2018 started with comparatively high drama in tech circles, as the leak, reveal and operating system patches for the Meltdown security flaw in CPUs unfolded. To keep it simple, it’s an issue that’s mostly affecting Intel and has the potential to seriously affect their CPU performance depending on the workload.
While the fixes aren’t really impacting everyday users with everyday computing needs, from gaming to content creation or simple web browsing, they are much more likely to affect the cloud and online services.
Following recent login issues and instability in Fortnite, Epic Games posted an explanatory message on their forums showing the dramatic increase in CPU load that was seen once one of their three cloud service hosts was patched to address Meltdown.
Yup, that huge jump in the green line that literally doubled the CPU load? That’s the immediate cost to Fortnite for fixing Meltdown. Epic say that there may be more instability in the weeks to come as further fixes are rolled out, but that they hope to smooth the process as much as possible.
Thankfully, the fixes shouldn’t be quite so keenly felt in the long term, but we’ll have to wait and see if there are teething problems across other games as fixes are applied.
Update: Here’s another example of the impact that Meltdown can have, with CCP tweeting out the before and after load stats from their servers for EVE Online. As you can see this is again a sizeable increase in CPU load.
the impact of the #intelbug on ESI’s production cluster: (rolling upgrade from 9:30-9:55) #devfleet#tweetfleetpic.twitter.com/3F1J9SZ6zi
— CCP SnowedIn (@CCP_SnowedIn) 4 January 2018
Meanwhile, the increase that can be found on Amazon Web Services from patches to mitigate the related Spectre security flaw. While end users might not see the damage this has done to performance, behind the scenes, this is the new normal.
The #Meltdown patch (presumably) being applied to the underlying AWS EC2 hypervisor on some of our production Kafka brokers [d2.xlarge]. Ranges from 5-20% relative CPU increase. Ooof. pic.twitter.com/fXM0OzfdKx
— Ian Chan (@chanian) 6 January 2018
Also, from my understanding, this isn’t temporary. This is now the new baseline. Double Ooof.
— Ian Chan (@chanian) 6 January 2018
Source: Epic Games
Tony Cawley
This makes a lot of sense. All of a sudden both me and my son were getting problems with the game, joining 8 minute log in queues and being kicked mid game.
I’d put it down to Christmas and the increased amount of people owning consoles and the game, but this makes more sense.
Stefan L
It’s a mixture. Games have always had difficulties during high load periods, whether it’s a major release or something seasonal, but now each person trying to play costs Epic double the CPU resources. Additionally, there’s a hit to storage performance from these fixes, which exacerbates things when lots of people are trying to login and the servers are having to reference the database, log player stat changes etc. etc.
I’ll update this later today with a few other examples that I’ve now seen this morning – EVE Online’s servers have jumped up by at least 30% as well, and the cost is not insignificant to AWS and other cloud services – but this is the new normal for devs to have to deal with.
Servers will need to be beefed up significantly to handle major game launches, DDoS attacks will be much easier to pull off, etc. etc.