over 4 years ago - CCP_Dopamine - Direct link

Hello, friends,

We will have a couple of extended downtimes this week to perform a range of database testing. The downtime will be extended on:

  • Wednesday, 24 June
  • Thursday, 25 June

Each downtime may take up to 30 minutes and we expect Tranquility to be back accepting connections by 11:30 UTC at the latest. However, we will let players in earlier if work is finished earlier. Updated will be posted in this thread and on EVE Status Twitter as they become available.

Thank you for your understanding and sorry for any inconvenience this may cause.

over 4 years ago - CCP_Dopamine - Direct link

UPDATES:

over 4 years ago - CCP_Dopamine - Direct link

UPDATES:

over 4 years ago - CCP_DeNormalized - Direct link

We’re excited to test this in production tomorrow!! Anyone care to guess specs?? :slight_smile:

poc_db_server_cpu_ram4032×3024 1.47 MB

over 4 years ago - CCP_DeNormalized - Direct link

You can hold to the faith… it has significantly more ram and cpu cores than that :slight_smile:

over 4 years ago - CCP_DeNormalized - Direct link

It is a dell chassis, R7xxxx :slight_smile:

over 4 years ago - CCP_DeNormalized - Direct link

(post withdrawn by author, will be automatically deleted in 24 hours unless flagged)

over 4 years ago - CCP_DeNormalized - Direct link

some of those numbers are correct!

over 4 years ago - CCP_DeNormalized - Direct link

no, this is testing new db hardware for tq

over 4 years ago - CCP_DeNormalized - Direct link

close enough!! This is a R7525 with a single EPYC 7742

over 4 years ago - CCP_DeNormalized - Direct link

AMD!! We’ll see how 128 cores compares to 32 tomorrow!

over 4 years ago - CCP_DeNormalized - Direct link

The core of the simulation is still on Windows (DB and Sol Nodes). This particular server is Win 2019 - and I hear you with lots left on the table, but not just with linux. SQL Server in general really benefits from tuning to the workload being thrown at it.

over 4 years ago - CCP_DeNormalized - Direct link

we’ll do a dev blog for sure

over 4 years ago - CCP_DeNormalized - Direct link

in the near future we’ll also test more traditional setups - dual socket, 8 core CPU’s from intel/amd - this will match our current 32 cores, with as you say, max cpu performance.

We’ll do a devblog later and have more info on what we’re measuring with more cores/less numa nodes, etc…

And licensing is a giant thing :frowning:

over 4 years ago - CCP_DeNormalized - Direct link

No, I don’t think so :frowning: The DB typically isn’t a bottleneck in big fights. But this will likely help with slowness for the first while post-startup and also on busy weekends the current DB struggles.

Market is one of the big things that should be snappier with this type of change

over 4 years ago - CCP_DeNormalized - Direct link

this is 16x128gb modules for 2 tb :slight_smile: it would have been 4 tb of memory but the vendor said all 256gb chips were sold out worldwide.

but correct on the dual sockets / unused memory slots

over 4 years ago - CCP_DeNormalized - Direct link

it would of been REALLY interesting to have the other socket populated the same - 128/256 ht cores and 4 tb of memory, hmmm!

over 4 years ago - CCP_DeNormalized - Direct link

EYPC 7742, yup

over 4 years ago - CCP_DeNormalized - Direct link

I’m not sure :frowning: But things like that are not typically DB heavy

over 4 years ago - CCP_Dopamine - Direct link

Good day, everyone! The extended downtime has started. We will keep you posted here and on EVE Status Twitter on the progress.

over 4 years ago - CCP_Dopamine - Direct link

Tests went well. Thank you everyone for your patience. Tranquility is now back online accepting connections!

Please remember that tomorrow we will have another extended downtime with estimated duration of 30 minutes.

over 4 years ago - CCP_Dopamine - Direct link

If you are having payment issues or did not receive PLEX you purchased, please contact customer support for assistance.

over 4 years ago - CCP_DeNormalized - Direct link

We may have had a CPU spike just after failing over… but its all good now :slight_smile:

poc_db_server_cpu_ram_max_cpu1408×813 111 KB

over 4 years ago - CCP_DeNormalized - Direct link

It is a MS SQL server, and its not fun in any way to hint or think about the license cost :slight_smile:

over 4 years ago - CCP_DeNormalized - Direct link

Close! 64 cores is right (we have SMT enabled, so 128 HT cores) but it has 2 TB of memory

over 4 years ago - CCP_DeNormalized - Direct link

This is actually just a proof of concept… it’s a loaner box to see how we like it and how the AMD architecture will work for EVE (I love our vendors!). But I can’t imagine we would ever fill the other socket, at least not with another CPU like this.

over 4 years ago - CCP_DeNormalized - Direct link

we’ve dropped from 30% CPU to well under 10% - at this time of day things are typically calm - CPU is rarely balanced across cores due to the nature of how SQL works - it could using a single core for a call or a parallel plan that would hit multiple cores.

I can say that we’re seeing much more balanced NUMA nodes (groups of cores) which is one of the major things we were after here compared to our old setup

image1653×890 42.6 KB

over 4 years ago - CCP_DeNormalized - Direct link

We’ll do a devblog at a minimum and have had other people mention a stream… will look into it further!

CCP RAM and I did a presentation a few years ago at fanfest - it’s a mix of how we came to CCP and daily Operational tasks/tech talk at the end: https://www.youtube.com/watch?v=w8rGZCj6rgQ