Many people doubt that clouds are cheaper than their own hardware - we find this point of view quite often in ActiveCloud. Some use the clouds because of the flexibility, the second want to get away from the routine, the third need centralization, the fourth - security. However, the clouds are not only convenient, but also beneficial, and we will try to explain why.
When evaluating the efficiency of the transition to the cloud Customers often tend to compare the cost of owning clouds and iron in the forehead. For example, if we buy 5 servers with 40 processor cores and 256 GB of RAM, then we also request similar resources from the cloud provider (40 * 5 = 200 vCPU + 256 * 5 = 1280 GB vRAM), and then we compare the costs for 3 years or even 5 years old.
Unfortunately, in most cases, this approach will not be objective, since it does not take into account a number of important nuances that directly affect the cost of ownership.
1. The resources required for fault tolerance are not taken into account.
In the cloud, issues of fault tolerance are
already thought out - virtualization hosts are clustered, and in a cluster, resources of at least one virtualization host are reserved so that client servers in case of a host failure or shutdown during scheduled maintenance can be moved to the backup host. At the same time, the service provider does not require an additional fee (in excess of the indicated cost of resources) for such a reserve.
In the case of your hardware, redundancy losses will have to be deducted from the resource pool potentially available on the equipment. In our example with 5 servers, the Customer, while maintaining fault tolerance, will not be able to load servers on the processor and memory by more than 80%; otherwise, in the event of a failure of one of the hosts, some servers will not be able to be restarted physically due to lack of resources.
Of course, you can expect that in the event of an accident you will be able to temporarily stop a number of non-critical services, but this approach in real conditions usually does not work.
The same is true for cloud storage - the price of virtual disk space already includes multiple-fault-tolerant raid types and backup hot-swap disks, and in the case of its own storage, 45-65% of raw disk space will be available depending on the selected type of RAID array.
2. Does not take into account the load limit of servers and storage systems
Practice shows that x86 servers should not be constantly loaded by more than 70-80% on the processor and 80-90% on memory, otherwise performance drops can occur, especially noticeable during startup of maintenance tasks (for example, during backup). In this case, short-term surges in the load of iron, as a rule, are experiencing normal, but with long-term work in this mode, the performance degradation of the IT systems located on this hardware is very likely.
For this reason, many service providers not only follow the rules for permissible marginal disposal of their equipment *, but also fix such obligations in an agreement with the Customer.
Excerpt from the SLA contract for our VMware cloud
On the data storage system, it is also not always possible to dispose of all the available volume resulting from the assembly of RAID arrays, since a reserve is required for the operation of various functional storages (for example, hardware snapshots or tiring). The size of such a reserve may be 10% or more of the space available for addressing.
When using storage systems on mechanical disks (HDD), marked-up moons are not recommended to fill more than 70-80% with an intensive disk load in order to avoid degradation of storage performance. SSD drives of such a problem are deprived, but all-flash storage is still quite expensive today, and not every company has a profitable purchase.
3. The costs associated with limited scaling are not taken into account.
It is no secret that when planning the resources consumed there is always a certain reserve in case of growth. The size of this reserve is individual, but the factors affecting it are more or less known.
You can build on the statistics of consumption growth in previous years - if this data is available, then we are laying the same dynamics and a little on top of unintended urgent initiatives of business units. If there are no such statistics, then you will have to make an assumption yourself based on the plan of future projects and the company's strategy for growth and geographic expansion. Such a forecast is unlikely to be accurate, but it is still better than to be without the power needed by the business "here and now"
In a period of high competition in almost all b2c markets, businesses are forced to rely on the speed of launching and launching new products and services on the market, because often the first in the market removes the cream, and the second gets losses.
In such conditions, to maintain the desired pace, relying on its own infrastructure, can only afford the major players who can invest in their own expertise, or even IT platform, and reallocate resources between projects, if necessary.
For smaller and medium-sized companies, it will be much faster to scale as needed on cloud capacities, since you will not be able to quickly increase your hardware. Even if the company has implemented standards for the equipment and adjusted the process of calculating the configuration in the selected vendor, it will take at least 2-3 months before the iron is calculated, supplied, mounted and configured for possible operation. If the procedure for the acquisition of fixed assets is quite complicated, as often happens, and corporate regulations require a tender each time with the participation of several competing manufacturers, then the process of expanding the resource pool can easily stretch for six months or more.
Of course, in the case of the cloud, planning and budgeting is also necessary, but if planning is wrong, the company will not suffer significant losses - cloud capacity can always be scaled by actual consumption, and there is no need to pay extra until it is really needed.
