Human intelligence vs. machine intelligence
I sat with Joe (not the same Joe we mentioned last week and another disguised name) to show him the “just-in-time forecast” feature we released in Dec, 2016. It is based on X-Fit I explained in our day 7 tour. Joe got excited. “Instead of machine intelligence, I have done the same work also using intelligence,” he chuckled, ” human intelligence.”
Each week, Joe monitored the usage and tried to estimate how long the cluster can sustain without running out of capacity. If he sensed he needs to expand his cluster, he would look back the history of the usage. Sometimes he even had to study the screenshots of the charts he captured a year ago to get the sense of annual or monthly seasonality. He then had to use a spreadsheet to plug in the numbers, including the actual workload usage, the assumptions of future usage, the current hardware configuration, and the possible future hardware configuration. “It is at least a three-day process, and I hate to spend that much time on this task.” Then he added, “This is just the simplest form. Many times, I need to go back and forth with application teams and management if there are new projects involved.”
Joe needs the help from the machine intelligence.
Joe hit several challenges in his process. First, he had to guess a lot. He had to guess how long the infrastructure can support the workload. He had to guess what the workload will behave. He had to guess what hardware he can use and add. Machine learning can take out those guessworks. We have designed X-Fit to relief Joe from those pains.
Second, he may have missed other options that can extend the infrastructure life without purchasing new nodes. He could remove dead VMs to recoup the disk spaces. He could also identify the over-provisioned VMs and increase the available CPU and memory capacity for additional provisioning. If Joe needs to add workloads from a new project. He could check whether other clusters can host the extra capacity required.
Last, but not least, Joe’s time is expensive. He could have spent the time on addressing higher value tasks.
Just-in-time forecast (JITF) is developed to help Joe get the time back and save his company’s money. JITF is designed with three differentiations to help Joe.
X-Fit is its brain. JITF uses X-Fit to understand the current usages patterns and predict the future demands. This puts IT infrastructure deployment aligned with the business growth.
The recommendation engine in JITF lays out the timeline of your infrastructure deployment schedule to match the workload demands. As a result, it will prevent Joe’s company from spending unnecessary capitals for the capacity they need not use.
JITF has the built-in simplicity, including one-click recommendations and workload-friendly scenario definitions. Joe can now complete planning tasks within 3 minutes, compared to the three days he used to spend.
Questions that JITF help answer
Just-in-time forecast, JITF, can help Joe answer many questions. Here are some of them.
How much expansion is needed if the cluster expects a capacity shortage
Whether a cluster has enough runway after adding new workloads
Whether you will need to expand the cluster if the current workload changes its behavior (e.g., high workload volume due to a demand surge, M&A, marketing promotions, etc.)
When and how much capacity you will need for a new cluster to support its workloads
Which cluster is the optimal place to support your new workloads
The impact on capacity if you move nodes into your production cluster from your staging environment
In the video below, I will show you two examples on how Joe can use JITF.
We are bringing new furniture in the rooms we have visited. and we are building new rooms. I will invite you back when we have them ready for you. I will list new tours in our day 1 page. I also will announce that on the twitter.
In the meantime, I will start a new series – “Nutanix Prism – Now you know” – to showcase tips and tricks of how you can use Prism. Stay tuned.
Disclaimer: This blog is personal and reflects the opinions of the author, not necessarily those of Nutanix. It has not been reviewed nor approved by Nutanix.