Most of the blog posts on the Apprenda blog come from our more propeller-headed team members or our CEO (ok, ok, for anyone that knows Sinclair he’s the lead propeller-head). As the head of global sales for Apprenda, I really get excited by quantitative stuff like ROI. The goal of my blog post is to provide enterprise technology buyers and users with real-world examples of how Apprenda is directly helping our forward-thinking customers solve complex challenges, while also helping them to be more productive and reduce costs.
What will you learn from this post?
- How heavily virtualized environments still have significant inefficiency
- How PaaS reduces costs
- How much can be saved using Apprenda
- What future benefits and options does private PaaS provide
I’ll also be hosting a webinar at the end of the month where I’ll be covering this information and more.
A number of months ago we began working closely with a large global manufacturer (Fortune 100) to help them with their transformation to private cloud. With development teams spread around the globe, they have written thousands of custom applications to support their business, with a majority of them in .NET.
The move to private cloud is driven by many factors, and Apprenda is helping this world class organization solve numerous challenges, but for this post I would like to explain how Apprenda’s private Platform-as-a-Service will reduce their infrastructure needs by as much as 75%.
The Current Environment
First, let’s get a general sense of their virtualized environment. Using large blade servers with lots of RAM they provision 15-20 VMs per physical server, with a standard 12 GB VM for each front end and middle tier instance, and all applications generally use a shared database tier. Each of those VMs then need to be configured with the appropriate OS and software, which includes Windows, IIS, SQL Server, and numerous other likely dependencies . With the exception of Windows (Windows DC Edition is licensed at the physical server level, but allows for unlimited VMs) each VM incurs licensing costs for things like middleware, security, etc.
As we look across the application portfolio we find that a large number of the applications have low memory requirements of 500MB to 1GB (remember that regardless of the memory requirements they are provisioning a standard 12 GB VM), many are just two tier, with some 3 tier.
You might ask yourself, why would they provision a 12 GB VM for an app that only requires 500 MB of memory?
In many cases, it is because the organization wants to use standard configurations, and secondly to provide for additional headroom to support peak usage, which often may only be a few weeks a year (think of a financial app that is used by 10 people in finance for most of the year, but gets 500 users the last week of each quarter as they close out the quarterly books). Since most applications are memory bound, RAM tends to be the boundary factor that people focus on for headroom.
What is the impact on company?
In a word, waste. Unfortunately, this is the reality of a world without something like Apprenda’s Platform-as-a-Service. Apprenda helps focus on driving finer grained resource utilization to squeeze boundaries in a way that represent the more realistic needs of applications.
How does Apprenda reduce VM sprawl?
First, let me explain a couple of functional components of Apprenda to provide context.
Apprenda is a software layer that stiches together any number of Windows servers into a peer-to-peer grid, creating a self-service cloud. (Read more about our product) Part of Apprenda’s functionality allows for something called Resource Allocation Policies (RAP), which, simply stated, allow the owners of the platform to define various sizes of CPU and memory quotas that get enforced inside the OS kernel at the process level rather than via coarse grained OS instance VMs. In layman’s terms, you could almost think of this as virtualizing virtualization (or virtualizing the application stack at the process level); meaning it allows you to provide even more granular slices of your infrastructure, without sacrificing scalability or high availability. In fact, the application can automatically be scaled out when required, and scaled down later to let other applications use the capacity. Remember that financial app above? The app owner would simply have chosen a “Small Policy” (say, 1Ghz CPU/1GB RAM), then towards the end of the quarter he would log in to Apprenda and scale out 2 more “Small Policies” during the peak time, and then reduce it later as appropriate (this would take about 5-10 mins with Apprenda)
OK, but how does Apprenda reduce VMs?
For simplicity, let’s look at how the customer might use Apprenda RAPs to provide much better efficiency of just one VM to create 6 Small (500mb RAM), 3 Medium (1GB), 3 Large (2GB). Apprenda would also provide that same capability for CPU.
So the customer is now able to support many more applications on the same footprint, while also shortening time-to-market from 60 days to 10 mins (that’s a post for another day), gaining automatic HA, and total elasticity. Let’s not forget the APIs they have access to for building next generation cloud applications. For those that would like to see the math
Because a VM is allocated 12 GB VMs regardless of actual need you can see how RAM is inefficiently used. And let’s not forget that it costs money to manage each VM, and more money to put software licenses on each VM
How and where Apprenda reduces cost
Apprenda reduces costs for our customer in the following areas:
Our customers routinely see technology paybacks of 30-180 days, and ROIs between 15-50%.
Many large organizations use an outsourcing firm to administer the VMs and physical servers. Because we can’t disclose the actual outsourcing costs, we’ll use the Azure costs as a benchmark (outsourcing can be higher). Azure charges $4,200 per large VM /year.
$4200 x (36 – 4 VMs) = $134,000 savings each year on operations
Depending on how many VMs per physical, the VM reduction will drive hardware reduction, which also drives some other licensing savings that are tied to physical machines.
With each VM reduction there is an associated software licensing cost. We have seen this range for $3,000 – $10,000 per VM depending on what is required. This can add up for a large global company with thousands of applications to many millions of dollars.
What future benefits would we see post-implementation?
Our customers see numerous future benefits, but some of the top ones are: faster time to market with applications, improved agility, 50% increase in server utilization, higher confidence in IT, ability to write and architect multi-tenant cloud applications 18 months faster, better adherence to enterprise standards, automatically scale out to hybrid clouds, applications can automatically inherit compliance requirements (PHI, HIPAA, EAR, ITAR, etc)
But, what if I’m different from a Fortune 100 manufacturing company?
Our sales teams can work directly with you and the various stakeholders to build a custom ROI and quantitative business case so you can see exactly where and how you will see savings. Call us at (518)383-2130 or contact sales now.
If you have found this post interesting and you’d like to see more on this topic, I would encourage you to sign for my webinar taking place on October 30th.