Started monerod 0.10.1 on an AWS EC2 t2.micro instance, why does it sync so slowly?

Question

I started monerod on an AWS EC2 instance of type t2.micro. This means it has 1GB of RAM, which should be enough from what I've read in various places. It is syncing (from scratch) since two days already, and arrived at roughly 60 percent of the block height only so far.

CPU usage is below 1 percent, but memory above 700 MB, so I guess it is swapping a lot.

Do I have to set some specific parameters for runnning better on machines with little memory?

EDIT: I should add that it's an AWS instance from the free tier, if that makes any difference.

Answer 1

I now found the reason, and it perfectly makes sense (took me a while to realize though, being a first time AWS user).

My instance is running on an SSD (more specifically, the volume type is called gp2), but of course I have to share the hardware with a lot of other users, so Amazon limits the IOPS. Currently you get 3 IOPS per GB as baseline performance (but at least 100 IOPS for small volumes), and you're allowed to burst up to 3000 IOPS as long as you have what they call credits. Initially you get 5.4 million credits, which allows you to burst at 3000 IOPS for 30 minutes. After that you fall back to baseline performance, and during times when you don't use the volume your credits are filled up again at the baseline rate. More details here: Amazon EBS volume types

This matches the behaviour that I observed: Synching started out quite fast, but became incredibly slow later. Based on the above, it must have slowed down after around 30 minutes, which wasn't enough to sync the whole blockchain from scratch. Hence, when running on such an instance, it might make sense to download the blockchain from the monero website.

Answer 2

Your instance is probably stored on a HDD, rather than a SSD, and those take significantly longer to sync up the first time.