How to Cut Your Public Cloud Cost by 50 Percent

Cloud has become the new platform for enterprises to build, test and run their applications. Public clouds like AWS have been a preferred choice for a lot of companies due to a wide variety of instance size options, storage options, and services. Any team within mid to large enterprises that is looking to quickly build an application needs a self-service environment for application development. The options are either to go to public cloud or wait on their IT to provide an on-premises cloud using VMware vCloud, Azure Pack, OpenStack or some other cloud stack. Since the latter option can take weeks to months, most developers choose to go to public cloud and start consuming resources with a quick credit-card swipe.

Public Cloud Cost Grows Rapidly

Once they start using a public cloud, people realize that the overall cost grows rapidly. There is a separate cost associated with each of the resources. For example, a typical workload just leveraging IaaS features will have:

1. VM instances: EC2 pricing
2. EBS volumes: storage capacity pricing
3. Storage IOPS: IO based pricing
4. Outbound data transfer pricing

Now, there are some obvious solutions that one can use to lower the cost to some extent:

1. Reserved instances with 1-year reserved pricing
2. Long term reserved instances with 3-year reserved pricing and upfront cost
3. Spot instances

These solutions can help lower the cost to some extent, but it still doesn’t come anywhere close to on-premise infrastructure prices. Consider a dev team with 50 developers, using 250 VMs overall. The requirements for such a team can be handled using 10-15 on-premises physical servers. Here each physical server can run up to 20 VMs of medium size with ease. The cost of 10 servers, each with 32 cores, 256GB RAM, 6 TB SSD storage, and 20 Gbps network bandwidth is around $10,000 x 10 = $100,000. Let us look at the cost of public cloud for similar resources.

Public Cloud Cost (3 years)

Getting the CPU, memory and storage capacity equal to 10 physical servers, on AWS for 3 years will cost within the range of $1M to $1.5M. The lower end of the range can be obtained with 3-year reserved instances and the upper end with on-demand pricing. Table I below captures the exact amount for various configurations:

Given $100K vs. $1M as the gap in 3-year cost, why not have a hybrid architecture where some of the consistent workload is run in on-premises cloud and more elastic workloads are run in public cloud? The main reason is that the cost of on-premises cloud is significantly higher than the cost of infrastructure.  Most customers talk about their lack of interest in dealing with infrastructure and operations due to the following factors:

1. Cloud software costs
2. Cost of people needed to build the cloud
3. Time and effort needed to hire and manage the right experts

Once you add up all the costs with existing on-premise cloud solutions, it can get close to the public cloud costs and will require a big upfront investment.

So how can one get an environment that provides the benefit of both on-premise and public cloud economics?

The ZeroStack Solution & Its Cost

At ZeroStack, we have built a solution where a customer can get an on-premises cloud with fully integrated infrastructure and software stack. It has three main features:

1. Built-in cloud software with cost comparable to the hardware
2. No experts are needed to build or operate the cloud
3. Operations are handled by self-healing software and cloud-based monitoring done by ZeroStack
4. The solution provides a seamless migration to/from AWS with a dial to optimize cost and efficiency across the two environments

The total cost of a 10-server ZeroStack cloud for a development team comes to less than $500K for 3 years, which is at least 50% less than the public cloud costs. This includes the cost of hardware, cloud software, power, cooling, and operations.

How Do You Save 50%?

First, we provide an automated migration feature built into the platform to move workloads. We provide 1:1 mapping between various AWS concepts to similar ZeroStack concepts. In cases, where customers will not be able to move all of their workloads to on-premises infrastructure (depending on their use of cloud services), we can move one application at a time. For each of them, the savings will be 50% or more. In addition, customers will get data locality, lower latency to their development teams, and 4-6x higher I/O performance compared to public cloud.

If you want to know more and figure out how you can get flexibility and an automated dial to control your private and public cloud cost & efficiency, contact us at


Leave a Reply