Myths about the infrastructure in the cloud: what kind of illiteracy we face in Russia every day

My name is Alexander, and I, among other things, do what I communicate with those who have just learned about the clouds. And he wants to transfer part of his infrastructure there. Most often, these are productions not from Moscow, which only recently learned about virtualization as such.

In half of the cases, these are not the most advanced IT guys. Some live in the last century. Friends come to them, they say that there are cloud providers. Then begins communication with us.
')
My favorite episode is the miscalculation of prices for new equipment in the office (a rack of cheap segment servers or converted PCs) and clouds:

- So. License for hypervisor. Why pay for it? We never bought a license ... Russia, hehe! Support for iron in the rack? To hell, we will manage without support, our admin fumbles. What else is there backup? We already have a singed ...

There is also a popular myth that if the server is in the office, then when the verification comes, you can take and pour boiling water from the kettle, and then no one will take the data. It is so common that sometimes it seems to me that the domestic market of low-end servers directly holds on it.

How does the company understand that you need to look at the cloud

It's about two things: money or responsibility.

With money, everything is obvious: the transfer of infrastructure to the cloud provider allows you to save. Almost everyone understands this. True, they usually think about it at the moment when something breaks. The financial director remembers the words “financial responsibility” and makes his count. He sees that the service costs 7 million, and the same inside the company - 10 million per year. Plus there is a financial responsibility at the level of SLA.

Another responsibility is the question that someone will be responsible for the infrastructure, except for the admin or CIO / CTO. Very often, if a business gets up due to IT breakdowns, this is the problem of the IT manager. The fact that he said two years ago that it would be good to purchase a couple of hard drives in the server for worn out RAID, or that he insisted on including support for hardware in the budget, but he was not given any money, these are frequent stories.

It turns out that the risk control instruments were deprived of it, and responsibility for these risks was left.

Therefore, the request most often comes from either the director or the head of the IT department. Both roughly represent what they want, but clarification is needed at the meeting. Then the following happens: they tell you what the needs are, and we are trying to translate this into a kind of technical requirements and roughly calculate. At this stage of discussion, what is needed and who needs it, a person who can talk to people, understands the needs of the business, and at the same time understands the technique, is necessarily involved. As a rule, this is a former engineer who developed along a branch of organizational skills. Its analytics functionality: talked and clarified needs. I thought how to close these needs with our solutions. He formulated a proposal for closure and has already presented, presented, substantiated, proved why it is correct.

Then the stage “why are you a good company?” Begins. In fact, this question is also about responsibility, or rather, about its transfer to a large company. This is such an additional informal, but very important argument. Because if something breaks! and the contractor is small, this one. And if something breaks down and the contractor is big, this is no longer a head jamb, but practically force majeure. It looks like this: “I found a big company, it provides a service. If she messes up, it is they who have to grunt, not me. They were chosen by the competition commission on the basis of factors ... ". It is clear that all the same, the manager, in the case of downtime, grows dryukaty, but the corporate world is so arranged that a large contractor is punished much less. Because the fault is on external circumstances. It’s like when the trunk provider once again drops the channels, it’s not the IT person’s fault, it’s the insuperable forces of nature.

And then, in the process of coordinating the proposal, we inevitably begin to disassemble myths.

