What’s really great about these low prices is that you can get on and build a small but scalable cloud architecture for yourself and try your startup ideas in a way it won’t vacuum your wallet empty and you won’t be scratching your head when your startup starts booming and you need to scale.
One other thing I just LOVE about their pricing page is that they list the prices monthly, not just per hour. There are no hidden costs (compare that to AWS’s traffic pricing, EBS volume prices, fixed IP prices…),
Let’s see how EC2 instances perform compared to DigitalOcean droplets (DO droplet == EC2 instance).
The servers I tested:
Smallest DigitalOcean droplet: 512 Mb RAM, 1 CPU - $5 / month
More potent DigitalOcean droplet: 2GB RAM, 2 CPU - $20 / month
EC2 m1.medium: 3.75 GB RAM, 1 VCPU - $87.84 / month
EC2 m1.large: 7.5 GB RAM, 2 VCPU - $175.68 / month
I left out the t1.micro EC2 instance on purporse as it’s a very low performant, yet it costs $14.64 / month
Ouch. A serious burn for AWS.
Yikes. Would anyone want to host anything disk-intensive on EC2 anymore?
You can clearly see that we’re getting very powerful computing power for much less money at DigitalOcean.
I’m really missing ELB. Elastic Load Balancing is a great way to do fire-and-forget load balancing without having to worry about scaling. Its pricing is also really great and you can easily add and remove nodes using their API.
You don’t have anything like this at DigitalOcean (A great opportunity, DigitalOcean guys! Please create such a service, I’d gladly pay for it).
You don’t have DynamoDb or ElastiCache either. No RDS. No nothing. You’re back to your oldschool ops mojo and you have to build your infrastructure yourself. This can be a good thing actually, though building a repicated HA DB cluster is one of the challenges few are willing to take on.
It’s not so hard to create a HA cloud though using Chef or anything similar (I’m using SaltStack btw). Here’s what I did:
DigitalOcean is clearly a great and cheap way to start experimenting with the cloud, it’s infrastructures and create a low-cost, scalable solution for your Next Great Idea™ but it has many ways to improve, e.g. removing the burden of the oh-so-typical infrastructural parts - load balancing or replicated databases with in-house solutions.
Due to a request by dzsol I benchmarked a c3.large instance too. This instance’s characteristics: $40/mo over 3yrs if you pay the total amount upfront, 3.75GB RAM, 7CU, 32GB SSD.
The results are (without a fancy graph):
We can see the CPU is powerful (no wonder the c-line is called compute-intensive) but the I/O still sucks big time. I had a hard time believing it really had an SSD but I double-checked and it’s not a rotational disk according to sysfs.
The benchmarks I used can be found here