What’s better, the cloud or servers? Comparing IT infrastructure platforms

Clouds and servers are often compared when talking about IT infrastructure solutions. However, not many sources address the technological differences between these solutions transparently and in detail. In this article, we will determine which platform is the best choice based on their technical capabilities, as well as compare the economic aspects of renting either solution.

Cloud and server comparison criteria

Before comparing the IT infrastructure platforms, it is worth determining the comparison criteria and the definitions of certain terms related to the solutions.

Firstly, we compare clouds and servers in the context of IT infrastructure provider services. The platform where the solution is hosted directly affects the technological capabilities of the entire infrastructure; local data centers and professional data centers rarely offer comparable conditions. For the purposes of this article, we assume that the cloud and the server are hosted in a provider's data center. In addition, it is difficult to determine the average quality of local solutions; providers' offers more reliably demonstrate the real capabilities of the platforms.

Secondly, remote solutions in general are often mistakenly referred to as «clouds». In fact, a cloud is a computing platform based on multiple physical servers which uses the virtualization technology to create virtual machines with flexible configurations. Remote access is not a feature unique to clouds; when you rent a server in a provider’s data center, you access it remotely as well. In this article, we will compare the technological capabilities of these options, as well as the economic aspects of working with each one.

Category 1. Fault tolerance

Virtual machines in the cloud use parts of the resources of several physical servers at the same time. In the event that a component or an entire server fails, the VM will automatically use other capacities to complete its tasks. In addition to this key functionality, a well-designed cloud uses component redundancy on the architectural level. For example, the SIM-Networks cloud uses 2N + 1 redundancy: each component has a parallel copy and another one in reserve. This way, a given element of the system will fail only if all 3 devices fail.

Some clouds also offer a way to protect the infrastructure from large-scale failures, such as damage to the power source due to a natural disaster. Data can be distributed across multiple availability zones, completely independent segments of the cloud. These zones are part of the same infrastructure, but each of them uses its own own computing resources, power supply & cooling systems, etc. A large-scale failure in one availability zone will not affect the operation of the cloud in another zone.

Among traditional server solutions, the cluster is the most fault-tolerant one. A cluster is a system of several interconnected servers under centralized control. Same as the cloud, a cluster distributes its tasks in the event of a hardware failure. However, the workload will be moved from one physical server to another - this system cannot distribute a single process between several machines at the same time.

As a rule, there are fewer physical machines in a cluster than in a cloud. Unlike a public cloud, a server cluster is an individual solution that is used by a single client. The main job of the cluster is to keep the services running in the event that one or several machines fail, so that a system administrator can quickly fix them. In the case of the public cloud, which distributes resources among many users, fault tolerance is also achieved through scale: there are always enough servers for the user to continue working.

Clusters are often used for large databases – they require the performance of several dedicated servers and high fault tolerance, but do not need the flexibility of cost & configuration that the cloud offers. This way, the level of fault tolerance is higher in the cloud; however, a cluster may be a more appropriate solution depending on the task.

Category 2. Additional features

Cloud infrastructure is not only a pool of computing resources, but also a set of tools that are responsible for various aspects of working with IT services. Let's look into these features using OpenStack – a platform for deploying clouds – as an example. OpenStack consists of multiple modules, and each is responsible for a specific function. Here is what some of the core modules do:

• Neutron allows you to configure the cloud network infrastructure, set up VPNs and firewalls, create virtual networks, etc.;
• Glance stores virtual disk images which are used to install software on virtual machines and structures them into a library;
• Keystone distributes data and resource access rights between users through authentication;
• Swift stores data as unstructured objects and supports distributed data storage;
• Horizon provides the user with a graphical interface for managing cloud resources and other modules.

These features are part of the basic functionality of an OpenStack cloud, and each of them is included by default. Additionally, some providers offer a cloud backup service – BaaS, Backup-as-a-Service. To create a backup, all you need to do is specify the frequency and depth of the backup, as well as select the location of the backup storage. For example, SIM-Networks offers two options - BaaS Local and BaaS Remote. The first solution stores backups in the same data center as the main infrastructure; the second uploads backups to a remote data center which improves security.

Dedicated servers are also referred to as a "bare-metal servers" – as a rule, server solutions do not offer the same level of automation as the cloud. Providers offer a range of additional services for dedicated servers, but many of them are not included in the rental price. For example, if the vendor doesn't offer a backup service, the client will have to set up the system themselves or look for a third-party tool. Compared to the cloud, such a system is difficult to manage: in cloud infrastructure, additional service management is carried out through a single control panel. This makes the cloud the best choice if you need a set of ready-made tools in addition to computing power.

Category 3. Configuration flexibility and scaling

Flexible configurations are one of the main advantages of the cloud. Resources in this kind of infrastructure are represented as a pool, meaning an unordered set. Users create virtual machines from this set of computing resources. Each VM has the same characteristics a physical server has – CPU cores, RAM, data storage, and so on. To create or reconfigure a VM, a user specifies the necessary parameters in the control panel – the process will take no longer than a couple of minutes. This means that a user can create an instance for a particular task at any time and immediately start working.

