Unused resource management on The Cloud
Something we’ve been discussing recently is IT cost management in the Cloud. One of the things that we’ve addressed on a few occasions, is utilised compute instances. It's common for engineers to be using these machines for testing purposes, so they have valid reasons to be running in the first place but more often than not they don’t need to be running for the length of time that they are. This is probably one of the most common situations for unnecessary costs when running virtual machines on the cloud.
There’s a few ways we can address this issue. For example, Google Cloud offers a feature called Recommender, which provides insights and recommended actions, in AWS this is Compute Optimizer. At Mesoform we previously spent some time working on automation code which uses tags to decide whether machines should be considered for termination. We run this at midnight, every day in our test account. Links to each of these below. Now go destroy some machines ?
https://aws.amazon.com/compute-optimizer
https://cloud.google.com/compute/docs/viewing-and-applying-idle-resources-recommendations#gcloud
https://aws.amazon.com/solutions/implementations/instance-scheduler/
The cost of operations for startups
I'm generally sceptical of studies commissioned by the subject of the study itself but I do agree with the overall conclusion and in fairness to Google, they do point out the fact. Many wouldn't.
We work with companies of all sizes and have views on when and why you would invest in your own dedicated machines. Even if you do, unless you already own it, it should probably be leased in-situ, instead of signed, sealed and delivered.
For a startup, or early operating business, Cloud offers exceptional value for money. The study highlights this, and we agree, but... There are some things that need to be handled to achieve it. If you don't, it could be more expensive.
One example is doing a bit of forecasting and using features like Reserved Instances (AWS) or Committed Use (Google Cloud); or planning for different workload types and where those workloads are short-lived use features like AWS Spot Instances or Google Pre-Emptible Instances.
In case you miss it, the study does point out these specifics with, 'IDC calculated that SMBs spend 26% less over three years with Google Cloud Platform with preemptible VM instances... contributing to the lower costs.'
There's ways to manage cost better with dedicated machines (containers, for example) but The Cloud rules when it comes to the speed and agility with which businesses can build, test and throw away ideas. Something which is essential when market or customer demand changes quickly.
Are your IT services portable?
Beauty of containers is their ability to be ported easily. Done correctly, a container image should be agnostic to it's environment and given the right variables when being deployed, sets the resulting container workload (application) in a multitude of ways: for development purposes, testing, production; in one region on a Cloud provider or in an on-premise datacenter in another region. Even, on your local laptop. Done right, they can drastically simplify service delivery.
Recently we've been working on the reliability and security of a client's Docker Swarm cluster. In particular the pitfalls which can arise out of Docker's use of (Linux firewall) IPTables, and persistent storage and backups.
Since technologies have moved on a lot in the last few years, we’re also taking the time to update our Concierge Paradigm to use an encrypted service mesh and simplify the codebase. Look out for Concierge 2.0!
Why else would you use containers? Check out AWS’s IDC report: https://d1.awsstatic.com/analyst-reports/ModernAppContainersCloud.pdf
How would you migrate your applications to containers on the Cloud? Check out Google's docs: https://d1.awsstatic.com/analyst-reports/ModernAppContainersCloud.pdf
If you would like to discuss any of these topics in more detail, please feel free to get in touch
