is the long term solution. Itā€™ll never truly be a solution as such, because there will always be a physical limit to how big a fight can get, but the ceiling will continue to increase while CCP continues to maintain and develop the game.

VM is a compute requirement.

Network and storage virtualisation donā€™t rely on VMs. Containers running directly on a hypervisor donā€™t require a VM.

What you are writing isnā€™t correct.

Sure.If you canā€™t read.

This is why getting into any discussion with you is stupid. You donā€™t know what you are talking about but carry on anyway.

Iā€™m an idiot for even writing this follow up post, but you won. You got me to waste some more time.

A hypervisor (or virtual machine monitor, VMM, virtualizer) is computer software, firmware or hardware that creates and runs virtual machines.

and we are not talking about network or storage virtualization, but about ā€œhosting the physics engineā€ which is where the issues arise.

Also network and storage virtualization is not virtualization in the sense of hypervisor.

Nice personal attack that you made here, showing that you know you are wrong but canā€™t admit it.

I know. Nothing more idiot than making personal attacks when you are wrong instead of trying to present your position in a civil way.
Wait ā€¦ maybe moving the topic on something else, like ā€œvirtual networkā€ instead of virtualization is actually more idiot ? duno, but you definitely made more efforts to look like an idiot than you needed.

Yup. Sad.

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.