Container-Native Means Big Savings

May 11, 2016 - by Bill Fine

Next generation container-native (G4) and KVM-based (K4) instance package families are now available on Joyent’s Triton-powered public cloud. These new instance packages replace prior generation T4 and G3 instance package families and provide a wider range of options at lower costs. Dig into the new instance packages and you’ll find they align with existing AWS instance types and pricing. Dig deeper and you’ll realize that the total cost of ownership on the container-native Triton Cloud is significantly cheaper.

A container-native cost advantage

Workloads are more efficient on Triton Cloud. This is because Triton allows you to run containers natively, without having to pre-provision (and pay for) virtual machine hosts. The result is less waste and more cost savings for you.

Consider our recent blueprint to run WordPress in containers. A minimum running implementation requires six g4-highcpu-128M instances and costs just over $13 per month. That minimal site may be perfect for a small blog or staging a larger one. Should you need to scale it, you can resize the containers without restarting them or scale horizontally with docker-compose scale…(or another scheduler of your choice). Triton automatically places each container on a compute node with the resources you request. You pay only for the containers you use, and you don’t need to worry about VM bin-packing or wasted resources on VMs running under capacity.

Here’s a scaled out version of that Wordpress implementation on Triton:

Instances Containers $/month
g4-highcpu-1G Nginx $18.98
g4-highcpu-1G Nginx $18.98
g4-general-4G WordPress $48.91
g4-general-4G WordPress $48.91
g4-general-4G WordPress $48.91
g4-general-4G Memcached $48.91
g4-general-4G Memcached $48.91
g4-general-4G Memcached $48.91
g4-highram-16G Percona $151.48
g4-highram-16G Percona $151.48
g4-highram-16G Percona $151.48
g4-bigdisk-32G NFS $503.70
g4-highcpu-128M Consul $2.19
g4-highcpu-128M Consul $2.19
g4-highcpu-128M Consul $2.19
Total $1,296.13

The management and cost efficiency of running each container in its own instance on Triton is most clear when we compare it to how we might try to bin-pack the same containers into AWS VMs:

Instances Containers $/month
m4.xlarge Nginx $174.47
WordPress
Memcached
Consul
m4.xlarge Nginx $174.47
WordPress
Memcached
Consul
m4.xlarge WordPress $174.47
Memcached
Consul
r3.2xlarge Percona $485.45
Consul
r3.2xlarge Percona $485.45
Consul
r3.2xlarge Percona $485.45
Consul
d2.xlarge NFS $503.70
Consul
Total $2,483.46

In this example, the cost of running on Triton is about half the cost of running on AWS. With enough experimentation and determination you might be able to narrow this cost gap by more efficiently bin-packing your containers into VMs on AWS, but on Triton those efficiencies are built in and the cost and complexity of VM host clustering is removed. Each container just runs (on bare metal) with the resources you specify.

Triton’s container management efficiency is even clearer when we compare the application lifecycle on AWS and Triton:

Task AWS Triton
Deploy a minimal site $37.99/month; All containers packed in a single t2.medium VM instance $13/month; Each container on its own g4-highcpu-128M instance for maximum performance, reliability, and flexibility
Scale up for live traffic and for redundancy $2,483.46/month; Redeploy all containers into new VMs, possibly risking downtime or data loss $1,296.13/month; Resize instances without restarting for convenience and deploy additional instances
Scale for a big event Provision new VMs and redistribute containers Resizing or adding instances is easy
Scale back down to “normal” Re-allocate and redistribute containers to a smaller fleet of VMs; it may be difficult to reduce costs if the containers can’t be moved Because each container is its own compute instance, costs go down as you scale down the container count for each service

A price-performance cost advantage

Workloads also run faster on Triton Cloud. This is because Triton instance packages leverage container-based OS virtualization, advanced caching technology, and faster data retrieval from storage systems.

  • Elasticsearch clusters on Triton complete query requests 50% to 70% faster than comparable Elasticsearch clusters running identical data sets on AWS.
  • Sharded MongoDB clusters on Triton complete tasks 100% to 150% faster than comparable MongoDB clusters running on AWS.
  • Standard primary/replica Postgres configurations on Triton complete tasks up to 200% faster than on comparable Postgres multi-node virtualized database configurations running on AWS.

Better performance on Triton Cloud means you save money, because you need less infrastructure to power your application. Or, for the same cost, you can turbocharge your applications. It’s up to you.

Our secret sauce

How is it that Joyent can compete with (and beat) AWS on price? The answer is simple. Our cloud runs on containers, while AWS’s cloud runs on virtual machines. Ten years ago we bet big on containers, believing that we could run and operate containers much more efficiently than traditional virtual machines. We bet right, and now we are able to pass those efficiencies and savings on to our customers.

Bonus: The hybrid cloud cost advantage

Sometimes it’s cheaper to run workloads in your own data center. As an example, a born in the cloud company recently shared that, by moving to a hybrid cloud model, they were able to double their capacity while reducing their costs by two thirds. Triton makes it easy for you to deploy and operate a hybrid cloud to take advantage of these kinds of potential cost savings. The software we use to operate the Triton Cloud service is 100% open source and available for you to use to operate your own private data centers.

Whether you choose to deploy your application on the Triton Cloud service, or on your own Triton-powered data center, with Triton and our new instance packages it is easy to begin reaping the benefits of a true, container-native architecture.