Home Big Data GoDaddy benchmarking leads to as much as 24% higher price-performance for his or her Spark workloads with AWS Graviton2 on Amazon EMR Serverless

GoDaddy benchmarking leads to as much as 24% higher price-performance for his or her Spark workloads with AWS Graviton2 on Amazon EMR Serverless

GoDaddy benchmarking leads to as much as 24% higher price-performance for his or her Spark workloads with AWS Graviton2 on Amazon EMR Serverless


It is a visitor put up co-written with Mukul Sharma, Software program Improvement Engineer, and Ozcan IIikhan, Director of Engineering from GoDaddy.

GoDaddy empowers on a regular basis entrepreneurs by offering all the assistance and instruments to succeed on-line. With greater than 22 million prospects worldwide, GoDaddy is the place individuals come to call their concepts, construct an expert web site, appeal to prospects, and handle their work.

GoDaddy is a data-driven firm, and getting significant insights from knowledge helps us drive enterprise selections to please our prospects. At GoDaddy, we launched into a journey to uncover the effectivity guarantees of AWS Graviton2 on Amazon EMR Serverless as a part of our long-term imaginative and prescient for cost-effective clever computing.

On this put up, we share the methodology and outcomes of our benchmarking train evaluating the cost-effectiveness of EMR Serverless on the arm64 (Graviton2) structure towards the normal x86_64 structure. EMR Serverless on Graviton2 demonstrated a bonus in cost-effectiveness, leading to vital financial savings in whole run prices. We achieved 23.85% enchancment in price-performance for pattern manufacturing Spark workloads—an final result that holds great potential for companies striving to maximise their computing effectivity.

Resolution overview

GoDaddy’s clever compute platform envisions simplification of compute operations for all personas, with out limiting energy customers, to make sure out-of-box value and efficiency optimization for knowledge and ML workloads. As part of this imaginative and prescient, GoDaddy’s Information & ML Platform staff plans to make use of EMR Serverless as one of many compute options beneath the hood.

The next diagram reveals a high-level illustration of the clever compute platform imaginative and prescient.

Benchmarking EMR Serverless for GoDaddy

EMR Serverless is a serverless possibility in Amazon EMR that eliminates the complexities of configuring, managing, and scaling clusters when working large knowledge frameworks like Apache Spark and Apache Hive. With EMR Serverless, companies can take pleasure in quite a few advantages, together with cost-effectiveness, sooner provisioning, simplified developer expertise, and improved resilience to Availability Zone failures.

At GoDaddy, we launched into a complete research to benchmark EMR Serverless utilizing actual manufacturing workflows at GoDaddy. The aim of the research was to judge the efficiency and effectivity of EMR Serverless and develop a well-informed adoption plan. The outcomes of the research have been extraordinarily promising, showcasing the potential of EMR Serverless for our workloads.

Having achieved compelling leads to favor of EMR Serverless for our workloads, our consideration turned to evaluating the utilization of the Graviton2 (arm64) structure on EMR Serverless. On this put up, we concentrate on evaluating the efficiency of Graviton2 (arm64) with the x86_64 structure on EMR Serverless. By conducting this apples-to-apples comparative evaluation, we intention to achieve worthwhile insights into the advantages and concerns of utilizing Graviton2 for our large knowledge workloads.

By utilizing EMR Serverless and exploring the efficiency of Graviton2, GoDaddy goals to optimize their large knowledge workflows and make knowledgeable selections concerning essentially the most appropriate structure for his or her particular wants. The mix of EMR Serverless and Graviton2 presents an thrilling alternative to boost the info processing capabilities and drive effectivity in our operations.

AWS Graviton2

The Graviton2 processors are particularly designed by AWS, using highly effective 64-bit Arm Neoverse cores. This practice-built structure gives a outstanding enhance in price-performance for numerous cloud workloads.

