-
- 0-OUTLET - SUPER PROMOTION
- Racks
- BLOG
- RUKAM University
- RUKAM Certified
- Contact
Choosing a server does not simply mean looking for the system with the most cores, the most memory or the largest number of drives. The right configuration depends on the workload, number of users, applications, expected growth, availability requirements and budget.
In this first part of Rukam's Server Buying Guide, we explain how to evaluate the three components that most strongly affect server behavior: processor, RAM and storage.
Before selecting a configuration, answer a few questions:
There is no universally “best” configuration. There is a configuration that is better suited to the workload.
Before CPU, memory and drives, it is also important to define the physical platform:
Platform choice should consider space, expansion, power consumption, cooling, management and the infrastructure already available.
The CPU executes instructions from the operating system, applications and server services. When comparing processors, the main points include:
A processor with a higher clock speed is not automatically faster in every application, just as a processor with many cores is not automatically the best choice. The workload determines which characteristic matters most.
Workloads that depend mainly on a small number of threads may benefit more from high per-core performance. Virtualization, multiple simultaneous services and applications that can parallelize tasks may make better use of a larger number of cores.
For database servers, virtualization and applications licensed per core, the choice should also consider licensing, number of VMs, actual application behavior and CPU generation.
Processors should therefore be compared as part of complete platforms, rather than by looking only at GHz or core count.
Threads represent execution flows that a processor can handle. With technologies such as Intel Hyper-Threading, a physical core can expose more than one logical processor to the operating system.
This can improve the use of internal processor resources for certain workloads, but an additional thread is not equivalent to an additional physical core. The performance gain varies by application.
RAM holds data and instructions that need to be quickly available to the processor. In servers, the required amount depends directly on the application.
When sizing memory, consider:
In addition to total capacity, consider DIMM type, supported speed, memory-channel population, number of installed processors and platform-specific rules.
Installing more memory than the workload actually needs does not necessarily improve performance. The goal is sufficient capacity for the current workload, appropriate headroom and room for future expansion.
HDDs and SSDs serve different purposes.
HDD remains an important option when the priority is large capacity at a lower cost per TB, especially for file storage, backup, repositories and predominantly sequential workloads.
SSD offers lower latency and higher I/O performance and is generally better suited to operating systems, virtualization, databases and applications with heavy random-access activity.
The two technologies can also be combined, using SSDs for performance-sensitive workloads and HDDs for high-capacity storage.
The main storage interfaces found in Enterprise servers include:
When choosing drives, check interface, backplane, controller, physical form factor, speed, capacity, compatibility and hot-swap support.
The nominal interface speed should not be confused with the actual performance of the drive. For example, SATA 6G and SAS 12G interfaces are expressed in Gb/s, not GB/s.
Physical form factor also affects the configuration:
The number and type of server bays should be considered together with required capacity, redundancy, future expansion and the installed controller.
RAID level is also part of server sizing.
RAID can provide redundancy, performance or a combination of both, depending on the level used. Choosing between RAID 1, RAID 5, RAID 6, RAID 10 or other options should take into account drive count, required usable capacity, write performance, fault tolerance and rebuild time.
It is important to distinguish:
Possible platforms: Dell PowerEdge R720 / HPE ProLiant DL380 Gen8 or equivalent systems, when compatible with the application and support requirements.
The configuration should be validated according to the operating system, application, data volume, number of users and backup policy.
Possible platforms: Dell PowerEdge R730 / HPE ProLiant DL380 Gen9 or equivalent systems, according to application and support requirements.
For SQL Server, virtualization or other workloads licensed per core, processor selection should also consider the licensing impact.
Two servers with similar core counts and memory capacity can differ significantly in performance, efficiency, management features and expansion capability because they belong to different generations.
When comparing used or refurbished equipment, also consider:
Previous-generation Enterprise equipment can continue to serve many workloads very well, often at a significantly lower cost than a new platform.
The decision should consider not only age but also equipment condition, provenance, tests performed, parts availability, power consumption, performance, compatibility and expected service life.
At Rukam, equipment covered by the RUKAM Certified standard goes through the applicable selection, testing and preparation processes before being offered for sale.
Do not choose a server based only on the specification sheet.
CPU, memory and storage need to work in balance and according to the application. An oversized configuration can increase cost without providing a real benefit; an undersized configuration can compromise performance and expansion.
If you have questions about compatibility or sizing, Rukam's technical team can help evaluate the configuration best suited to your environment.
In Part 2: platform, redundancy, RAID, networking, expansion, remote management and other important points when choosing a server.