Myths

1. We take the Mercedes "Gelentvagen" and put on it half of the spare parts from the UAZ. Including brake pads. In order to save on individual components. It turns out a great car model "Gelentuazen." We are about the so-called configuration, where everything is fine with respect to computing resources, but for the DBMS, drives are used up close to speed: those that are usually intended for archival storage or storage of “cold” data. I don’t know what kind of alchemy works in the head of a number of people, but reducing configuration on the DBMS at the last moment to reduce the total budget is a frequent case. "Gelentuazen" outside looks like a cool car, but can not brake. From the DBMS, the story is reversed. It can brake.
2. Often the customer comes and says that he has important data. And he needs a certain grade certification. For example, on 1K. And rolls out the requirements for such an infrastructure. First, 1K is not used for many years. Secondly, it may well turn out that he has data of the third category, and he simply does not know about it, because someone is either incompetent, or does not know how to properly execute documents, or decided to be safe. It happens that a lawyer is armed with a set of myths and confidently defends them.
3. Scaling speed There are two more types of customers who do not always know how to calculate the cost of iron. These are research institutes and retail. Retail has large seasonal peaks, for which it is not always ready even at the level of the site. And scientific research institutes have projects with a big miscalculation. For example, they may consider exploration a month, and for another three months their iron in the rack will stand idle. Therefore, the financial director also begins to look at recycling and clutches at his head. There are many more peaks: on payment aggregators, on transport workers in season. And, if they are, you should always think that instead of buying a new iron, it may be worthwhile to consider a partial migration. These companies have a very long planning step. Iron is bought for 3 months, and sometimes it is needed right yesterday. But no: 2 weeks - analysis, 2 months - tender, month - contracting. Then delivery. If lucky, commissioned - after 3 months. But not everyone is so experienced. In the cloud, however, many things are being done at the level of “today a letter of guarantee - tomorrow they deployed twice as many virtual machines”. Naturally, this is not the case everywhere. If you need to increase capacity by half, not everyone is ready to work on warranty. It is better to find out this question in advance.
4. Medium and small businesses assume that when the inspection bodies arrive, the safest thing is the storage room, where the counter stands, and the kettle boils next to it or the fire hammer lies. Next, the server is flooded with water, or the hard disk is broken to shreds. In such cases, we simply show 30 machine rooms, each with 100 racks. Guess where the cloud is, and which racks specifically refer to your data? And how to make them so as not to drop someone's banking or IT state-owned enterprises? Separate exactly where whose physically impossible. And to withdraw everything at once is a national incident. And yet, even if suddenly someone removes everything, from the equipment that you pulled out, you still need to collect something working. And it most likely will not work.
5. "You will not give away my data." If any one customer, even a small bakery with a check of 5 thousand rubles a month, suddenly will not be able to collect the data, or we will steal it, or lose it, or put it in open access, or merge it with someone, or otherwise discredit us, as a cloud provider, Khan. It is impossible to hide a pack if you are a public service provider. Therefore, you make an effort to avoid it. Well, immediately about the level of intervention: we don’t go further than the hypervisor in principle, that is, we don’t even know if the customer has licensed operating systems or not. And they should not know. And do not ask.

Testing, rakes and conclusions

Before confirming the start of a project, the customer usually begins to take an active interest in other enterprises from adjacent fields: they talk up, change their impressions, hit each other on the liver with expensive alcohol, exchange sincere impressions and take the minimum set of tests. Some take everything at once, because they rightly assume that the greater our coverage of services, the more complex the responsibility, and the easier it is to ask us for obligations.

Here, it seems, everything was agreed, all the questions were asked, the answers were received - it's time to start the test implementation. This is where mid-level performers come on stage, who have a completely different motivation and mode of action, different from the tops.

The largest "setup" at this stage: it happens that the CIO wants, but his subordinates do not. Because there were 50 IT professionals, and after the move, there will be 30. The cloud to some extent replaces their work, and they are afraid of losing it. And then begin the traditional stick in the wheel. Alas, but the fact that after the transition to the cloud infrastructure begins to reduce, it is a fact.

Here I can not refrain from lyrical digression. In my opinion, an advanced approach, usually peculiar to Western companies, when the IT department is not just something like cleaners, but a full-fledged business unit , this is a big breakthrough in the market of virtualization providers. Why? Because two factors:

1. The IT department issues internal accounts to departments, and in the cloud it can say exactly who consumed resources and how much (with the exception of routing payment resources like VM routers, but this is a few percent).
2. CIOs have a task to increase business efficiency. Or increase control. They even have goals in the money from the business: risk reduction, transition to opex, increase in labor productivity. Therefore, all the money saved and the risks that were not worked out immediately increase the premium, and it can be 2-3 annual salaries, if everything is good. Therefore, immediately begins the most efficient use of capacity.

Well, the last myth is that the cloud is a must. No, not always: there are tasks when everything is decided exactly by the office desk. Or there are companies whose budgeting is such that it is easier to buy iron every 3 years and support immediately, rather than operating expenses. Well, and, of course, there are people who have always lived on physical hardware and simply don’t want to bring something beyond the organization’s perimeter: their stands are also important for them. Somewhere on the border of these tasks there is a layer of myths, and these myths are sometimes quite actively fueled by server vendors.

The text was prepared by Alexander Kuznetsov, head of strategic sales management at Technoserv Cloud .