Background As the global financial market becomes increasingly complex and competitive, the financial industry is undergoing an unprecedented digital transformation. This transformation not only requires financial institutions to enhance the efficiency and security of business processes, but also to achieve new heights in return on investment, cost-effectiveness, and technological development trends, as well as in data management, analysis, and AI computational power. To meet the specific needs of the financial industry, many financial institutions have begun exploring virtualization solutions based on indigenous innovation to build a more efficient and flexible IT infrastructure. In this context, choosing the appropriate underlying server architecture has become one of the key decisions. Currently, the two mainstream architectures on the market are ARM and C86, each with its own advantages and limitations, suitable for different business applications or device requirements. How to make a scientific choice and what difficulties one might face in the selection process have become the main concerns.This article is derived from a community online seminar on the theme of “Choosing Between ARM and C86 for Financial Industry Virtualization” , focusing on key issues such as how to determine suitable underlying server architecture for the financial industry, how to differentiate in specific applications, the limitations of each architecture, and the current ecosystem. Below are selected excerpts from the peer exchanges in the financial industry and the final formed consensus, which have reference value for peers in making related architecture choices and solving difficulties.
Moderator and Expert Author of this Article:
Yao Yafei, Senior Engineer at a City Commercial Bank
Sharing Experts (Community ID):
Effort Mover, System Operation Engineer at a Joint-Stock Bank
nkj2021, System Architect at a Securities Company
waring_id, Technical Manager at a Company
xjwangbo100, System Architect at a City Commercial Bank
JAGXU, Storage Operation Management at a Financial Company
leqiyufeng, System Engineer at ewell
*The content of this article reflects the views of users and does not represent the position of the community.
Focus Topic Discussion Record
(1) ARM or C86, How to Choose Indigenous Innovation Servers?
◉ waring_id, Technical Manager:
The mainstream chip architectures today are X86 and ARM, both dominated by foreign entities. The degree of control over the instruction set determines the level of localization. Domestic CPUs usually operate under three models through licensing or self-research:
1. MIPS instruction set architecture represented by Loongson and Alpha architecture represented by Shenwei. Shenwei has basically achieved complete autonomy (Shenwei 64 has fully formed its own architecture), and Loongson is the earliest domestically developed chip in China, based on the MIPS architecture with permanent authorization, and has developed a completely autonomous instruction set called LongARCH on this foundation.
2. Domestic chips based on ARM instruction set authorization represented by Feiteng and Huawei Kunpeng, which can design core CPUs based on the instruction set architecture, providing a higher degree of security and control.
3. Domestic chips represented by Haiguang and Zhaoxin that have obtained X86 IP core authorization, although Haiguang later obtained X86 instruction set (AMD) authorization.
It can be seen that in terms of the degree of autonomy and control, Loongson and Shenwei > Feiteng and Kunpeng > Haiguang and Zhaoxin. A comprehensive comparison of the six major domestic CPU manufacturers:
• In terms of performance, the Haiguang series X86 architecture and the Kunpeng series ARM architecture have advantages.
• In terms of ecosystem, Haiguang and Zhaoxin have ecological advantages based on the X86 instruction set, while Feiteng and Kunpeng, represented by ARM architecture, are building their own ecosystem. Shenwei and Loongson do not have advantages in ecological expansion due to their self-researched instruction sets.
• In terms of autonomy and control over the instruction set, Shenwei and Loongson self-researched instruction set > Kunpeng and Feiteng based on ARM instruction set authorization > Haiguang and Zhaoxin based on X86 IP authorization.
◉ xjwangbo100, System Architect at Hami City Commercial Bank:
1. Performance gap
Computational performance: Although indigenous innovation servers have made certain progress in recent years, especially in self-developed processors (such as Loongson, Feiteng, etc.), there is still a certain performance gap compared to the mature X86 architecture, especially in high-performance computing and big data applications. This is due to the more mature optimization of X86 processors and the comprehensive support of hardware and software ecosystems.
Storage and memory bandwidth: The storage and memory bandwidth of indigenous innovation servers can sometimes become a bottleneck, affecting overall performance.
2. Compatibility issues
Operating system: Currently, the operating system support for indigenous innovation servers is not as widespread as that for X86 architecture. For some special applications or commercial software, additional adaptation work may be required. Therefore, when choosing, it is necessary to ensure that the selected server can be compatible with existing operating systems (such as domestic operating systems like Kylin, Galaxy Kylin, etc.).
Application software: Many software that relies heavily on X86 architecture (such as databases, middleware, etc.) may face compatibility issues on the indigenous innovation platform, which may require additional migration or modification work, especially in some cross-platform application scenarios.
Hardware compatibility: Although indigenous innovation servers are gradually improving hardware support, it is still necessary to ensure that the selected server has good compatibility with existing hardware resources (such as storage, network devices, etc.).
3. Ecosystem support and maturity
Software ecosystem: The X86 architecture has rich software support, while the ecosystem of indigenous innovation servers is still developing, especially in emerging fields like cloud computing and big data, which may face certain immaturity.
Industry support: Some industries provide strong support for indigenous innovation servers, such as government and finance, gradually increasing the pace of domestic substitution, but may still face limitations in other fields.
4. Selection advice:
Clarify requirements: First, clarify application scenarios and requirements, such as whether high-performance computing, virtualization, or containerization is needed. If the primary application scenario is general office or lightweight applications, indigenous innovation servers may meet the requirements. If a lot of data processing and high-performance computing is needed, a deeper assessment of performance gaps may be necessary.
Compatibility assessment: When selecting indigenous innovation servers, the compatibility with existing operating systems, application software, and hardware should be evaluated, especially during system migration and adaptation, which may require additional resource investment.
Select vendors with support: Some vendors have already done a lot of work in supporting indigenous innovation servers, such as Inspur, Huawei, etc. Choosing vendors with mature ecosystems and support may reduce compatibility issues.
Performance testing: Conduct actual performance testing, especially for workload performance comparisons, to ensure that indigenous innovation servers can meet business needs.
In summary, when selecting, one should not only focus on the performance improvement of indigenous innovation servers but also evaluate their compatibility with existing systems and applications in detail to ensure that the entire environment can run smoothly.
◉ Effort Mover, System Operation Engineer at a Joint-Stock Bank:
In comparison to C86 architecture, the ARM architecture currently has better compatibility. This is because previous business systems were mostly developed based on the X86 architecture, and the entire industry’s technology ecosystem has gradually built and improved around the X86 architecture. The related middleware, databases, and other software have undergone long-term optimization under the X86 architecture, making the compatibility system quite mature and stable, so adaptation will be simpler. In terms of performance, the ARM architecture can formulate personalized performance tuning plans according to different business requirements, and it can more accurately meet performance requirements when facing diverse business challenges, thus having an advantage.
(2) Which Architecture is More Stable for Supporting Financial Industry Applications, ARM or C86?
◉ nkj2021, System Architect at a Securities Company:
1. The ARM architecture has high compatibility and broad ecosystem support, which means its hardware and software development environment is more mature, providing a stable infrastructure for financial enterprises. The support for multi-core processors allows the ARM architecture to better utilize resources in complex financial operations, enhancing processing capacity.
2. The C86 architecture has advantages in specific embedded scenarios, but its availability is relatively low. Its resource-constrained characteristics lead to stability issues under high load or complex business requirements. The support capability of the C86 architecture is limited for the stability requirements of the financial industry.
3. The wide support and mature ecosystem of the ARM architecture give it high compatibility in the financial industry. More software development tools, ecosystem components, and third-party services are based on the ARM architecture, making system expansion and maintenance easier. It has high security, effectively preventing vulnerability exploitation and ensuring system stability.
4. The C86 architecture has certain shortcomings in compatibility within the financial industry. Its ecosystem is relatively closed, with fewer supported software tools, increasing system maintenance costs. For complex financial business requirements, the support capability of the C86 architecture is limited.
