These are the 5 most effective measures to improve the efficiency of your servers

The servers installed in today’s data centers are characterized not only by their power, but also by their high energy consumption. They consume so much that more than half of what is spent in an entire data center is consumed by them. For this reason, there are quite a few people responsible for this type of installations that seek different ways to reduce your energy consumption. Not only for environmental issues, but because lower consumption also implies lower spending. But without affecting performance.

There are different ways to achieve this, but according to the Uptime Institute, there are several that, with their application, achieve reduce the bill and electricity consumption without a decrease in performance. To reach these conclusions, the center has prepared a report based on testing 429 server platforms.

To test them, the SERT (Server Efficiency Rating Tool) database was used. The SERT suite is an industry standard for measuring server efficiency. Uptime has analyzed data from Intel and AMD servers (mostly AMD EPYC and Intel Xeon chips) in this suite, noting the advantages and disadvantages of each type of processor based on workloads. It has also analyzed server generations from 2017, 2019 and 2021. Based on the data obtained, the institute has selected the five measures that allow the efficiency of servers to be further increased. They are the following:

1 – Replace old servers with more efficient ones

Without a doubt, older servers have a lower level of energy efficiency than current ones. Thus, for example, the efficiency of servers with an Intel chip has improved by 34% between 2017 and 2019 when their CPUs showed 50% occupancy, while in the case of servers with AMD, the improvement has been 140%. %, according to the aforementioned report.

In addition, between 2019 and 2021, servers with Intel CPUs improved efficiency by 32%, while those with AMD CPUs improved in this regard by 47%. Intel servers were more efficient than AMD servers in 2017 at all levels of CPU usage, but since 2019 AMD has taken a leap forward in this regard. Thanks to that, with 2021 servers at 50% usage, a server with an average AMD is 74% more efficient than one with an Intel chip.

2 – Do not underuse servers

Servers that are not running at full power are a waste of energy. In 2020, only 36% of companies used their servers at 50% or more. In 2022 this percentage was still very low: 47%. And this data may be inflated, because some companies that answered the institute’s questions may have referred in their answers only to their servers with the best performance.

Usage rates would generally be lower, because many applications do not run consistently. For example, business software is used a lot during office hours, but it drops significantly outside of that period. Therefore, the use of servers can be increased by having those that host business apps execute tasks for which the execution time is less important outside of office hours.

To achieve maximum impact, companies should assess increased usage, as well as upgrade servers to the latest models. Combining increased usage with server replacement can more than double server efficiency. This leads to a more than 100% increase in workloads processed with the same amount of energy.

When done on a large scale, the savings in money and operations, as well as the increase in energy efficiency due to reduced energy consumption, can be very high. In addition, with improved system performance. Of course, it is necessary to take into account that the direct replacement of an old server with one or a higher capacity without increasing the workload supported by the old one reduces the usage rates of the system, and reduces some of the advantages offered by the change. .

This leads to the need to optimize task scheduling to increase their execution level on servers when hardware is upgraded. In this way it is possible to carry out all workloads with fewer servers. Hence, replacing systems can be cheaper than originally thought.

3 – Choose more powerful machines

Purchasing more powerful hardware can result in a higher level of energy efficiency. In the particular case of AMD servers, efficiency improves sharply as the working capacity of a server increases. In the case of servers with Intel, the improvements are also notable, but less high.

4 – Increase server cores

Another way of improve notably The efficiency of servers involves increasing the number of their processor cores. In the case of 2021 AMD servers, which recorded an increase in cores from 8 to 64, their efficiency was tripled. In the case of Intel’s 2021 ones, this increase was smaller, but no less notable.

At this point it is important to note that not all workloads are capable of using all available cores. In fact, some will operate more efficiently with processors with fewer cores, so it is important to match the capabilities of the processors with the needs of the applications running on the server on which they are installed. Only in this way will the greatest possible efficiency be achieved in this regard. In some cases, hypervisors and virtual machines can be used to maximize utilization, but not all applications are suitable for these environments.

5 – Manage consumption effectively

The server power saving features they can further improve their energy efficiency scenario. So much so that they can improve it by more than 10%. These features mean that the voltage and frequency of the CPUs can be raised or lowered, and cores that are not in use can be moved to a low-power standby state. But many companies do not use them because they are worried about performance or latency problems.

According to the Uptime Institute report, power management can increase latency by 20 to 80 milliseconds, which is not acceptable for certain workloads. In addition, there are some applications that companies decide not to use because they may cause performance or response problems. Of course, there are tasks and applications in which these slight delays will not have an impact on the business, and with which they can be used without problems.

Of course, in this case the biggest problem in achieving optimization is that some companies want to avoid taking risks at all costs. That is why it is recommended, at the time of purchase of new servers and run your performance tests, Make sure you test if power management negatively affects to the applications or not.

If it does not affect them, you can use the energy saving and management functions without problem, and implement a set of energy saving functions that allow you to save on your consumption without reducing the response time and performance necessary for the execution of tasks. be affected.

Therefore, to optimize the energy consumption of the servers, it is necessary to take five measures: upgrade them to a more current generation of servers, choose servers with a high computing capacity in terms of number of transactions per second, increase the number of cores by choosing processors Furthermore, be aware that although a server may be more energy efficient its overall power consumption may increase even as its efficiency improves and use power management features by reducing CPU frequency and voltage as usage increases, putting the idle cores into a standby state.

If you are thinking about changing your company’s servers for ones that are more energy efficient but have better performance, and you don’t know which one to decide on, take a look at this HPE guide to choose the most suitable one for your business.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *