Clouds from an unknown country. Cloud FAQ
A light cloud floated right above them.
- Listen, let's go to Tili-mile-trodydy! - suggested Little Bear. “We can speak theirs.” Look, what a good word: “Tryam”!
- To you? A very good word, said the Hedgehog. “What does it mean?”
- Tryam - in tili-mile-tryam means "hello!"
Cloud technicians use numerous terms that describe the features of virtualized solutions. These terms entered the vocabulary gradually and to describe technical or marketing specifics. In this case, their meaning was initially described, and then they were given names. A decade after the beginning of the rapid development of “clouds,” professional slang makes it difficult to understand the essence of cloud products for businesses and people who previously did not know or perceived cloud technologies as a news background without delving into the content.
In the article, we will try to act as a translator from the marketing and technical language of the cloud provider to a more understandable one, restricting the use of words located on the periphery of the vocabulary of the language and not having the quality of communicative general validity.
Why is cloud computing?
In 1993, Eric Schmidt came to the conclusion that the network and the computer are related, this idea became widely known later from the Sun Microsystems motto “Network is a computer”. “When a network becomes as fast as a processor, the computer as such ceases to exist, it will spread across the network,” Schmidt said about 25 years ago. In August 2006, Schmidt was the first to use the term cloud computing in an interview after the Search Engine Strategies Conference:
“We are witnessing the birth of a new model of computer systems, and it seems to me that there are not many people who are able to understand the emerging perspective. Its essence is that services that support data and architecture are hosted on remote servers. The data is on these servers, the necessary calculations are performed on them ... And if you have the appropriate browser and the appropriate access rights, then you can access this cloud regardless of the device used. ”
Back in 1959, science fiction writer Kurt Vonnegut mentions a cloud that “thinks about everything and for everyone together” in the book “Sirens of the Titan”.
There is also an opinion that the Internet in the schemes of computer networks and systems was often indicated by the image of a cloud, and so the name became popular for the use of computers via the Internet. But we believe that it is Eric Schmidt’s comment that can be taken as the beginning of the use of the term “cloud” in the narrow sense in which it is used in IT, since it is the remoteness of the servers that the user accesses via the Internet that characterizes the cloud consumption model.
We give our definition:
A cloud is a pool of resources from which a user, interacting via the Internet, can get the volume and composition of IT services he needs and, if necessary, independently change these consumption characteristics.
Everything as a Service or what you can get from the cloud
Much depends on the responsibility and rights that you want to keep when using the cloud. The initial level is the program-as-a-service (saas - software as a service) model. With this business model, the provider himself creates and configures the infrastructure, giving the client access to the finished program / application, for example, corporate mail or antivirus. This type of cloud service is well-known even to people far from IT due to cloud solutions for storing photos and documents. However, it should be understood that this is not the only application of the model and the functionality and purpose of the program in the cloud can be any.
For customers who want to independently configure “cloud servers”, install the necessary software on them and, importantly, connect them over the network, the services are provided according to the Infrastructure-as-a-Service model (IaaS - Infrastructure as a Service).
Since the cloud is only a consumption model in which you can get any IT resource with access to it via the Internet, marketers have come up with an infinite number of abbreviations like Everything as a Service (XaaS) corresponding to each type of service.
How it works?
The network has become fast enough to make full-fledged interaction with a computer possible via the Internet. This has created prospects for a new business. So there were cloud providers. To provide their services, they use physical servers, the resources of which are combined into a pool, and then, using a program (hypervisor), it is "cut" into virtual machines (cloud servers, vds / vps). This process of emulating software-defined or virtual machines is at the core of virtualization technology. Cloud4Y uses the virtualization platform from VMware and we will describe the operation of the cloud using our example.
Below, in the first two photos are examples of servers that we use. These are the HP ProLiant BL460c Gen8 Blade Servers. Each blade (blade) is combined using a chassis, which is responsible for power supply, network connection and other processes.
The tenant of the provider’s infrastructure can create virtual machines with the characteristics they need (CPU, RAM, etc.) and quickly change them. Of course, for this you do not need to select the appropriate physical server each time and “move” to it. Physical resources of data center servers, such as computing power, disks and networks, are combined into large pools of virtual resources. In the future, parts of these resources are provided in the form of “prefabricated” virtual data centers (data centers) that are allocated to tenants. This provides the flexibility and scalability that are the benefits of cloud infrastructure.
To support a large number of virtual machines on one physical server, more memory, more connections for data warehousing and more network connections are required, so we chose HP servers certified for VMware and built with virtualization in mind. VMware certification enables stable use by all cluster options of the VMware virtualization platform , which significantly increase the efficiency and reliability of the entire cloud:
What is a VMware cluster and how is it organized?
In the beginning, we will designate that in the framework of this article we will mean a cluster as a group of hosts (physical servers) under the control of a single service for joint performance of certain functions as an integrated system that communicates through a network.
On the VMware vSphere virtualization platform, you can build 2 types of clusters: High-availability cluster (HA) and Distributed Resource Scheduler cluster (DRS).
A HA cluster will mean that a certain number of physical servers are clustered and virtual machines are launched on them. In the event of failure of one of the hosts, virtual machines are launched on other servers from the group on which a place was previously allocated for this. As a result, the downtime is equal to the boot time of the virtualka operating system.
If you want to reduce downtime to a minimum time, we recommend using VMware Fault Tolerance technology. The main idea of the option can be described as creating a synchronously working replica of a virtual machine on another server and instantly switching to it when the main host fails.
VMware DRS technology is used to balance the load in the cluster. To do this, at the initial stage, cluster resources are combined into a pool and then load balancing between hosts occurs by moving virtual machines. DRS can recommend a move with the necessary confirmation from the administrator or do it automatically. This happens using the vMotion “live migration” utility, due to which migration does not require VM stopping. Users continue to work with one instance of the VM until the data is transferred to another host. At the last moment, the latest changes are copied from the RAM, the user sees a slight short-term decrease in system performance and, after a moment, is already working with the same VM, which in fact is already on another physical server.
How VMware HA + DRS
In the case of a VMware cluster, a group of 2 or more ESXi servers is centrally managed by VMware vCenter Server. Actually, you can create virtual machines on the same host with the VMware ESXi hypervisor installed, but there will be no HA, DRS and others. You can simply “slice” your physical server into several virtual ones, and its inoperability will mean the downtime of all VMs.
To use all the cluster capabilities, you must use the VMware vSphere platform, which includes the ESXi host management server and storage, the so-called vCenter Server. Also, to build a cluster, you need to connect a storage system. In it, in a special VMFS cluster file system, sections with virtual machine files are stored, which are accessible for reading and writing to all ESXi hosts in the cluster. Due to the storage in one place and the independence of the virtual machine from the physical platform, quick movement and recovery is achieved with HA, DRS, FT, vMotion.
VMware vSphere Platform
VMware vCenter Server, in simple terms, is a set of services and a database. Each of the services deals with its specific task list and interacts with other ESXi services and / or hosts. vCenter Server is a command point to which ESXi hypervisors on hosts are subordinate. Communication between them occurs through host agents of VPXA. You can do even more from the vCenter Server control panel than by connecting directly to ESXi. If you can create / delete virtual machines in ESXi, then using vCenter Server you can additionally create and configure a cluster for them and all the necessary cluster options, some of which are described above. VMware vCenter Server can work both on a separate physical server, and inside a virtual machine on the same host, which it itself manages.
The topic is certainly interesting and extensive, but the deployment of such infrastructures requires large material costs. If you want to use all the features that increase the fault tolerance and reliability of the system, you need to purchase at least two servers and storage, as well as buy a license for the VMware vSphere platform from one of the distributors. Installing, configuring, and administering a VMware cluster will also require time and money.
In cases where the IT infrastructure requires the high reliability that the VMware vSphere platform provides, but there is no possibility or sense to make significant capital investments, many corporate clients choose the infrastructure rental service (IaaS).
Clients do not use VMware vCenter Server. The provider is responsible for managing the clusters and physical equipment. Customers get a significant amount of control over their virtual data center using the convenient VMware vCloud Director self-service portal , which we wrote a separate article about . Creating a data center for the client occurs as soon as possible, while the required number of virtual machines with the necessary characteristics and operating systems, routed and isolated networks with any topology can be created, flexible Firewall rules and much more are configured.
Can I build my own cloud? The main differences between public, hybrid and private clouds?
The public cloud is a business model of providing services in which all the physical resources of the data center, such as computing power, disks and networks, are combined into large pools of virtual resources, in the future part of these resources are provided in the form of “prefabricated” data centers, which are allocated to tenants. The virtual machines of different tenants are isolated from each other, so you should not understand the publicity of such a cloud as the lack of data security. The publicity of the cloud means that the data can physically be stored on the same physical server with the data of other companies, but they do not have access to your data.It is impossible to say on which physical equipment your virtual machines will be located, because when stored in a cluster, virtual machines move between servers to balance the load and increase fault tolerance. It is the allocation of “private” resources from the common “public” pool that makes the cloud public, but user data is protected by modern organizational and technical and software tools and solutions.
Private or private cloud- this is the opposite of a public cloud, which means that a pool of physical resources (physical server) will be provided to only one tenant / organization. Moreover, where exactly the equipment of a private cloud is located does not matter. The cloud will be considered private not only if the equipment is located on the territory of the company, but also options are possible when the equipment is located in the data center and is adjacent to the equipment of other customers. A private cloud can also be provided by a cloud provider. For example, Cloud4Y has a Private Cloud 2.0 solution .
Private Cloud Pros:
- high level of security;
- complete isolation of infrastructure;
- equipment control;
- the possibility of placing on the customer’s site.
Summing up, it can be noted that in both cases the model is cloudy, which means that there is no binding of equipment to the place, all resources are available via the Internet or the network where necessary. However, in a private cloud, the customer receives completely isolated equipment, and in the public cloud only his virtual resources are isolated. Often when equipment is placed in the data center, customers who are especially sensitive to the security of their data or fulfill the requirements of certain standards order server racks with a perimeter enclosing structure, the so-called “cage”. In this case, a private cloud is perhaps the only solution, but even without a separate “cage”, the equipment in the modern TIER III data center is safe and under round-the-clock surveillance, including video recording.
Hybrid cloud is a model of consumption of IT resources, in which the organization part of the system is located in a public “cloud”, based on the equipment of a cloud provider, and part is in a private “cloud”, on servers owned by the company itself or leased as a whole. The concept of a hybrid "cloud" (hybrid cloud) allows you to combine in a single cloud space the internal (onsite) corporate cloud and the external (offsite) cloud service provider. The main idea of the “hybrid cloud” is that when there is not enough internal capacity, you can use external resources. Hybrid clouds are used as an opportunity to go beyond the company's cloud capacities in the event of peak loads or in order to abandon capital costs in favor of operating expenses.
The cloud is not a luxury
A VMware cloud has three types of models that allocate resources.
- ALLOCATION POOL -% of the resource is guaranteed, and the maximum possible limit is set in the resource pool.
- PAY-AS-YOU-GO - there are no guaranteed resources and maximum limits set in the reservation pool. Resource limits are set at the virtual machine level.
- RESERVATION POOL - guaranteed resources and maximum limits are equal, all resources are allocated. No resource parameters are set at the virtual machine level, however, the user can change the restrictions and reserve resources on the virtual machine.
A client who needs a fixed set of resources can work with guaranteed resources, or choose PAY-AS-YOU-GO when there is no data on how much resources they will consume in the cloud. Due to the elasticity of the pool, the provider can avoid the redundancy of physical data centers and reduce capital costs by adding physical hosts only as needed without stopping work. For a client, payment may occur upon consumption at the end of the month, which also allows you to abandon capital costs in favor of operating expenses.
An agreement with a provider (SLA) should provide for the level of availability of services of a certain quality.In the case of Cloud4Y, this figure is 99.982%. In addition, we set the minimum acceptable performance indicators for the CPU and RAM of the system. The number of MIPS per vCPU is at least 2,900, which guarantees customers the stated processor speed. Also, "oversubscribing" physical RAM is not allowed, RAM Swaped is 0%. This means that the Configured Virtual RAM allocated during creation of the virtual machine, which the guest OS will see, is 100% allocated physical memory that is available to the virtual machine at any given time. In practice, this approach avoids reducing the performance of cloud servers in cases where the operating system accesses blocks in RAM and expects a quick response,
This creates the conditions under which cloud servers for performance at any time can fully replace a physical server with the corresponding characteristics for clients, and due to the reliable equipment in the network of TIER III data centers, virtualization in the cluster and round-the-clock technical support, the quality of services and fault tolerance are ensured.