Optimizing Hardware Specifications for HeLa Nodes

In the world of blockchain technology, ensuring optimal hardware specifications for node operators is crucial for maintaining efficient and cost-effective operations. While many blockchain platforms offer recommended hardware specifications, these recommendations are often designed for general use cases and may not be tailored to the specific needs of HeLa Nodes. This article aims to provide a comprehensive analysis of determining optimal hardware specs for HeLa Nodes based on the research conducted by the Klaytn Foundation and their recommendations for Klaytn nodes.

Understanding HeLa Nodes and Their Requirements

HeLa Nodes play a vital role in the HeLa blockchain network, responsible for transaction processing and block creation. To determine the optimal hardware specifications for HeLa Nodes, it is important to consider key factors such as storage, memory, and CPU performance.

1. Storage: The amount of accumulated data in the blockchain can significantly impact the speed of block creation and processing. Therefore, it is essential to ensure sufficient storage capacity with high Input/Output Operations Per Second (IOPS) for efficient performance.
2. Memory: Memory size directly affects performance by reducing the frequency and volume of Input/Output (I/O) access. However, it is important to strike a balance as excessively increasing memory size may not necessarily decrease the cache miss rate, a crucial performance metric.
3. CPU: While parallel processing is important for various functions, such as transaction management and block synchronization, single-core performance has the most significant impact on overall node performance. This is particularly important for critical tasks like block processing.

Benchmarking and Experimentation

To determine the optimal hardware specifications for HeLa Nodes, the Klaytn Foundation conducted an experiment on a private HeLa network. The experiment involved varying the CPU, memory, and storage specifications of Consensus Nodes (CNs). The goal was to identify the hardware configuration that delivers the highest Transactions Per Second (TPS) while ensuring stability and avoiding out-of-memory (OOM) issues.

The experiment revealed several key findings:

1. CPU and Memory: The performance of HeLa Nodes is influenced by the number of CPU cores and memory capacity. Nodes with higher CPU core count demonstrated better TPS, emphasizing the importance of CPU performance. Additionally, memory size should be optimized to achieve the lowest cache miss rate, which typically requires a minimum of 100 GB.
2. Storage and IOPS: Higher IOPS resulted in better overall performance. It is recommended to use storage solutions with high IOPS, such as gp3 with 16,000 IOPS and 500 MiB/s throughput.
3. the CPU benchmarks and recommendations: we have used the PassMark CPU Mark score as the main metric to measure the CPU performance. The PassMark CPU Mark score is a composite score that reflects the overall performance of a CPU based on various tests and benchmarks. The higher the PassMark CPU Mark score, the better the CPU performance. We have also used the Amazon Web Services (AWS) EC2 pricing as the main metric to measure the CPU cost. The AWS EC2 pricing is the hourly cost of renting a virtual machine with a certain CPU configuration on the AWS cloud platform. The lower the AWS EC2 pricing, the lower the CPU cost. We have also considered the power consumption and the availability of the CPUs when choosing the CPU benchmarks and recommendations.
4. The table below shows the PassMark CPU Mark scores, the AWS EC2 pricings, the power consumptions, and the availability of some of the CPUs that are commonly used for HeLa Nodes. The CPUs are sorted by their PassMark CPU Mark scores in descending order.

              Fig CPU Mark Score

Based on the table and the above figure, we can see that the AMD EPYC CPUs have higher PassMark CPU Mark scores and lower AWS EC2 pricings than the Intel Xeon CPUs, which means that they have better performance and lower cost. However, the AMD EPYC CPUs also have higher power consumption than the Intel Xeon CPUs, which means that they have higher energy and cooling requirements. Moreover, the AMD EPYC CPUs and the Intel Xeon CPUs have similar availability, which means that they are both widely accessible and compatible with various platforms and services.

Based on these factors, we recommend the following CPUs for the different types of HeLa Nodes:

• Validator Nodes: We recommend the AMD EPYC 7763 CPU for the Validator Nodes, as it has the highest PassMark CPU Mark score among the CPUs, which means that it has the highest processing power and speed. This CPU can handle the high network load and performance pressure of the Validator Nodes, as well as the large number of transactions and blocks that need to be processed and produced. However, this CPU also has the highest AWS EC2 pricing and the highest power consumption among the CPUs, which means that it has the highest operating cost and the highest energy and cooling requirements. Therefore, this CPU is suitable for the Validator Nodes that have a high budget and a high performance goal, as well as a reliable and efficient power and cooling system.
• Budget and performance requirement: The AMD EPYC 7402 CPU offers a cost-effective solution for Full Nodes that need to balance between performance and operational costs. With its moderate power consumption, it also presents a sustainable option for organizations that are conscious of their energy usage and environmental impact. Thus, this CPU is ideal for Full Nodes that aim for efficiency and eco-friendliness without compromising on the necessary computational capabilities.

Updated Recommended Hardware Specs for HeLa Nodes

Based on the experiment results and analysis, the Klaytn Foundation has updated their recommended hardware specifications for HeLa Nodes. These specifications aim to reduce operating costs while ensuring optimal performance. The recommended specs are as follows:

1. Consensus Nodes (CN):
   • AWS: m6i.8xlarge or similar (32 cores or more, 100GB or more)
   • Azure: D32s v5 or similar (32 cores or more, 100GB or more)
2. Proxy Nodes (PN):
   • AWS: m6i.4xlarge or similar (16 cores or more, 60GB or more)
   • Azure: D16s v5 or similar (16 cores or more, 60GB or more)
3. Endpoint Nodes (EN):
   • AWS: r6i.2xlarge or similar (8 cores or more, 60GB or more)
   • Azure: E8 v5 or similar (8 cores or more, 60GB or more)

Cost Considerations and Customization

While the recommended hardware specifications provide a good starting point, it is important to consider individual usage needs and adjust the specifications accordingly. Operating costs can vary based on factors such as cloud provider, usage patterns, and deployment scale. Therefore, it is advisable for HeLa Node operators to carefully assess their requirements and determine the optimal hardware specifications to minimize costs.

Determining optimal hardware specifications for HeLa Nodes is crucial for achieving efficient and cost-effective blockchain operations. By considering factors such as storage, memory, and CPU performance, HeLa Node operators can ensure optimal performance and minimize operating costs. The research conducted by the Klaytn Foundation and their updated recommendations provide valuable insights and guidelines for HeLa Node operators. However, it is essential for operators to customize the hardware specifications based on their specific requirements and usage patterns to achieve the best results.

References:

Please contact me for details of HeLa spec. We had some performance testing on the different Hardware spec

https://klaytn.foundation/node-operator-optimal-specs/