Four performance test competitions of Alibaba Cloud, Tencent Cloud, and Huawei Cloud Server (with detailed data)

The recent opening of the World Cup has attracted attention to the strength of several cloud vendors behind the live broadcast business. Low latency, high sound quality, and high picture quality are the basic demands of all video owners for live broadcast services. This requires the public cloud to have strong cloud codec capabilities, as well as greater bandwidth resources and network quality.

As for the real situation? Our evaluation of the World Cup cloud-based services Ali cloud , Tencent cloud , Huawei cloud , to see how the real data.

Prerequisites for evaluation

Taking into account the issue of fairness, the participating cloud servers adopt the specifications closest to the real production environment, that is, the computing model 4 core 16G, the system CentOS 7.2, the common type, and the unified
cloud server based on the latest Skylake CPU. The overall strategy is to buy hardware with the same configuration and similar price, then execute the same commands, run the same programs, and compare the execution results.


1. Price comparison: The monthly price is similar, and the annual price advantage of Huawei Cloud is relatively more obvious.


2. Overall performance comparison:

1. Test conclusion:

a) Looking at the evaluation scores alone, Tencent Cloud leads, Huawei Cloud follows closely behind, and Alibaba Cloud ’s evaluation results are low;

b) Comprehensive price factors, Huawei Cloud has a higher cost performance, followed by Tencent Cloud .

2. Tool description: UnixBench, an old-fashioned system performance test tool, will be used to perform 11 individual tests, then compare the results with the benchmark system and output a value. The higher the value, the better the performance.

3. Note: This data is the overall performance data. Although it does not completely match the specific business requirements, you can have a general understanding of the quality of the cloud server.

UnixBench performance test results


3. Floating-point computing capabilities: For live broadcasts, video encoding and decoding take up a lot of computing power. Video owners are most concerned about single-core and multi-core floating-point performance, mainly floating-point operations. You should know that the CPU utilization during peak business hours often reaches more than 90%. Therefore, the computing power of the CPU and memory system is crucial to the live broadcast business.

1. Test conclusion:

a) There is not much difference between the three under single core, and Huawei Cloud is slightly better;

b) HUAWEI CLOUD takes the lead under multi-core . On the whole, in the comparison of floating-point computing capabilities, Huawei Cloud > Tencent Cloud > Alibaba Cloud .

2. Tool description: The multi-platform performance test tool Geekbench was selected, focusing on the investigation of the Sanduoyun CPU and memory system capabilities.

GeekBench identifies host information

Benchmark test results (average of 5 tests)


4. Memory bandwidth: Memory performance determines the performance of the system to a certain extent, especially for scenarios where a large amount of memory is accessed, such as relational databases.

1. Test conclusion: Huawei Cloud still performs well under this test.

2. Test tool: Use the comprehensive performance test assembly STREAM; it is written in two advanced and efficient languages, Fortran and C, which can give full play to the memory capabilities.

Note: The test here is the maximum continuous memory bandwidth, not the general theoretical maximum.

STREAM test results (average of 5 tests)


5. PPS test: PPS is the number of data packets that pass 4FE1 per second, which is suitable for scenarios with high network requirements such as live broadcast. The larger the single-instance PPS, the stronger the network performance, which often means that it can support a larger business volume.

1. Test conclusion: Huawei Cloud has obvious advantages, followed by Tencent Cloud .

2. Test tool Iperf: It can test the maximum TCP and UDP bandwidth performance, has a variety of parameters and UDP characteristics, can be adjusted as needed, and can report bandwidth, delay jitter, and packet loss.

After several rounds of testing, overall, Huawei Cloud still has bright spots in terms of price and performance. Of course, this evaluation focuses on the live broadcast industry, so it focuses on the relevant performance in the live broadcast scene.


Finally, let’s talk about the issue of “ Huawei Cloud defeated Alibaba Cloud and Tencent Cloud to bid for the server for China Mobile’s Migu Video World Cup broadcast. Migu was stuck during the World Cup live broadcast”.

Factors that affect the fluency of live broadcast: network speed, memory, graphics card, operator line, platform bandwidth, platform settings for the live broadcast room, viewer's network speed, memory, and graphics card.

Internet live broadcast sources are generally satellite live broadcast signals (CCTV5 this time), which are then transcoded in real time before being transmitted to the client's terminal equipment for playback. Approximate process: from CCTV5 live signal (HLS mode) → CDN distribution → operator network transmission (fixed network broadband or 4G) → user terminal decoding and playback

The above 4 links may cause image stuttering. Excluding the live signal, network transmission and Migu APP's own problems (recovered the next day), the more likely problem lies in the CDN. As we all know, CDN is a content distribution service, which determines the fluency of the network to a large extent. In the China Mobile Migu bidding some time ago, the CDN that Huawei Cloud won was not:

HUAWEI CLOUD Won the Procurement of Migu Public Cloud


Above, welcome to the discussion of the technology gods.

Guess you like

Origin blog.csdn.net/NicolasLearner/article/details/112790078