Private cloud: "blue tape" for harsh enterprise

    It seems like clouds are great. Rent yourself virtual resources on a virtual server, spread over the data centers of the entire planet. And do not worry about anything other than topping up your account - all support, etc. automagically performed by itself. But one fine day, Hetzner, or even Amazon, suddenly apologizes for the completely force majeurely dropped server and irretrievably lost data. As a compensation for the downtime, you will definitely be transferred $ 10, according to clause 122.3 (d). Weak support against contractors and customers gathered around the central office with torches, cans and a strange gleam in their eyes.
    At such moments, you want to buy a ticket somewhere in Ecuador or Puerto Rico. And the look falls on a cozy dusty server in the back of the office. What to do?

    The question "buy your own or rent someone else's" has been relevant for thousands of years. Ever since the times of ancient Egypt, which outsourced the ancient Greek hoplites to the country's defense. Which successfully changed their employer and subsequently captured the same Egypt along with the Persian ruler Cambyses II. History has not had the desired effect, and people continue to step on this rake over and over again.

    Greek hoplites defend Granik. Roughly 334 BC
    When computing power, data storage, managed systems like Ansible and the like are clearly not what your head should hurt, the cloud immediately looks interesting. Let's say you have your own plumbing stall. And I really want a site so that fittings, couplings and locknuts 3/4 look beautiful against a purple background. And the server in the pantry is not happy at all - the mixers are already there. In this situation, everything is simple and logical - everyone should do what he understands. Otherwise, we would hardly have left the subsistence economy.
    Lack of expensive IT staff. Payment of resources as needed, in the form of OPEX. Infrastructure flexibility - you can add a couple of terabytes to the storage at any time or use a dozen more virtual machines. These advantages actively contribute to the development of the cloud market, the emergence of all these SaaS, IaaS and all sorts of PaaS. And the accounting has already been transferred to the cloud. It seems that soon even the assistant secretary will be on lease and with per-minute billing.

    Responsibility and money

    Not everyone knows how to fry the right steaks. The choice of a beginner looking at glossy packages with pieces of beef is painful. It is doubly disappointing to spend a lot of money to get a rubber sole due to the wrong piece or errors in roasting. Triple if everything was conceived as an ideal romantic dinner. In such cases, you think that it was worth shifting responsibility to invisible professionals who work somewhere in the bowels of a red-hot kitchen, and immediately enjoy the excellent result. However, nothing is given for nothing, and the work of such specialists is paid accordingly.

    From this perspective, you can consider the cloud service models - as the costs of a “romantic dinner” increase:

    1. IaaS (Infrastructure as a Service) - infrastructure as a service. No need to worry about choosing meat. You have already picked up a good piece of thin edge, packed, attached a set of spices, butter and a sprig of rosemary. It remains to come home, ideally fry the meat, set the table, arrange glasses and pour champagne.
    2. PaaS (Platform as a Service) - platform as a service. The courier will deliver perfectly fried steak-ribeye in a thermally insulated container within a few minutes. It remains only to arrange the cutlery.
    3. SaaS (Software as a Service) - software as a service. You and your soul mate go to a restaurant and enjoy good music, excellent wine and tender meats.

    Private clouds

    Transferring responsibility is great, but many technology companies already have at least a few “steaks” in the refrigerator. Iron steaks, the purchase of which was accompanied by excruciating capital expenditures. The desire to maximize the use of this iron resource led to the development of virtualization technologies - a private cloud model.

    Let's try to figure out what is a private cloud. Perhaps it is enough to publicly declare that “we do not use public clouds, since our private cloud is quite enough for us”, implying the existing IT infrastructure and all the software that has been used on it for the past 15 years?

    The US National Institute of Standards and Technology (NIST) defines the cloud with five key characteristics:

    • self-service on demand
    • free network access
    • pooling resources
    • quick elasticity and / or expandability
    • measured service

    If all this was achieved in the process of IT evolution, then the company can safely say that it has its own “private cloud”. Reality is often fundamentally different from such a utopian picture. Cloud technologies are extremely interesting for business, as they provide full transparency of costs, provide a quick time-to-market and are inexpensive due to their efficient use. On the other hand, security guards are worried about these companies and want to see them on their own site behind seven steel doors.

    In our case, three key components were missing for the implementation of the private cloud:

    • self-service portal - for on-demand service
    • measurement and billing capabilities of a particular service
    • resource automation mechanisms

    The situation was aggravated by a significant variety of technologies, including not only the ageless x86 classic with virtualization, but also the “favorite” enterprise solutions: high-end storage and RISC servers, clustering and geo-redundancy. Such a zoo imposes restrictions on the choice of platforms that can be considered to automate the provision of services. In addition, the choice should take into account compatibility with existing software - because the new platform should organically fit into the landscape of ITSM systems, realizing its purpose of automating query steps and providing resources.

    But choosing a platform is only the first step. The following are many others:

    • designing a standardized service catalog,
    • design and implementation of the cloud platform itself,
    • phased automation of services from the catalog (and the more heterogeneous the IT landscape, the longer and more difficult this step)
    • development and automation of a model for calculating the cost of services. You need to be able to translate the cost of hundreds of physical servers of several models into the cost of a virtual core, without forgetting to put in it a fraction of the cost of LAN, SAN, switching equipment and everything. That is still entertainment.
    • changing internal IT processes (this is not Amazon - you need to coordinate the provision of the service)

    It would seem that here it is happiness. But it was not there! Now is the time to migrate to this brave new world of legacy applications. But this is another story, which is not even about the clouds.

    Interestingly, in the end, a private cloud is most often attractive for two opposing types of companies. At one extreme, there are huge monsters from a harsh enterprise with their own developers, admins, and everything else. They get all the benefits of saving on a large scale. At the other extreme are companies that are unable to maintain their own data centers, but have great flexibility in organizing the infrastructure and can afford to assemble the most bizarre designs from material at hand, connecting them with almost blue electrical tape. Which is excusable as long as you can count your servers on the fingers of one hand. It is here that you can find unstable versions of software and the most exotic solutions. This is the advantage of a private cloud over a public one - you have free hands for any interesting experiments.

    Also popular now: