Notices by Moritz Heiber (moritzheiber@mastodon.social), page 8

@Thegcat @netzvagabund Isso ist super, kann ich sehr empfehlen.

@nolan yes! So if you’ve switched before, from t2 to t3, you should no longer be exhibiting these issues (hopefully)

@nolan try the unlimited mode? It’ll be an extra $1.2 per day (max!) when the t3.medium is running in that mode..

Or, try a C or M instance. They aren’t all that much expensive but provide an equal share between performance and reasonable costs

There are two sparrows in the secure area of the TXL terminal E, and it’s the cutest pair of passengers I’ve ever had the pleasure of waiting for my flight with..

I have no idea how they got in, and apparently so doesn’t the security guard who is (very unsuccessfully) trying to “escort them out” 😊

@JPEG @haptap 🤗

@nolan @markus here’s the reference documentation: https://aws.amazon.com/ec2/instance-types/t3/

@nolan @markus You can keep track of the accrued and remaining CPU credits in CloudWatch, and the graph should clearly indicate whether or not this has been a problem before

@nolan @markus Also, your instances failing on a single T2 instance just _might_ be because you ran out of CPU credits, because if the instance is still under load in “boost mode”, and is suddenly running out of credits, it’ll probably not respond in a more timely fashion afterwards but instead chug along painfully slow

@nolan @markus All instances prior to T3 started out in “Standard” mode, ie once the credits have run out your performance will fall back to a non-boost level.

T3s start out in “Unlimited” mode, where extra charges may apply (very negligible ones I might say though).

It’s a essentially a guaranteed performance vs guaranteed credit baseline kinda thing

@nolan @markus The way T instances work is that they accrue CPU “credits” over a 24h time period, which they can then use afterwards to “boost” performance should it be required. This is possible up until the credits run out usually.

Unless you’re in “unlimited” mode, where, once the credits are used up, you’re paying an extra 5 cents and hour until either you no longer need the boost or the credits have a accrued again, whichever comes first

@nolan newer processor, little more RAM, better hypervisor integration of the underlying architecture