The pool of resources in the cloud is made up of the capacities of all physical machines. This way, the upper limit of scaling in the cloud is equal to the amount of unused resources in the entire infrastructure. This kind of pool is not easily exhausted; public clouds are built to provide scaling capabilities to many users at the same time.

The configuration of a physical server corresponds to its components, which means that it can only be scaled by replacing components. Using a server as the base, you can create a VPS – a virtual server similar to a VM in the cloud. However, even if you combine several servers into a cluster, a virtual server will only be able to use the resources of its host. The configuration of the VPS cannot be changed – you will have to deploy a new virtual server with the necessary parameters. If you want to learn more about the difference between VPS and virtual machines in the cloud, read this article: «What is the difference between the cloud and VPS?».

Clustering is one of the ways to scale your server infrastructure. This method is less flexible than cloud scaling. A user will have to order a new server, set it up and connect it to the cluster, which takes time. Besides, the company in question may not need an entire dedicated server to handle new workloads; if the task is temporary, the new server may end up being redundant afterwards. When using the cloud, virtual machines can be created or deleted at any moment. If flexibility is a priority for your business, cloud solutions are the better choice.

Category 4. The economics of the solutions

Public clouds are paid according to the «pay-as-you-go» model, which means that the client pays only for the resources that they’re using at the moment. If the user turns the cloud resources on or off, the price increases or decreases instantly. This model is a consequence of flexible scaling in the cloud; it is impossible to foresee which configurations the client will use, for example, within the next month.

This advantage has a downside: the public cloud may end up being economically unviable for a company. In the early stages of business development, flexibility is needed in many aspects of work. However, as the company grows, it is important to determine which tasks would benefit from a more permanent solution - for example, a dedicated server or a private cloud with a fixed cost. At the same time, a public cloud can be used for temporary workloads and combined into a single system with a private cloud – this type of infrastructure is called a «hybrid cloud».

The cost of renting a dedicated server is fixed and does not depend on actual resource usage. As a rule, this is unsustainable for companies in their early stages of development: the cost of renting servers may exceed the actual need for computing power. In addition, server infrastructure is inefficient when working with seasonal loads. For example, if a company experiences an influx of customers in spring and autumn, deploying a virtual machine in the cloud is considerably easier and cheaper than ordering an entire dedicated server with a fixed configuration.

The server is a cost-effective tool for permanent tasks with stable workloads. If the server configuration meets the requirements of the services, this solution may be more cost effective than renting an equivalent pool of resources in the cloud. However, if a company cannot predict the workloads, or if the task does not require the capacities of an entire physical server, the public cloud is a good choice.

Category 5. Security and protection measures

The issue of security is often raised in the context of discussions on cloud computing. It is believed that the public cloud is an unsafe infrastructure because the data of several users is hosted on the same machines. The assumption is that hackers can easily gain access to someone else's infrastructure; however, that is not the case. Cloud resources are divided into virtual machines using a hypervisor, a technology that isolates VMs from each other as if they were two separate dedicated servers. In addition, there are also physically isolated cloud platforms, private clouds.

Cloud architecture also contributes to data protection. The cloud infrastructure supports distributed data storage, which significantly reduces the risk of information loss or theft. Critical services are not stored on a single server, which makes them more difficult to access. Reliable clouds also use hardware-based encryption and software-based measures such as authentication services, firewalls, DDoS protection, and so on.

A dedicated server is a completely isolated platform that is used by one client only. However, the basic level of protection in a server is limited by physical isolation and the capabilities of the data center; other security measures are the client’s responsibility. It is also worth considering that hackers often use vulnerabilities in outdated software to gain access to data. Updating basic software in the cloud is the responsibility of the provider; in the case of a dedicated server, this is done by the user.

Setting up a secure infrastructure based on dedicated servers is a solvable task; however, the cloud offers more protection by default. This allows you to save money and your technical staff’s time both at the stage of building the infrastructure and in the long term – the provider is fully responsible for the security measures they offer. In conclusion, the cloud is better protected than a dedicated server by default and requires less investment in security.

Which one do I choose, a server or the cloud?

Each of these solutions has its pros and cons; however, the cloud is the winner of most categories considered in the article. By default, a server only provides computing power; the cloud is an ecosystem that consists of many physical machines, programs, additional services, etc. A dedicated server needs additional software to provide the functionality that is accessible in the cloud the moment you order it.

Bear in mind that the functionality of the public cloud is aimed at covering the needs of as many users as possible. A bare-metal server can become the basis for a larger system built to your individual specifications, and you don't have to do it on your own. We, SIM-Networks, build IT infrastructures based on any platform - public clouds, private clouds, dedicated servers and other solutions. If you want to learn more, contact the Customer Care team – simply describe the tasks your business faces, and we will offer a solution for your needs.

Did you like the article?