Awareness of the need to implement cloud services is usually followed by the choice of a cloud provider. How to choose a reliable supplier from all the diversity of the market? What are the criteria for checking a candidate?

Unfortunately, within the framework of one article it is hardly possible to tell about all the pitfalls and nuances, but we can guarantee – after reading the article you will know:

• what to look for when choosing a supplier,
• what comparison criteria to use,
• is it worth asking for official documents and so on.

We have summarized our expertise and the experience of our customers to collect all the important information about choosing a reliable cloud platform. Here we go!

Used hardware

Let’s start with the most basic. How do you imagine a data center? Cleanliness, high level of discipline, access by passes and a lot of equipment. On this it is worth paying attention in the first place. When choosing a service provider, take the time and check the hardware on the check list:

Models of equipment. Outdated hardware is a guarantee of regular crashes and reduced performance. A reliable service provider regularly upgrades the hardware complex and will not allow the appearance of old bare-metal.

Equipment class. Vendors, trying to occupy a favorable position in the market, produce various lines of equipment – from corporate (enterprise) solutions to medium (midrange) and cheap (low-end). It is logical that you should focus on the provider with enterprise solutions.

Manufacturers. Using the solutions of manufacturers from the budget segment increases the likelihood of failures and equipment outages.

Duplication, redundancy, lack of a single point of failure. This can be achieved by:

• duplication of power supplies and cooling servers;
• duplication of storage or the presence of two independent controllers;
• presence of the necessary supply of server resources in case of a scale in consumption;
• duplication of all network equipment.

If we are talking about a reliable platform with high availability, no failed element should affect the performance of the entire complex.

Disk subsystem performance

When renting a virtual infrastructure at a basic level, you need to determine the needs for the following resources:

• CPU clock speed (in GHz);
• RAM amount (in Gb);
• amount of disk space required (in Gb).

However, it should be noted that the operation of a service or application in the cloud is affected not only by the allocated amount of disk space, but also by the performance of the disk subsystem. As a rule, the second factor plays a more important role. Even if you ask the provider for a specific type of drive, you won’t get a 100% performance guarantee, because it is also affected by the load from the “neighbors” in the cloud.

Objectively to evaluate performance you can using two indicators:

• IOPs – number of input-output operations;
• Latency – level of delays in use.

It is imperative that the provider is ready to openly discuss the disk arrays used and be ready to provide you with the appropriate IOPs management capabilities. However, if the service provider guarantees these metrics for you, there is no need to go deep into the details.

Platform Reliability

When choosing an IaaS provider, be sure to prove the reliability of the site on which the cloud platform is located. Many providers conduct short excursions or show you the site upon request – a good provider has nothing to hide.

However, this is not always enough. As a rule, any client would like his applications to be located in a data center, the level of which is confirmed by specialized companies specializing in this. To avoid confusion among the many proposed documents, pay attention to those issued by the Uptime Institute:

• Tier Certification of Constructed Data Center Facility;
• Tier Certification of Data Center Design Documents;
• Tier Certification of Data Center Operational Sustainability.

Detailed information on certificates can be found on the official website of Uptime Institute.

Support service

One of the pillars of the site’s high availability is well-developed processes for supporting virtual infrastructure. The list of these processes includes:

• incident management;
• processing service requests;
• change management;
• optionally – knowledge of managing the support staff.

The incident processing process should be as formalized as possible, which can be achieved using a single system for recording user requests and messages from the monitoring system. This approach allows you to work with incidents as they occur, take into account every single notification and message and, as a result, reduce the affect on systems and their recovery time.

One of the attributes of a well-structured change management process is a single regulation for making changes to the infrastructure. It includes:

• careful planning of any changes;
• testing;
• plan to roll back the changes made in case something goes wrong;
• minimizing the risks of violating SLA.

As a rule, a client who has a request or has a problem first of all “knocks” precisely to technical support employees. However, it’s not enough just to put engineers on the phone — corporate-level services must necessarily be accompanied by a high level of service. To provide it, the support service must meet certain criteria:

• 24/7/365 service;
• several communication channels: phone, email, access system on the site
• compliance with the declared service parameters:
• reaction time;
• the maximum term for solving a typical request;
• maximum time to resolve the incident.
• mandatory and timely reporting of work and the elimination of incidents;
• assistance in solving tasks that are beyond the scope of the service.

Company profile

In the worldwide market, cloud services are provided not only by providers specializing in this, but also by companies of other profiles. Who can I contact to rent a cloud and other services?

Cloud providers. Companies of this profile specialize in providing cloud services to customers – they form the core of their portfolio of services. A reliable service provider hosts its own equipment in selected data centers.
Data centers. The growth dynamics of the domestic cloud market motivates more and more data centers to expand the list of their services by renting virtual infrastructure and other services. Some provide these services on their own, others do it through IaaS providers, concluding partnership agreements.
System integrators. It is unlikely that there will be at least one large system integrator without its own cloud. However, a specific feature of such companies is their versatility and the provision of a wide range of services, so there is a considerable chance that the level of cloud services will suffer due to a lack of attention and technical expertise.
Telecom providers. Telecom providers, like data centers, are interested in expanding the portfolio of services through cloud services. And their participation in the development of the cloud market is more than justified: in order to effectively interact with resources in the cloud, you need a good communication channel.
As you can guess, the sophistication of services from a technical and organizational point of view often depends on the profile of the company. It is logical that by the level of service those companies for which virtual infrastructure rental play the role of the main activity — cloud providers and data centers — will stand out.

Hypervisor

Customers are not always interested in the hypervisor used by the provider, however, for certain reasons, we recommend that you pay attention to this:

• hypervisors may differ in functionality, which directly affects the performance of virtual machines and the level of availability of services provided;
• different hypervisors can support different OS;
• the level of security in terms of information security and vulnerabilities may also differ.

The main hypervisors now are VMware ESXi, Hyper-V from Microsoft, Citrix Xen and KVM from Red Hat. However, today the undeniable leader of the virtualization market is VMware, which in fact created it from scratch. One of its main advantages is a high level of security. The VMware hypervisor is “sharpened” for virtualization tasks and cannot perform any other tasks, which minimizes the attack area.

If virtualization is already used in your company, choose a provider with the same hypervisor that you are already using.

SLA

Before concluding a contract for renting a virtual infrastructure, it is necessary not only to conduct testing whenever possible – as a rule, providers provide resources for these purposes for free – but also familiarize themselves with SLA, an agreement on the level of services provided. What criteria can be used to check SLA:

• all typical requests must correspond to the solution time parameters recorded in the SLA; for atypical requests, a fixed period of time must be indicated, during which the decision time must be determined.
• for violation of SLA, the service provider must be held financially responsible.

Remember that the mere existence of a Service Level Agreement does not guarantee the effectiveness of the service level management process. The service provider should regularly collect SLA metrics, analyze them and make adjustments.