By way of value, Graviton2 presents an interesting benefit. As indicated within the following desk, the pricing for Graviton2 is 20% decrease in comparison with the x86 structure possibility.

   x86_64  arm64 (Graviton2) 
per vCPU per hour $0.052624 $0.042094
per GB per hour $0.0057785 $0.004628
per storage GB per hour* $0.000111

*Ephemeral storage: 20 GB of ephemeral storage is out there for all employees by default—you pay just for any further storage that you simply configure per employee.

For particular pricing particulars and present data, check with Amazon EMR pricing.

AWS benchmark

The AWS staff carried out benchmark checks on Spark workloads with Graviton2 on EMR Serverless utilizing the TPC-DS 3 TB scale efficiency benchmarks. The abstract of their evaluation are as follows:

  • Graviton2 on EMR Serverless demonstrated a median enchancment of 10% for Spark workloads by way of runtime. This means that the runtime for Spark-based duties was decreased by roughly 10% when using Graviton2.
  • Though nearly all of queries showcased improved efficiency, a small subset of queries skilled a regression of as much as 7% on Graviton2. These particular queries confirmed a slight lower in efficiency in comparison with the x86 structure possibility.
  • Along with the efficiency evaluation, the AWS staff thought-about the associated fee issue. Graviton2 is obtainable at a 20% decrease value than the x86 structure possibility. Taking this value benefit under consideration, the AWS benchmark set yielded an total 27% higher price-performance for workloads. Because of this by utilizing Graviton2, customers can obtain a 27% enchancment in efficiency per unit of value in comparison with the x86 structure possibility.

These findings spotlight the numerous advantages of utilizing Graviton2 on EMR Serverless for Spark workloads, with improved efficiency and cost-efficiency. It showcases the potential of Graviton2 in delivering enhanced price-performance ratios, making it a horny alternative for organizations looking for to optimize their large knowledge workloads.

GoDaddy benchmark

Throughout our preliminary experimentation, we noticed that arm64 on EMR Serverless constantly outperformed or carried out on par with x86_64. One of many jobs confirmed a 7.51% improve in useful resource utilization on arm64 in comparison with x86_64, however as a result of cheaper price of arm64, it nonetheless resulted in a 13.48% value discount. In one other occasion, we achieved a powerful 43.7% discount in run value, attributed to each the cheaper price and decreased useful resource utilization. Total, our preliminary checks indicated that arm64 on EMR Serverless delivered superior price-performance in comparison with x86_64. These promising findings motivated us to conduct a extra complete and rigorous research.

Benchmark outcomes

To realize a deeper understanding of the worth of Graviton2 on EMR Serverless, we performed our research utilizing real-life manufacturing workloads from GoDaddy, that are scheduled to run at a each day cadence. With none exceptions, EMR Serverless on arm64 (Graviton2) is considerably less expensive in comparison with the identical jobs run on EMR Serverless on the x86_64 structure. In actual fact, we recorded a powerful 23.85% enchancment in price-performance throughout the pattern GoDaddy jobs utilizing Graviton2.

Just like the AWS benchmarks, we noticed slight regressions of lower than 5% within the whole runtime of some jobs. Nevertheless, on condition that these jobs can be migrated from Amazon EMR on EC2 to EMR Serverless, the general whole runtime will nonetheless be shorter as a result of minimal provisioning time in EMR Serverless. Moreover, throughout all jobs, we noticed a median velocity up of two.1% along with the associated fee financial savings achieved.

These benchmarking outcomes present compelling proof of the worth and effectiveness of Graviton2 on EMR Serverless. The mix of improved price-performance, shorter runtimes, and total value financial savings makes Graviton2 a extremely engaging possibility for optimizing large knowledge workloads.

Benchmarking methodology

As an extension of a bigger benchmarking EMR Serverless for GoDaddy research, the place we divided Spark jobs into brackets based mostly on whole runtime (quick-run, medium-run, long-run), we measured impact of structure (arm64 vs. x86_64) on whole value and whole runtime. All different parameters have been saved the identical to attain an apples-to-apples comparability.

