ARM64 processors use less energy, allowing data centers to fit more servers into the same space compared to x86 hardware. This efficiency means more processing power per rack. Each ARM-based server rack consumes less power and needs less cooling than its x86 counterparts.
Scott Sellers, CEO of Azul, offers an alternative to the Oracle Java Development Kit with Azul Platform Core, designed for enterprise Java applications. He spoke with Computer Weekly about how processor architectures influence enterprise software development and the relevance of Java’s “write once, run anywhere” philosophy.
Nowadays, it’s not just Intel or AMD that hold the market. Companies like Amazon, Microsoft, and Google are investing heavily in ARM64-based server architectures for cloud services. Sellers emphasizes that ARM is often more cost-effective than x86. “Performance is now comparable, if not superior, to x86, while power efficiency is significantly improved,” he says.
As demand grows for ARM64 workloads, developers deal with multiple programming languages in public clouds—some require source code changes between x86 and ARM. But with Java, that’s different. “The beauty of Java is that the application doesn’t have to be modified. No changes are necessary. It really does just work,” Sellers explains.
Replatforming can be labor-intensive and require extensive testing when moving from x86 to ARM64. Using Java allows developers to write code once, and the Java runtime compiler adapts it for the target processor when the application runs, giving IT managers the flexibility to pick a platform based on cost or performance needs.
Java runs well on both x86 and ARM64. Azul’s customers report performance gains of 30% to 40% with their Java runtime engine on both platforms. Sellers notes that moving workloads to ARM64 not only saves costs but also increases speed, as the workload needs less processing to achieve similar performance. He points out that ARM64 compute nodes typically cost about 20% less than x86 equivalents, a factor that benefits the tech sector and keeps competition sharp among major players like Intel and AMD.
Some larger customers are now combining x86 and ARM64 in a hybrid cloud to get the benefits of both architectures. While more clients want ARM64, x86 still dominates public cloud setups. Sellers expects that to shift over time but acknowledges that many Azul clients are limited by supply when it comes to ARM64 nodes.
On the topic of GPUs, demand for Nvidia’s graphics chips has skyrocketed for AI workloads. GPUs excel at running many simple tasks simultaneously, making them ideal for AI. But they don’t fit every use case. For LMAX Group, a financial exchange that Azul works with, using GPUs wouldn’t work due to their workload’s complexity.
Sellers explains that while GPUs are great for specific uses, enterprise applications often require intricate parallel processing, making them less suitable for those situations. He argues that while Python is popular for AI, it primarily serves as a front end for GPU computation. Java, on the other hand, offers robust capabilities for traditional enterprise applications, particularly with the introduction of virtual threads in Java 21, enhancing multithreading and vectorization.
In a cloud-native future, decision-makers need to recognize that x86 is no longer the only option available. The landscape is changing, and so should the strategies for developing enterprise applications.