◉ JAGXU, Storage Operation Management at a Financial Company:
It should be C86, at least its compatibility is smoother.
(3) How are the Ecosystems of ARM and C86?
◉ nkj2021, System Architect at a Securities Company:
1. The ARM architecture is known for its high compatibility and broad software ecosystem, supporting high availability requirements. The ARM architecture can ensure that systems continue to operate normally even when critical components fail, thus ensuring service continuity. A broad support community and rich development resources help quickly resolve potential issues.
2. The C86 architecture is more suitable for specific embedded devices and has certain advantages in resource-constrained environments. However, due to the singularity of its instruction set and expansion limitations, fully supporting high availability requirements will face challenges. The scalability and maintainability of the C86 architecture are not as good as those of the ARM architecture.
3. The standardization level of the ARM instruction set is high, with good compatibility with various processors and development tools. A well-developed software ecosystem allows ARM-based servers to easily integrate existing software stacks and technologies. It supports multi-platform and multi-device development environments, suitable for complex financial enterprise indigenous innovation needs.
4. The C86 instruction set may be relatively singular, with certain limitations in compatibility with mainstream development tools and technology stacks. For complex financial business system integration, the C86 architecture requires additional adaptation and compatibility measures, increasing development and maintenance complexity.
◉ Effort Mover, System Operation Engineer at a Joint-Stock Bank:
The C86 fully compatible with mainstream X86 has this foundation. Currently, both in terms of ecosystem and compatibility, it is better than ARM. However, as the indigenous innovation transformation deepens, more and more enterprises have joined the ARM ecosystem, which will become increasingly robust in the future. If focusing on compatibility with traditional systems and the completeness of the software ecosystem, and needing to achieve a relatively smooth indigenous innovation transformation in the short term, the X86 architecture may be a more suitable choice. If focusing on the autonomy and control of indigenous innovation and willing to explore and adapt during the gradual improvement of the ARM ecosystem, the ARM architecture can be considered.
(4) How is the Performance of ARM and C86?
◉ nkj2021, System Architect at a Securities Company:
1. The ARM instruction set integrates efficient computing capabilities and powerful optimization mechanisms, making it particularly suitable for handling high-performance tasks. In high-load scenarios such as financial transaction processing, data analysis, and distributed computing, the ARM architecture performs excellently, meeting the system’s performance requirements. The support for multi-core processors provides significant advantages for parallel computing in the ARM architecture.
2. Due to resource constraints, the C86 instruction set faces performance bottlenecks when handling complex tasks. In terms of performance optimization, the flexibility of the C86 architecture is low, making it difficult to fully utilize multi-core processors and complex computing demands. For tasks that require high concurrency and low latency, the C86 architecture does not perform as well as the ARM architecture.
3. The ARM architecture performs excellently in the indigenous innovation virtualization underlying server architecture for financial enterprises. Its high compatibility, strong performance support, comprehensive security mechanisms, and good scalability make it more suitable for complex financial business needs. However, the C86 architecture has certain advantages in specific embedded scenarios, especially in resource-constrained and cost-sensitive situations.
◉ yyf123, System Engineer at Weihai City Commercial Bank:
There are two aspects to consider. One is the performance of indigenous innovation virtualization, and the other is the performance of the server. In specific selection, we also need to consider the continuous development capabilities of the products, especially at the indigenous innovation virtualization level.
In the current indigenous innovation virtualization selection, the main products rely on KVM for encapsulation, and each product cannot be compatible with other products’ virtual machines. More active products include Alibaba, Tencent, Huawei, Inspur, H3C, Yunhong, etc. Personally, I believe that when selecting products, priority should be given to vendors engaged in IaaS while also having public cloud offerings, as these vendors’ virtualization products are relatively stable and provide better service. Next, select vendors of private cloud products, and finally choose pure software vendors. The main consideration is that cloud is the trend, and enterprises usually directly use the virtualization products provided by the cloud; pure software vendor solutions are relatively singular and have lower risk resistance.