The staff adopted these steps:

  1. Put together the info and atmosphere.
  2. Select two random manufacturing jobs from every job bracket.
  3. Make essential adjustments to keep away from inference with precise manufacturing outputs.
  4. Run checks to execute scripts over a number of iterations to gather correct and constant knowledge factors.
  5. Validate enter and output datasets, partitions, and row counts to make sure similar knowledge processing.
  6. Collect related metrics from the checks.
  7. Analyze outcomes to attract insights and conclusions.

The next desk reveals the abstract of an instance Spark job.

Metric  EMR Serverless (Common) – X86_64  EMR Serverless (Common) – Graviton  X86_64 vs Graviton (% Distinction) 
Whole Run Price $2.76 $1.85 32.97%

Whole Runtime


00:41:31 00:34:32 16.82%
EMR Launch Label emr-6.9.0
Job Sort Spark
Spark Model Spark 3.3.0
Hadoop Distribution Amazon 3.3.3
Hive/HCatalog Model Hive 3.1.3, HCatalog 3.1.3

Abstract of outcomes

The next desk presents a comparability of job efficiency between EMR Serverless on arm64 (Graviton2) and EMR Serverless on x86_64. For every structure, each job was run not less than thrice to acquire the correct common value and runtime.

 Job  Common x86_64 Price Common arm64 Price Common x86_64 Runtime (hh:mm:ss) Common arm64 Runtime (hh:mm:ss)  Common Price Financial savings %  Common Efficiency Acquire % 
1 $1.64 $1.25 00:08:43 00:09:01 23.89% -3.24%
2 $10.00 $8.69 00:27:55 00:28:25 13.07% -1.79%
3 $29.66 $24.15 00:50:49 00:53:17 18.56% -4.85%
4 $34.42 $25.80 01:20:02 01:24:54 25.04% -6.08%
5 $2.76 $1.85 00:41:31 00:34:32 32.97% 16.82%
6 $34.07 $24.00 00:57:58 00:51:09 29.57% 11.76%
Common  23.85% 2.10%

Observe that the development calculations are based mostly on higher-precision outcomes for extra accuracy.


Primarily based on this research, GoDaddy noticed a big 23.85% enchancment in price-performance for pattern manufacturing Spark jobs using the arm64 structure in comparison with the x86_64 structure. These compelling outcomes have led us to strongly advocate inside groups to make use of arm64 (Graviton2) on EMR Serverless, besides in circumstances the place there are compatibility points with third-party packages and libraries. By adopting an arm64 structure, organizations can obtain enhanced cost-effectiveness and efficiency for his or her workloads, contributing to extra environment friendly knowledge processing and analytics.

Concerning the Authors

Mukul Sharma is a Software program Improvement Engineer on Information & Analytics (DnA) group at GoDaddy. He’s a polyglot programmer with expertise in a big selection of applied sciences to quickly ship scalable options. He enjoys singing karaoke, enjoying numerous board video games, and dealing on private programming tasks in his spare time.

Ozcan Ilikhan is a Director of Engineering on Information & Analytics (DnA) group at GoDaddy. He’s captivated with fixing buyer issues and rising effectivity utilizing knowledge and ML/AI. In his spare time, he loves studying, mountain climbing, gardening, and dealing on DIY tasks.

Harsh Vardhan Singh Gaur is an AWS Options Architect, specializing in analytics. He has over 6 years of expertise working within the discipline of massive knowledge and knowledge science. He’s captivated with serving to prospects undertake greatest practices and uncover insights from their knowledge.

Ramesh Kumar Venkatraman is a Senior Options Architect at AWS who’s captivated with containers and databases. He works with AWS prospects to design, deploy, and handle their AWS workloads and architectures. In his spare time, he likes to play together with his two children and follows cricket.



Please enter your comment!
Please enter your name here