Monday, July 24, 2017

Open learning and the race to the Cloud

Last year I got a chance to work on a customer solution deployment based on a Cloud heavy Apache Stratos + kubernetes + mesos setup. It was a brash hands on introduction to the circuitry that power many a cloud solutions. At the end of that engagement, I made a mental note to pay a second visit to some of the technologies I got exposed too when time was at hand.

Not all of us Sri Lankans' have the mental abilities or the tough stomach needed to digest and perform the academic aerobics needed to secure a good free education in our country. Even fewer have the financial capabilities to fund a good education abroad. So naturally, many of us who wish to continue learning find ourselves doing so utilizing platforms such as Coursera, Udacity and Khan Academy. With these platforms you can now access the latest content curated and delivered by experts hailing from prestigious universities halfway around the world. All of these changes tip the balance in favor of those who are truly interested in a subject.

Apart from the platforms mentioned above, as an engineer who’s interested in keeping up with the technologies in my space I’ve found free trials offered by many vendors to be ridiculously useful.

From a provider’s perspective, it’s a great way to market your offering, especially if the long term success of the offering depends on the adoption and loyalty of specialized consumers. This happens to be the case with the Cloud. In the long run, the ultimate winner or winners will be decided by the level of developer traction secured. It is because of this very reason, I believe you can now get a 300$ worth of computation/storage time on the Google Cloud Platform for free, some services on AWS for free for extended periods of time and why you can try out products like Apigee Edge and RedHat OpenShift without having to spend a cent.

Back to the topic at hand, this month I thought I’d get back to where I left off last year. To get my hands a little dirty and get a test of the latest from the Cloud space. I cashed in my GCP green and took RedHat OpenShift for a spin. This post captures a few thoughts from the experience.

Google Cloud Platform

Coming from a middleware background, it’s easy to categorize GCP as an an end-to-end middleware platform in the cloud(forgive my ignorance, I’m sure it's a lot more than that). As such it offers all the components one could hope for when modernizing IT systems.


Infrastructure

Users may ascertain as much or as little control over their infrastructure as they wish. Those who are seeking IaaS level control over their systems may build up staring at the Compute Engine. If the user wants to delegate work from this layer to the platform they may build up from the Container Engine or App Engine instead.

Storage

A bucket is an atomic storage unit in the GCP context, based on accessibility and performance requirements needed for the buckets, the storage options can be grouped into three main categories, Standard; which provides the best SLA for storage that needs to be frequently and globally accessed. RDA; for those storage needs that are less taxing. Nearline/Coldline; for storage that is rarely accessed such as those needed for backup or disaster recovery. 

GCP provides a foray of storage solutions meant for application consumption and insight generation. Everything from RDMS services such as Cloud SQL to no SQL services such as Cloud Datastore. Storage for analytics is provided through BigTable which can provide high write capabilities needed for scenarios which push large number of data to be processed later on.


Utilities

Utility capabilities such as inflight data transformations, reliable messaging and identity management capabilities are provided through solutions such as Cloud dataflow, pub/sub and IAM. 

For anyone who wishes to experience the platform and its breadth, this Coursera course would be a great place to start[1].


What I liked liked about the platform and the free trial,
  • There seems to be a one to one mapping between GCP services/products, and capabilities/offerings you would expect from a middleware vendor. This makes mapping solutions to the platform easier.
  • The platform wide logging and tracing capabilities provided that just works. 
  • The utilities provided to make development work easier such as API for all the services and client side libraries to make tie in work easier, CLI(figure 1) which wraps the service API to provide a convenient means of access to developers.
  • The flexibility provided in creating the consumption architecture; for example for someone who is creating a backend for a mobile application, they may directly consume some of the GCP services(such as storage) or utilize GCP API creation capabilities to consume the services similar to the way GCP API are consumed but with better control[2][3].
  • Responsive support system, I tested the waters with a container engine query and sure enough I got a prompt appropriate response. 
  • The ability to pay for only what you use(which is a given and a key selling point for the cloud). 
  • Looks like the USD $300 given by google can carry evaluation work far.
 


Figure 1
RedHat OpenShift

After building a billion dollar global organization which has withstood over two decades of industry battering, I don’t think there will be many naysayers when I say RedHat has mastered what it takes to make an Open Source business model work. Their key proposition is value addition(be it functional or operational) to generic Open Source, a simple but effective mantra that works!

OpenShift is aimed at the on-premise managed cloud space, which is a fancy way of saying putting the some of the best of tech that make cloud services like GCP and AWS work behind a corporate network, for added control. Therefore, to gain the most of the offering it should be deployed inside an organization's data center.


First impressions and what I liked about the trial,
  • The trial allows a 14 day all pass access to the PaaS solution deployed on either GCP or Azure. If you go with GCP you get about 6 hours of computation access at a time, which is plenty for anyone who wishes to evaluate the functional capabilities.
  • The product UI(figure 2) is easy to understand, difficult to get lost in. 
  • The CLI client effectively envelopes the kubernetes and OpenShift API, making the deployment and routing setup easier to do than otherwise. 
  • Access to docker hub from within the trial setup. 
  • The concise lab documentation that takes you through the key functionality. 
  • The product seems to support the leading conventions and technologies in the space with their own value added spin-offs or vanilla components when they are good as is.



Figure 2






No comments:

Post a Comment