We commissioned Forrester Consulting to conduct a study describing how innovative technology companies utilized Azul Platform Prime to reduce hardware expenses, increase the performance and scalability of JVM applications, and reduce the amount of effort their development teams need to expend on debugging JVM-related problems. The study found that the total economic impact from utilizing Platform Prime provided savings and benefits across a broad spectrum of key elements that factor into the costs of running a company whose product is a software service, with an initial cost that is repaid after three months of the investment in switching to Azul Platform Prime.
The study involved interviews with knowledgeable development and management staff from four global companies:
- Head of Software for a Financial Technology company that deploys 250 Prime servers
- VP IT for a Web Ad Technology company that deploys 1750 Azul Platform Prime servers and 2000 other Java servers
- Chief Platform Architect for a Financial Services company that deploys 1000 Prime servers
- Java Architect for an Online Shopping company that deploys 100 Azul Platform Prime servers and more than 10000 other Java servers
The study contains quotes and data from the individual interviewees; but the Forrester study also agglomerates the data from all four companies by creating a “Composite Organization” that evaluates: “What would the total economic impact be if all four of these companies were components of a single company?” Applying this methodology, Forrester arrived at two very important conclusions:
- the cost for the Composite Organization having switched to Platform Prime was paid back by the cost savings Prime provided within three months
- the return on investment (ROI) for the Composite Organization over a three-year period was 224%
Forrester Total Economic Impact Categories
To arrive at these numbers, Forrester evaluated several categories of economic impact identified both by the interviewees and Forrester’s own staff. The study found that the primary factors in determining the economic impact of Prime for a company are:
- Developer efficiency related to code simplification
- High-performance server costs related to JVM memory capacity
- JVM licensing costs
- Developer time spent troubleshooting and recovering from JVM-related issues
- Impact of JVM outages and failover strategies
- Cost of re-architecting performance-critical software components to facilitate increased scalability
Platform Prime Reduces the Complexity of Necessary Code Enhancements
The Forrester study reports that “All customers saw increased developer productivity after switching low latency workloads to Azul Platform Prime. Azul Platform Prime simplified coding by eliminating specialized coding required to minimize latency issues and allowed customers to use open source libraries to further reduce custom coding.”
This is a reduction in custom coding for a development staff simply to enable the company’s product to run correctly under evolving situations, including the need for increased performance and the ability to utilize the already-developed open source solutions, rather than requiring your developer staff to “re-invent the wheel” at high development expense.
Platform Prime’s Vastly Increased JVM Heap Memory Capacity Reduces Server Costs
The interviewees in the Forrester study noted that JVM heap memory was a decisive factor in determining how many servers were needed in order for them to serve their customer base. The study notes: “Heap memory size plays a significant role in avoiding garbage collection issues, both with the original JVM and with Prime. The original JVM for all of the customers could only address 32 GB of heap memory per server, while with Platform Prime some of the customers are addressing 1.5 TB of heap memory per server.”
Bottlenecks are always costly. The 32 GB of heap memory in standard JVMs is a problem for high-performance software that must be scalable. You may have the necessary processing power in your servers, but when you run out of heap memory in that server, you’re going to have to buy another server to run a new JVM that will provide you with another 32 GB of heap memory.
Azul Platform Prime’s capability to provide 1.5 TB of heap memory per server means that, rather than buying a new server, you can simply add memory to an existing server. Not only is adding memory to an existing server less costly than buying a new server, it also reduces the burden on your IT team that’s charged with configuring, managing, and maintaining your server farm.
The Forrester study states: “One rapidly growing customer was able to avoid adding 50 new servers for over the first two years that they had Azul Platform Prime.” Another interviewee said “We moved from horizontal scaling to vertical scaling.”
Azul Platform Prime Reduces JVM Licensing Costs
Eliminating the need to add many new servers reduces the number of JVM licenses your enterprise requires to implement your product. In addition, the cost for Azul Platform Prime licensing is substantially lower than the cost of competing JVM licenses. For the companies Forrester interviewed, the cost of Platform prime licensing was well under 1% of the cost the companies were previously paying for JVM licensing, producing an immense ROI.
Azul Platform Prime Substantially Reduces Developer Effort to Troubleshoot and Recover from JVM Issues
Forrester found that, prior to implementing Platform Prime, for all four companies about 20% of developer effort was expended on “troubleshooting JVM-related issues and resolving them.” In the Composite Organization Forrester created, this 20% was reduced to 5%, freeing the development staff to spend 95% of their time on advancing the company’s product. You hire developers to advance your product. Any time spent troubleshooting the underlying structure upon which your product is created is developer time not spent on advancing your product. Platform Prime reduced the developer effort required to troubleshoot the underlying JVM structure by 75% for the four interviewed companies.
Azul Platform Prime Reduces JVM Outages and Supports Quicker Failovers
If your company requires a near-instantaneous response from your software, then JVM outages are a major concern. As well, if your failover strategy involves booting a JVM that has a long warm-up period, then you suffer from customers needing your data immediately, while your servers take too long to respond to meet your customer needs.
The Chief Platform Architect with the financial services firm noted: “We work in microseconds — a pause of 3 seconds will lose 15K to 30K transactions, which is unacceptable.”
Platform Prime addresses these problems in two ways wherein it significantly increases JDK temporal efficiency: garbage collection and warm-up time. Almost every developer working on a large highly-scaled JDK-based product will be immediately familiar with these problematic difficulties with standard JDK products.
The Java Architect at the online shopping company said: “One system experienced frequent fail overs due to garbage collection issues that virtually disappeared when we implemented Azul Platform Prime.”
Another point cited in the Forrester study: companies often guarantee a certain number of 9’s in their Service Level Agreements (SLAs). For example, they guarantee that certain performance will happen at a 4 9’s level (99.99 percent of the time) or a 5 9’s level (99.999 percent of the time). It’s a big problem for your customers if you do not meet your promised SLA.
Forrester reported that, after converting to Azul Platform Prime, “Customers are consistently exceeding their service-level commitments. Note that two customers measure their performance in microseconds; and one customer previously failed to meet a 99.5% SLA while now performing between 99.99% and 99.999%.”
Azul Platform Prime Reduces the Need to Re-Architect Your Product to Facilitate Increased Scalability
If you’ve got a great product, the last thing you want to have to do is to re-architect it because your customer base is growing rapidly. You already created the product your customers want. Why should having too many new customers be a problem?
Well, that can indeed be a problem if your product is not readily scalable. Scaling is always a problem. But any solution that eases the process of scaling is going to be a major benefit to your company. Scaling applications that rely on traditional JVMs suffers from various performance weaknesses inherent in standard JVMs.
If you have to send developers into your low-level code in order to increase its performance, then you’re not actually working on your product, you’re working on making your product available to more customers without enhancing the product.
Forrester described this situation this way: “Within their existing architecture, some customers were seeing growth
with associated performance issues that would have required rearchitecting their performance critical solution components. All rearchitecting options were expensive, ranging from adding additional servers to gain more memory, to implementing code optimization schemes, and to the extreme of moving to a high-cost commercial development platform outside of Java.”
Imagine the cost of moving away from a JDK-based product due to scalability issues?
Conclusion
According to the Forrester study, a company move to the Azul Platform Prime (which requires no changes to the company’s application) pays for itself within three months. Furthermore, over a three-year timespan, the Total Economic Impact ROI for the company from switching from a standard JVM to Platform Prime is 224%. To access the full study, The Total Economic Impact™ of Azul Platform Prime, click here.
