We've got great news for the countless StarRocks users who've adopted the project for its unrivaled performance and cost of ownership. We have expanded our CPU architecture options to support ARM, and, even better, AWS Graviton customers can now take advantage of this development and enjoy an over 50% price-performance boost to their analytics operations.

In this article, we'll provide some background to these improvements and explain just how you can enjoy a dramatically lower TCO and even lower latency through these changes.

 

More Than Just x86: Embracing ARM

Support for ARM has long been on our roadmap. Users regularly request it, thanks to its more streamlined instruction set compared to x86, which offers lower power consumption. This cost efficiency is why ARM architectures are widely used by smartphone and computer manufacturers, and why it's attracting more and more cloud providers (like AWS).

Additionally, the growth of ARM architectures in data centers has been significant, particularly in the high-performance computing (HPC) field. ARM architectures save energy, are more environmentally friendly, and offer better cost-performance, especially for compute-intensive applications optimized for ARM. Taken together, it's easy to see why support for ARM has been such a common request, and why we've prioritized it.

 

What Is AWS Graviton?

For those not familiar with Graviton, here's a quick overview. AWS Graviton is a custom processor developed by AWS based on the ARM architecture, specifically designed to optimize cloud computing performance and cost efficiency. Compared to its predecessor Graviton2, Graviton3 offers significant performance enhancements, with a 25% increase in single-thread performance and a 50% overall performance boost. On top of this:

  • Graviton3 supports DDR5 memory, providing higher memory bandwidth and lower latency, making it exceptionally effective for large-scale data analysis, high-performance computing, and memory-intensive applications.

  • The Graviton3 processor integrates advanced hardware security features, including built-in memory encryption and more efficient encryption algorithm support, ensuring data security during transmission and storage, and meeting the requirements for handling sensitive data.

  • Graviton3 boasts higher energy efficiency, reducing power consumption at the same performance level, which is crucial for enterprises looking to lower operational costs and their carbon footprint.

 

It's clear why so many users prefer to work with Graviton, and it was a significant reason why ARM compatibility was prioritized on our roadmap. Understanding that, let's take a look at how all of this comes together in unbeatable cost-performance benefits for you.

 

New Benefits for AWS Graviton3 Users

For businesses that have been utilizing StarRocks for SIMD-powered real-time analytics and high-performance data processing, the transition to ARM promises reduced operational costs and improved performance. Similar to x86 architectures and their SSE/AVX instructions, ARM, has a similar NEON instruction set that StarRocks utilizes to achieve remarkable performance improvements. We accomplished this in the latest 3.3 release that deeply optimizes its NEON instruction set for bitshuffle, CRC, and filter_range.

You can see this in the average performance gain for different operators(higher is better for ARM):

Operator

ARM/x86

Scan & Bitshuffle

1.15

Aggregate

1.43

HashJoin

1.28

 

This optimization is available across all StarRocks solutions leveraging AWS Graviton3 and provides a seamless transition for those already invested in the AWS ecosystem. Initial benchmarks and user case studies have shown promising results, with clients reporting substantial improvements in both speed and cost efficiency., making StarRocks a unique and leading solution in the ARM space. Now users can enjoy cost-savings and performance once reserved only for x86.

If you're trying to stay ahead in this competitive business landscape (and who isn't), we encourage you to experience these benefits firsthand by deploying StarRocks on AWS Graviton3 today.

 

53% Improvements To Price-Performance: Test Results

So what can you expect from these changes? Let's take a look at StarRocks' latest benchmarking efforts with ARM.

 

Benchmark Test

Node types selected for this test:

 

ARM

x86

3 BE

c7g.4xlarge

c6i.4xlarge

 

500GB SSD EBS GP3

500GB SSD EBS GP3

1 FE

c7g.4xlarge

c6i.4xlarge

 

200GB SSD EBS GP3

200GB SSD EBS GP3

In a standard TPC-H benchmark test at 100GB and 1TB scale, compared to the C6i instance type, the C7g instance type showed no performance regressions and achieved an average performance improvement of 30%. Combined with a 15% cost optimization for the C7g instance type, this results in a price-performance ratio of 1.3 / 0.85 = 1.53, representing an overall improvement in price-performance of 53%.

 

Test Sets

C6i (ms)

C7g (ms)

C6i / C7g

TPC-DS 100G

77473

57427

1.35

TPC-DS 1T

573921

431123

1.33

 

These results are impressive, to say the least, and we encourage you to run your own tests to validate just how much these latest updates enhance your current environment.

 

Get These Results Yourself With CelerData Cloud

You can take advantage of this ARM compatibility utilizing StarRocks' source code, however, there is no binary available for open-source users, and you will need to compile it by yourself. There is an easier way, though.

The best way to leverage StarRocks on Graviton is through CelerData Cloud BYOC. When creating a new cluster, you can easily choose the desired architecture without the need to compile and manage additional dependencies. This leads to a fast and easy setup process.

celerdata cloud

Furthermore, CelerData Cloud BYOC will soon support smooth AMI upgrades from x86 to Graviton, allowing you to upgrade seamlessly in the background. This ensures a hassle-free transition and continuous operation without disruptions, making it the most efficient and user-friendly option for maximizing the benefits of StarRocks on Graviton. So what are you waiting for? Try everything for yourself with a free trial of CelerData Cloud today.

Join StarRocks Community on Slack

Connect on Slack
copy success