Manufacturers of slightly different Cloud server name: ali Cloud called Cloud server ECS (Elastic Compute Service), tencent Cloud called Cloud server CVM (Cloud Virtual Machine), AWS called EC2 (Elastic Compute Cloud), Azure, called Virtual Machine,
Although each cloud vendors on cloud server has a different name, but the product shape is relatively consistent,
When purchasing cloud server, need to pay attention to the following: first of all, you want to evaluate their own websites, how to configure to shore up the site visits,
Secondly, examine the qualification of cloud server providers and service level, IDC industry access conditions are often harsh, requirements host must have a self-built housing, and have a certain scale, also has requirements for technical management and security control,
Finally, whether the price is within budget,
Cloud server configuration specification affect prices, but also directly determines the computing power and characteristics of it, and important considerations when purchasing, so, what is the difference between different kinds of cloud server? How do you choose efficient?
Choose cloud server configuration, see the three dimension
Cloud server configuration specification mainly depends on the type, the generation, instance size dimensions of the three most important
A: dimension type
Cloud server "type" or "series," means with the same design purpose or performance characteristics of cloud server categories,
Generally speaking, the cloud vendors will provide a general equilibrium model, computationally intensive, memory optimization, graphics computational common type of cloud server, such as the type corresponds to the hardware resources of a reasonable proportion or targeted to strengthen, convenient when facing different scenarios, you choose the most suitable models,
VCPU number and memory size (GB), the proportion of is decided and one of the important basis to distinguish the type of cloud server,
General equilibrium model is usually the proportion of 1:4, such as 2 nuclear 8 g, this is a classic collocation, can be used in the site, application service and other common load, such as website and enterprise application backend server, etc.,
If the vCPU and memory than was 1:2, 1:1, even it is computationally intensive, it can be used to make scientific calculations, video coding, code compilation and other computationally intensive load,
Ratio of 1:8 and above, and were classified as type memory optimization, such as the collocation of 8-core 64 g, it in the database, the cache service, big data analysis is common application scenarios, such as
Graphics computational is virtual machines with the ability of GPU, commonly used in machine learning and deep learning model training and reasoning, with the heat of the AI, this kind of machine are increasingly appear in all kinds of research and development and production environment,
In mainstream cloud computing platforms, often use acronym to express cloud server series, for example, is the common type for the AWS M series, ali cloud memory optimization model for the R series, the calculation of Azure optimization model for the F series,
dimension 2: generation don't
"Generation" of the cloud server (Generation), is used to identify what Generation that is under the series of models,
Data center hardware and virtualization technology is evolving, cloud vendors need to keep the latest technology and the ability to market, so even if is the same series of models, different generations don't, there will be a big difference between
Replacement of the same kind of cloud server, tend to bring the corresponding hardware CPU and upgrading to ascend, because the CPU is constantly updated, so the cloud server single-core performance is not necessarily the same, sometimes, although the two cloud server nuclear number is consistent, but due to the underlying chip architecture and frequency, may have a bigger difference on the performance,
A new generation of models, often corresponds to the new special underlying physical server and virtualization facilities, can provide a higher price performance ratio,
three dimensions: instance size
Cloud server instance Size (Size), refers to the Size of the hardware computing resources,
Under the selected machine type and generation don't, we can choose different instances of the size, in response to different load calculation,
In describing the instance, the industry often USES medium, large, xlarge word for naming, such a description of the basic has become the DE facto standard, including the AWS, ali cloud, tencent cloud, several mainstream manufacturers are using,
Can be roughly such memories: standard corresponding is equipped with 2 vCPU, large xlarge represents four vCPU, and higher configuration generally use nxlarge to express, where n 4 vCPU is multiplication relationship with xlarge representatives, for example, 8 xlarge means that this is a 8 * 4=32 vCPU machine,
If to be more rigorous expression of configuration, use the vCPU rather than the number of nucleus (Core) to describe the number of cloud server processor, because widespread hyper-threading technology (HyperThreading), often a Core to two vCPU virtual force, but also some processor does not support hyper-threading, so vCPU is a more appropriate way of expression, it is not easy to cause confusion and misunderstanding,
In some scenarios, you may see "metal" or "bare metal" such a description of the specifications of the word, is called the "bare metal" in Chinese, they are cloud services as much as possible to the physical model in cloud product way exposed as an example, it is mainly used for some ultimate performance, or need to run the software in the virtualization environment scene,
cloud server naming rules
Cloud server model name generally by type, the generation, size of the instance that the abbreviation of several, sometimes with supplementary suffixes, the most representative AWS naming rules (ali cloud use is also very similar format) :
When you understand the naming rules of cloud server after the next see a specific model, can quickly understand the meaning behind, obscure string immediately becomes clear,
For example, decomposition r5.4 xlarge this model, which is first and foremost a R type fifth generation memory type machine, it should have 4 x 4=16 vCPU, memory size is 16 x 8=128 g (memory type machine CPU memory than the average for 1:8),
Of course, not all cloud must be similar AWS naming rules, Microsoft Azure with a slightly different naming system, can be roughly summed up as:
"E4v3", for example, can represent Microsoft Azure on 4 of the third generation nuclear 32 gb memory type machine, mastered the Azure format features, you can also read quickly identify specific meaning,
In naming formula, there is a call "suffix" optional parts, it can be seen in many model named, it is as a general model of hardware information an important supplement, this model compared with the standard version without the suffix, there are some significant differences or characteristics, such as cloud, ali express the meaning of the "network enhanced" suffix is "ne",
how to verify the models match the expected configuration?
Lscpu commands available in the Linux environment, to understand the cloud server CPU information, and comparing the specific model name of the machine, below is in an AWS m5a. Xlarge machine running on the results, you can see the AMD chip providers and key information, such as dual-core four threads are in conformity with the implications of the model named:
