HUAWEI CLOUD Account Purchase: Comprehensive Evaluation of HUAWEI CLOUD Elastic Cloud Server ECS
In the circle of cloud servers (IaaS), many people buy servers and close their eyes to be "North Ali and South Tencent". When it comes to Huawei cloud, the first reaction is often: "oh, that big factory that does hardware and government enterprises, they also do public cloud for developers?"
As an old bird who has jumped back and forth between major cloud vendors and moved servers countless times, I am responsible to tell you:
If you still look at Huawei Cloud with the old eyes, you will probably miss the "performance monster" in the current public cloud market.
In the past two years, due to some well-known reasons, Huawei has played its life on the bottom chip and the full stack self-research. Today, we don't look at the official "far ahead" PPT, and directly pull it to the production environment to conduct an all-round and heart-to-heart hard-core actual combat evaluation of Huawei cloud elastic cloud server ECS(Elastic Cloud Server).
1. Chassis Performance Test: Is Self-developed "Optimus Prime Architecture" Bragging?
To buy ECS, the core is to look at the three-piece set:
CPU power, memory latency, cloud disk I/O
.
HUAWEI CLOUD ECS currently focuses on the 6th and 7th generation instances based on their self-developed "Optimus Prime Architecture. To put it bluntly, this architecture is to offload all the virtualization, network and storage losses that originally needed to consume server CPU to a customized soft and hard collaborative chip.
1. Calculation performance: unexpected "stability"
Let's take a general-purpose example of the same specification (4 cores 16G)
sysbench
Pressure measurement.
With multi-threaded and high-concurrency computing, HUAWEI CLOUD Generation 7 instances have very solid CPU running scores and extremely low volatility. This illustrates two points:
First, Huawei Cloud has strict control over overselling (Overcommit). The nuclear you buy is a real exclusive computing power and has not been "stolen" by neighbors ".
Second, its self-developed hardware acceleration does push the virtualization loss to close to zero.
2. Cloud disk I/O (extreme SSD): the real antidote to high concurrency
For databases and high-concurrency services, the IOPS (input and output per second) of cloud disks is the lifeline.
Ordinary cloud disk IOPS thousands of days, while Huawei cloud
Speed SSD
Under pressure measurement, even small pieces of random reading and writing can be firmly stuck on the officially declared high mark line. The most amazing thing is its delay, which is basically stable
0.5 to 1 ms
Between.
Old Bird vernacular: If your business often times out due to database disk I/O blockage, replace the cloud disk with Huawei Cloud's ultra-fast SSD, and the effect will be immediate.
2. network chassis: the "privilege" of the government and enterprise level sinking
Huawei started out as a communicator, so in the network, Huawei cloud ECS does have some "Gao Fushuai" genes.
Intranet bandwidth and latency: Compared with ECS instances of the same specification, the intranet benchmark bandwidth and burst bandwidth provided by HUAWEI CLOUD are both
Much more generous. If your back-end architecture has microservices, or there are frequent large files and picture synchronization in the intranet, the transmission speed between HUAWEI CLOUD ECS nodes will make you very climax. The measured intranet delay is about 0.1-0.2 milliseconds, which basically belongs to the physical LAN level.
Exclusive Dark Thunder: Overseas Network Dynamic BGP. If your business involves going to sea, HUAWEI CLOUD's overseas ECS (such as Hong Kong, China, Singapore) and its global BGP lines are extremely impressive. It is connected to the direct connection of the world's top operators. It is actually directly connected to Hong Kong ECS from China. The packet loss rate is stuck below 0.1 all the year round, and the evening peak is as stable as a dog.
3. actual combat experience: which places let a person cool? Which places want to let a person smash the keyboard?
As a "real person" evaluation, we definitely don't flatter blindly. Huawei Cloud ECS has its own advantages and advantages, as well as its advantages that make me want to spit out.
👍Let a person feel very cool flash points:
HCE(Huawei Cloud EulerOS) operating system: HUAWEI CLOUD now focuses on the self-developed Euler system. This system is actually optimized at the command level on HUAWEI CLOUD ECS. Similarly, running a Spring Boot or Nginx, the memory footprint and startup speed under HCE system are indeed better than the native CentOS or Ubuntu. The most important thing is that it is completely independent research and development, so there is no need to worry about stopping and safety wind control.
Second-level resource pull-up: Thanks to the hardware and software collaboration of Optimus Prime architecture, 50 ECS are purchased in bulk or flexibly expanded in the background. From clicking OK to initialization completion and SSH login, the whole process actually only takes about 10 seconds. This speed can save lives in case of sudden large traffic and emergency expansion.
👎people want to hit the keyboard slot point:
The console UI interaction is a bit "straight male government enterprise style": Huawei cloud's console menu logic is very strict, but it also means that it has a very deep level. For students who are accustomed to the flat and silly operation of large Internet factories, they may find some confusion when looking for "security group" and "flexible public network IP binding" for the first time. It is more like a rigorous industrial-grade system that requires a certain learning cost.
The ecological document is slightly hard-core: although the official help document of Huawei cloud is very complete, it is written in a very "engineer style". Sometimes I encounter a small error report, and the documents I found are all telling you the underlying principles and strict specifications. For impatient developers who want to "copy and paste a line of commands in 3 minutes to solve problems", it is not enough to be grounded.
4. selection life guide: how to buy the most cost-effective?
The naming specification of HUAWEI CLOUD ECS is like a string of Morse codes (such
c7.large. 2
), novices are easily picky. old bird teach you three step selection method:
Business Scenario
Recommended Specification Series
Suggestions on Selection of Pit Avoidance
Personal blog/test environment/small flow display station
General Entry (
t6/s6)
Commonly known as "sharing type". The price is so cheap that rounding equals no money. However, note that the CPU is based on a points system and cannot run at 100 full load for a long time, otherwise it will "speed limit".
Standard web backend/core business API/production cluster
Universal (g7)/Computational (c7)
The most recommended oil. Exclusive CPU power, intranet bandwidth full. The c7 processor has a higher dominant frequency and is suitable for the back end of writing a large number of complex logic algorithms.
Highly Concurrent MySQL/Redis/Big Data Processing
Memory type (m7)
The ratio of memory to CPU is 8:1 (such as 4 cores 32G). HUAWEI CLOUD provides hardware-level error correction (ECC) for high-memory instances to run the most stable core data nodes.
Summary: Who Is HUAWEI CLOUD ECS Suitable?
After a ordeal, Huawei Cloud ECS gave me the overall feeling that:
Low profile, solid, extremely high lower limit
.
It doesn't play virtual flower work very much, but in the core
Compute stability, network bandwidth, and cloud disk I/O
On, acting like a partial science overlord.
If you are doing government-enterprise projects, medium-and large-scale outsourcing, or financial, industrial and offshore core businesses that require extremely high stability, the underlying hardware strength of Huawei Cloud ECS and the Euler ecology of full stack self-research are definitely your best choice.
If you just want to toss a personal toy casually, you can choose whoever makes a big deal to make a bargain. Any tough battle involving the production environment, Huawei Cloud, the "self-developed giant", is definitely qualified to enter your core selection pool.
