Item: IBM Cloud Virtual Servers for VPC
Rating: 6
Author: Amit Kumar Verma

Use Cases and Deployment Scope

We are in Deep Tech space ( or DeepLearning ). We use AI to create voice that will almost sound like human. So mostly, We use IBM's Virtual GPU servers for training and evaluating our deep-learning models.

Pros and Cons

On-demand pay feature, [because] GPU [is] costly. When they are shut down, they almost charge 0$
Bare-metal OS, No pre-install vendor software
Seeing the monthly pricing before creating VS

Having a market place for pre-defined images would be great (i.e DeepLearning with pytorch image)
It would be great if GPU drivers are installed during VS creating process (i.e i have find NVDIA driver and install it by myself, [which] takes a-lot of time)
While creating an VS. Extra HD space should be mounted automatically. Instead, [I have to] log in and mount it by myself.

Return on Investment

Reduce costs
Good Performance
Great output from Neural Services

Alternatives Considered

IBM Watson Natural Language Understanding

IBM Cloud Portfolio

IBM Watson Natural Language Understanding

Other Software Used

Amazon S3 (Simple Storage Service), Firebase, MongoDB

Likelihood to Recommend

I would really love if you can provide pre-installed images for OS. When i was setting-up my OS for Pytorch environment, I had to setup following things by myself: 1) Mounting extra space, 2) Finding and Installing NVDIA driver, [and] 3) Install[ing] pytorch and supporting software. This whole process took around 3-4 hours which [I belive] is a complete waste of time and money for me. I used to do my development on AWS. Perhaps [you] should look at their marketplace.

Having a marketplace images for Virtual Servers will help them getting started faster

Overall Satisfaction with IBM Cloud Virtual Servers