In terms of server performance, ARM is currently slightly better than C86. In the ecosystem, there is not much difference between the two domestic architectures on mainstream software, and specific comparative data is rarely provided.
In terms of application transformation, C86 has a slight advantage over ARM, but due to the migration to the indigenous innovation environment, replacing the operating system, middleware, database, etc., application transformation is necessary, so the advantage is not significant.
(5) Which Architecture Has Stronger Compatibility, ARM or C86?
◉ yyf123, System Engineer at Weihai City Commercial Bank:
ARM and C86 are the current mainstream architectures for financial indigenous innovation. The ARM architecture mainly comes from Huawei’s product system, which was later opened for use by various hardware vendors. Huawei customized the BIOS before opening it, and the starting point of the BIOS is relatively high. Currently, the ARM architecture performs slightly better than C86; in the early stage of financial indigenous innovation, using the ARM architecture has certain adaptation difficulties, as there is not much good experience. With the advancement of indigenous innovation, the mainstream database, middleware, and operating system products are basically clear, and choosing either architecture can meet the requirements. Our own database and high-performance applications are mainly based on ARM, while C86 is mainly used for application execution, which perfectly solves the dual-route issue and ensures a smooth transition of applications.
◉ nkj2021, System Architect at a Securities Company:
1. The standardization level of the ARM instruction set is high, with good compatibility with various processors and development tools. A well-developed software ecosystem allows ARM-based servers to easily integrate existing software stacks and technologies. It supports multi-platform and multi-device development environments, suitable for complex financial enterprise indigenous innovation needs.
2. The C86 instruction set is relatively singular, with certain limitations in compatibility with mainstream development tools and technology stacks. For complex financial business system integration, the C86 architecture requires additional adaptation and compatibility measures, increasing development and maintenance complexity.
3. The ARM architecture supports multi-processor and multi-core designs, easily coping with the increased hardware demand brought by future business growth. The flexible software stack design ensures good scalability and maintainability, adapting to complex financial system needs. It supports distributed computing and cloud computing environments, suitable for modern financial enterprise business expansion.
Although the C86 instruction set can perform excellently in resource-constrained environments, it has certain limitations in scalability and maintainability. Facing future business growth and hardware upgrade demands, the C86 architecture requires more customized development and adjustments, increasing maintenance costs.
◉ leqiyufeng, System Engineer at ewell:
This question does not require analysis; C86 is more compatible with X86. C86 primarily selects Haiguang’s U, which is based on AMD core and instruction modifications, optimizations, and upgrades. Its instruction set has gained a higher performance advantage after subsequent self-research. Therefore, C86 has an absolute advantage in compatibility while also providing excellent execution efficiency. The ARM architecture can be understood as moving from mobile to desktop and data center, a latecomer with much poorer compatibility. Additionally, for the financial industry, software updates are slow, and there is rarely proactive adaptation to the latest container technologies (most current containers provide X86_86 and ARM images), so in terms of compatibility and migration, it is not as good as C86.
(6) What Parameters Should Be Focused on When Selecting Indigenous Innovation Servers as Virtualization Host Machines?
◉ Effort Mover, System Operation Engineer at a Joint-Stock Bank:
When selecting indigenous innovation servers as virtualization host machines, the following parameters should be focused on:
1. Processor
Core count and frequency: A higher core count and frequency can support more virtual machines running simultaneously, enhancing computing power. For example, the Kunpeng 920 processor can provide strong power for virtualization at 64 cores.
Architecture: Choose architectures that support hardware virtualization technologies, such as Intel VT-x and AMD-V technologies in x86 architecture, to enhance virtualization performance and efficiency.
Instruction set: A rich instruction set can optimize specific workloads, such as ARM’s NEON instruction set, which is beneficial for multimedia processing.
2. Memory
Capacity: Sufficient memory capacity ensures stable operation of virtual machines, generally recommended to be above 32GB, and for large virtualization environments, it may require 128GB or even more.
Frequency: Higher memory frequency can accelerate data read/write, such as DDR4 3200MHz performing better than DDR4 2666MHz.
