Item: IBM Cloud Bare Metal Servers
Rating: 7
Author: jonathan gough

Use Cases and Deployment Scope

We use this as a compute server and a build server for deployments and dev work. It is being used by several different groups at a single client for a lot of backend management and monthly/quarterly intensive compute jobs. It addresses the business issue that Cloud Foundry or virtual boxes do not have enough CPU power as well as a dedicated place for specific tasks.

Pros and Cons

Very good for compute
Reliable because it is always up and not shared
Easy to set up and get running

Very slow to provision
Understanding billing is tough and takes a month to really see what your usage is
Limited compute options.

Return on Investment

Good ROI for compute intensive Batch jobs
Reliable for intensive build jobs
Slow to provision. If you need it you have to wait.

IBM Bare Metal Servers High Performance

We primarily use these for high intensity compute jobs. We have optimized our workloads to run in parallel on bare metal servers in IBM Cloud. We got roughly a 80% speed increase when using bare metal in comparison to using shared services. Our jobs use all cores at roughly 100% for 8-20 hours at a time. We saw and had to deal with a lot of failures when dealing with shared resources.

IBM Bare Metal Servers Provisioning

It is relatively slow, but it works. We have seen faster start times in other cloud providers. Admittedly, we have seen longer start-up times than 4 hours when using IBM cloud bare metal servers. other services like AWS or GCP have more reliable startup times that what we have experienced with IBM

Alternatives Considered

Amazon Elastic Compute Cloud (EC2)

In general we have found that the provisioning and permissions and switching between instances of bare metal and virtual machines were easier in AWS than in IBM. Also the billing in AWS has much greater transparency. The UI interface is also easier to navigate and understand. we also found that our users liked them better.

Other Software Used

AWS Lambda, Amazon SageMaker, IBM Watson Studio (formerly IBM Data Science Experience), Google App Engine

Likelihood to Recommend

Very good for batch processing.
Very good for CPU intensive jobs.
Good for reliable always up and not having to share resources.
On the down side, it is less appropriate for scenarios where you need on demand compute.
If budget is an issue, it is quite expensive.
Not great if you need resources quickly.

IBM Bare Metal Servers work, but they make you work too... but that is just the way it is.

Overall Satisfaction with IBM Cloud Bare Metal Servers