4. The costs of inaccurate sizing of key business systems are not taken into account.
Accurate resource planning for key accounting information systems can be a daunting task, since each implementation individually and often involves a large amount of customization for specific business processes. In this case, when sizing, it remains to rely on the recommendations of the system developers, who will certainly put possible risks - after all, they do not want to justify themselves to the Customer for the slow work of the system after implementation. Having received the recommendations of the manufacturer, its own IT specialists will also create risks because they do not want to listen to reproaches from users. As a result, it turns out that the manufacturer reinsured when providing recommendations one and a half times, and the IT department reinsured when ordering equipment one and a half times, and as a result, the purchased expensive iron stands idle for more than half, and you can’t take the excess back to the store.
The cloud not only saves from such situations, but also can greatly ease the lives of those who rent power at the time of implementation, thus performing sizing on a really deployed system. In this case, however, there are some risks of compatibility - after moving to your hardware, system performance may suddenly decrease for no apparent reason. But this approach is still better than buying equipment blindly.
5. It does not take into account the cost of maintaining its infrastructure.
Any hardware, be it network equipment, servers or storage, must be serviced. The same can be said for virtualization, monitoring, security, backup and other services that ensure the stable and secure operation of the company's key business systems. Staff capable of maintaining all this in working condition and repairing it in the event of an accident is quite expensive. Often - unnecessarily expensive for not the largest business.
Based on statistical data, including HH.RU ( https://stats.hh.ru ).At the same time, the cumulative complexity of the used stack of infrastructure technologies becomes higher every year, and along with it the requirements for the qualification of service technicians are constantly growing.
Retaining such people in the state, even with a plump budget, is becoming increasingly difficult, since, in addition to salaries, it is also necessary to ensure that specialists who are standing in front of them with a stream of tasks with increasing complexity and responsibility. It’s usually unprofitable to compete for them with core IT companies.
By placing the infrastructure in the cloud, the Customer thereby actually gains access to a diverse team of qualified architects and engineers whose experience and competencies are constantly growing. The more closely a company interacts with a service provider in its projects, the more it can consider it not only as a cloud service provider, but also as a constantly accessible IT partner who is ready to share experience with the client and solve its tasks in a comprehensive manner.
It is important to note here that renting facilities and services in the cloud, of course, will not relieve the company from the need to invest in its IT service, however, it will allow you to review the staff by selecting specialists with narrower competencies relevant to your business systems, shifting responsibility for infrastructure and border services
on the shoulders of the provider .
6. The cost of vendor support for the used software and equipment is not always taken into account.
No matter how famous the manufacturer is, no one is still insured against a manufacturing defect and errors in the operation of the software. It is especially annoying when, because of this, part of the functionality does not just work incorrectly, but operation of acquired IT assets as a whole is impossible, and it’s impossible to resolve the issue promptly with the manufacturer, even if expensive next business day support is purchased.
This happens for various reasons: sometimes the vendor has not yet built processes, and support is provided in a limited mode, sometimes there are no necessary details in the spare parts kit, sometimes custom firmware is required, which takes more than one week to write, but not one The result is always the same - the time spent, the nerves, the deadlines of the project and the call to the carpet to the business. I think many will agree that it is much more pleasant to transfer such risks onto the shoulders of the provider.
7. It does not take into account the cost of construction and maintenance of its server room.
Cloud providers serve a large number of customers, and the inaccessibility of the cloud for any reason strongly affects the business reputation of the service provider, even if the failure was not his fault. Clients do not care how big the tree fell on the data center, and what color was the excavator that damaged the optics. Therefore, the question of choosing and equipping the site is usually not worth it - we rent a Tier3 data center and connect several Internet channels with DDoS protection.
To provide a similar level of reliability and security in the local server, you will have to invest in it about $ 100,000 per rack, but even in this case the commercial data center will win according to a number of parameters.
Even if you do not set yourself the task of building a server according to international standards, and only provide acceptable operating conditions for your equipment, capital costs will still be noticeable - you will have to allocate a room, build the ACS, connect a couple of external communication channels, bring quality power lines that can endure a dozen other kilowatts, properly organize cooling and ventilation, take care of the UPS, and ideally duplicate the main engineering systems and put the industrial system on arotusheniya and ACS.
And of course, all of the above will have to maintain, which is unlikely to cost less than a quarter of a million a year for each full rack.
You can make your life easier and place your equipment in a commercial data center. However, the benefits of outsourcing non-core IT operations are quickly becoming obvious, and from the lease of racks, Customers are switching to a hybrid of their hardware and gradually replacing cloud services.
If, when comparing, we take into account the nuances described above, then in most cases it will turn out that clouds enable businesses to reduce costs, even if choosing a provider focuses on the quality of services and the number of available services.
Quickly calculate the cost of cloud resources will help the
calculator on the ActiveCloud website - you can immediately take into account in the calculation of backup, technical support, operating system licenses and other options. If you have a need to compare the cost of ownership of the cloud and your own infrastructure in detail - write to
dmitriy.yashin@activecloud.ru .