Channel count: Multi-channel technology can increase memory bandwidth, improving memory access efficiency. Servers supporting quad-channel memory perform better than dual-channel.
3. Storage
Interface type: Common types include SATA, SAS, NVMe, etc. NVMe protocol M.2 interface solid-state drives have fast read/write speeds and are suitable for system disks or virtual machines with high I/O requirements.
Capacity and performance: Determine capacity based on the number of virtual machines and data volume, while paying attention to read/write speed and IOPS. For example, enterprise-grade solid-state drives can achieve sequential read/write speeds of over 3000MB/s and random read/write IOPS of hundreds of thousands.
RAID support: Support for various RAID levels, such as RAID1, RAID5, RAID10, etc., to ensure data security and storage performance.
4. Network
Network card type: Choose network cards supporting 10G Ethernet or higher speeds to meet the data transmission needs among virtual machines.
Network protocol: Network cards supporting network virtualization technologies like SR-IOV can improve network performance and virtual machine network isolation.
Network bandwidth: The total network bandwidth determines the server’s communication capacity with the outside world. Multi-network card binding technology can increase bandwidth and provide redundancy.
◉ nkj2021, System Architect at a Securities Company:
When selecting indigenous innovation servers as virtualization host machines, the following parameters should be focused on:
1. Computing performance: The more cores a processor has, the more tasks it can handle simultaneously. For virtualization environments, sufficient core count ensures multiple virtual machines run efficiently at the same time.
2. Memory performance: Memory capacity determines the number of programs the system can run simultaneously and the amount of data it can store. In virtualization environments, each virtual machine requires a certain amount of memory resources, so the host machine needs sufficient memory to support multiple virtual machines.
3. Storage performance: Common hard disk types include mechanical hard drives and solid-state drives. Mechanical hard drives have large capacities and relatively low prices, but slower read/write speeds; solid-state drives have fast read/write speeds but are relatively expensive. In selecting indigenous innovation virtualization servers, appropriate hard disk types and capacities can be chosen based on actual needs.
4. Reliability and stability: Using redundant power modules can continue to power the server when one power supply fails, avoiding server downtime due to power issues. By binding multiple network cards, network card redundancy can be achieved. When one network card fails, the system can automatically switch to other normal network cards, ensuring stable network connections. Using RAID technology can combine multiple hard drives into a logical disk, improving data reliability and storage performance.
Peer Exchange Consensus
Based on the above opinions, ARM and C86 are mainstream choices for the underlying server architecture of financial indigenous innovation. In specific usage, the valuable experiences summarized by experts are as follows:
1. In terms of compatibility and performance: C86 inherits the advantages of X86, and its ecosystem is more mature. In terms of performance, ARM has more advantages. In specific usage, applications can utilize the simple adaptation advantages of C86 for deployment, while database servers can leverage the performance advantages of ARM for deployment.
2. In terms of stability: Considering the current situation, the ARM architecture is more stable.
3. In terms of product parameters: Refer to the detailed explanations of CPU, memory, network, and storage provided in the previous discussions when engaging in product exchanges.
You can click at the end of this article“Read the Original”If you find this article useful, pleaseshare, like or click“♡”, so more peers can see it
Recommended articles/data:
-
Choosing the Technology Route for Indigenous Innovation Servers: Interpretation of 10 Core Issues
-
Discussing the Fault Rate Comparison Between Indigenous Innovation Servers and Existing Commercial Chip Servers
-
Practical Experience of Migrating from x86 to ARM Kunpeng Architecture【Exclusive Data Release for the Alliance】
-
Design and Implementation of Monitoring Solutions for Indigenous Innovation Servers, Middleware, and Databases
Welcome to follow the community “Indigenous Innovation” Technology Theme , which will continuously update quality materials and articles. Address:
https://www.talkwithtrend.com/Topic/139279
Download the TWT Community Client APP
Long press to identify the QR code to download
Or search for “twt” in the app store
Long press the QR code to follow the official account
*The content published by this official account only represents the author’s views and does not represent the community’s position.