Pay to win by using IBM Cloud's Virtual Servers as a development machine that can download docker images at 1Gbps.
May 13, 2021
Pay to win by using IBM Cloud's Virtual Servers as a development machine that can download docker images at 1Gbps.
Score 7 out of 10
Vetted Review
Verified User
Overall Satisfaction with IBM Cloud Virtual Servers
IBM Cloud Virtual Servers are being used for machine learning training and for software development.
Immediate access to the V100 and P100 GPU and 32-core CPUs allows quick testing of machine learning models and methods along with the ease of saving templates and instantiating new servers provides an extraordinarily smooth development process.
JetBrains' Projector has been recently adopted as a means to develop code on a remote machine with a much better experience than remote desktops.
The IBM Cloud Virtual Server interface is simple enough to make the instantiation of a new instance seamless.
Immediate access to the V100 and P100 GPU and 32-core CPUs allows quick testing of machine learning models and methods along with the ease of saving templates and instantiating new servers provides an extraordinarily smooth development process.
JetBrains' Projector has been recently adopted as a means to develop code on a remote machine with a much better experience than remote desktops.
The IBM Cloud Virtual Server interface is simple enough to make the instantiation of a new instance seamless.
- Managing server resources (including configuration)
- Fitting these servers into the production pipeline (because of the templating system and ease of using them to create instances)
- Integration with other IBM Cloud products (such as Object Storage)
- Solid connection to servers (e.g. SSH connections never or rarely drop)
- Easy recovery (seamless backup options)
- Provide an environment for testing CLI scripts
- Maybe offer option to automate backup/restore on shutdown to allow no charging when powered down
- Make the dashboard UI more lightweight so that it's faster and smoother to navigate
- Supports businesses who want compute power on demand (instances are ready within minutes).
- It can give you access to a 32-core CPU at $0.29 for short running or interruptible processes - enabling cost-efficient number crunching.
- Given me a taste of the powers of remote compute power for compilation and development.
I ran a full day's worth of simulations during my lunch break for less than the cost of extra syrup in my latte.
With a few shell scripts and minor adaptations to the Java-Pommerman launch code, I was able to spread tests of the AI across 32 cores at little to no cost.
Preemptible instances can also be used for compiling large C++ codebases using resources that don't need to be constantly available. The net result is that you can get the benefits of having a compilation farm at relatively no cost. This can be used for producing daily builds, while also carrying out other computationally intensive procedures.
With a few shell scripts and minor adaptations to the Java-Pommerman launch code, I was able to spread tests of the AI across 32 cores at little to no cost.
Preemptible instances can also be used for compiling large C++ codebases using resources that don't need to be constantly available. The net result is that you can get the benefits of having a compilation farm at relatively no cost. This can be used for producing daily builds, while also carrying out other computationally intensive procedures.
IBM Cloud Virtual Servers is much more appropriate than IBM Cloud Bare Metal Servers as a remote development machine, as it is only being billed for a third of the day. However, it has fewer options (such as not having access to the T4 GPU).
Compared to IBM Watson Machine Learning, I preferred using IBM Cloud Virtual Servers despite IBM Watson Machine Learning's pipeline being very well refined. The reason is that I have the complete freedom to choose my own stack (e.g. Jupyter Lab) as well as running JetBrain's PyCharm on the virtual server via Projector to run analyses from my remote IDE.
Compared to IBM Watson Machine Learning, I preferred using IBM Cloud Virtual Servers despite IBM Watson Machine Learning's pipeline being very well refined. The reason is that I have the complete freedom to choose my own stack (e.g. Jupyter Lab) as well as running JetBrain's PyCharm on the virtual server via Projector to run analyses from my remote IDE.