When the time comes to buy server hardware, there are a lot of factors to consider, such as the number of processors, the available memory and the total storage capacity. Buyers should closely evaluate eight important features when comparing the servers available from the leading vendors.
These eight features cover the basic components to look for to buy server hardware, but they don't represent all the features that buyers should consider. Decision-makers at every organization must determine exactly what they need to support their existing and future workloads, keeping in mind the differences between rack, blade and mainframe computers.
Companies should view these eight features as the starting point to identify their requirements and evaluate the available products and should expand their research as necessary to ensure they're addressing every concern.
One of the most important components to consider when buying server hardware is the processor that carries out the data computations. Also referred to the central processing unit (CPU), the processor does all the heavy lifting when it comes to running programs and sifting through data. Most servers run multiple processors, usually with one per socket. However, a processor can also be made up of multiple cores to support multiprocessing capabilities.
Multiple cores usually translate to better performance, but the number of cores is not the only factor to consider. Buyers should also consider the processor speed -- CPU clock speed -- and available cache, as well as the total number of sockets, as these can differ significantly from one processor to the next.
For example, the NEC Express5800/D120h blade server supports up to two processors from the Intel Xeon Scalable product family. One of the most robust of these processors offers 26 cores, 35.75 MB of cache and a 2.0 GHz clock speed.
Compare that to the Dell PowerEdge M830 blade server, which uses Xeon E5-4600 v4 processors. The most robust of these offers 22 cores, 55 MB of cache and a 2.20 GHz clock speed. The Dell server also supports up to four processors rather than two.
With extensive research into the server market, TechTarget editors have focused this series of articles on server vendors with considerable market presence and that offer at least one product among blade, rack and mainframe types. Our research included Gartner, Forrester and TechTarget surveys.
Adequate server memory is essential to a high-performing system, and the more memory that is available, the better the workloads are likely to perform. However, other factors can also contribute to performance, such as the memory's speed and quality. Most server memory is made up of dual in-line memory module integrated circuit boards with some type of random-access memory.
Server memory might also include fault-tolerant capabilities or other features that enhance reliability. One of the most common capabilities is error-correcting code (ECC), a method to detect and correct common single-bit errors. When evaluating server hardware memory, you should look at the entire offering, keeping in mind the types of workloads and applications you run.
For example, Fujitsu's mainframe computers in the BS2000 SE series support up to 1.5 TB of memory. However, IBM's ZR1 mainframe, which is part of the z14 family, supports up to 8 TB of memory. The ZR1 also provides up to 8 TB of available redundant array of independent memory to improve transaction response times, a pre-emptive dynamic RAM feature to isolate and recover from failures quickly, and ECC technologies to detect and correct bit errors.
Servers vary greatly in the amount and types of internal storage that they support, in part because workflows and applications also vary. For example, a server hosting a relational database management system will have different requirements than one hosting a web application. In addition, the use of external storage, such as storage area networks (SANs), can also impact internal storage requirements.
When you buy server hardware, be sure to evaluate each prospective server to ensure it can meet your storage needs. Today, most servers support both solid-state drives (SSDs) and hard disk drives (HDDs). But buyers should certainly verify this support, as well as the server's supported drive technologies, such as Serial-Attached SCSI (SAS), Serial Advanced Technology Attachment (SATA) or non-volatile memory express (NVMe). Other considerations should include drive speeds, capacities, endurance and support for redundant array of independent disks (RAID).
For example, Oracle's X7-2 rack server can support up to eight 2.5-inch HDDs or SSDs, either SAS or NVMe, and multiple RAID configurations. Compare that to the Inspur TS860G3 rack server, which can handle up to 16 drives, either SSDs or HDDs, and support both SAS and SATA. However, the Inspur server does not support NVMe, which means the SSDs might not perform as well.
A server's ability to connect to networks, peripherals, storage and other components is essential to its effectiveness within the data center. The server needs the necessary connectors and drivers to ensure that it can properly communicate with other entities and process various workloads. Buyers need to determine exactly what type of connectivity is necessary and, from there, examine the server's specs to verify whether it will meet those requirements.
Servers differ widely in this regard, so buyers should look for specifics such as the number and speed of the Ethernet connectors, the number and type of USB ports, the availability of management interfaces, the types of protocols available, support for SANs and other storage systems, as well as whatever other components are necessary to facilitate connectivity.
Acer's rack server Altos R380 F3 is a good example of what connectivity features to look for when you buy server hardware. It includes two Ethernet ports, either 1 GB or 10 GB, an RJ-45 management port, three USB 3.0 ports, one USB 2.0 port, and a video port. In addition, the server offers up to seven Peripheral Component Interconnect Express (PCIe) 3.0 slots and one PCIe 1.0 slot.
Servers offer hot swapping capabilities to varying degrees. Hot swapping refers to the ability to replace or add a component without needing to shut down the system.
The term hot plugging sometimes refers to hot swapping, although, in theory, hot plugging capabilities are limited to being able to add components but not replace them without shutting down the system. Because of the confusion around these terms, it is best to verify how each vendor uses them.
One of the most common hot swappable components is the disk drive. For example, the Cisco UCS B480 M5 blade server supports hot swappable drives, as does the Huawei FusionServer CH242 V5 blade server and the Intel R2224WFQZS rack server.
With blade systems, the hot swapping capabilities are often within the chassis itself. One example is the chassis used for the Lenovo ThinkSystem SN850 blade server, which provides hot swapping capabilities for the fans and power supplies, in addition to the server's disk drives. However, these types of capabilities are not limited to blade servers. The Acer Altos R380 F3 system also supports hot swappable fans and power supplies even though it is a rack server.
Redundancy is important to ensure a server's continued operation in the event of a component failure. Most servers provide some level of redundancy, often for the hard drives, power supplies and fans. The Asus RS720-E9-RS12-E rack server, for example, offers redundant power supplies and the HPE ProLiant DL380 Gen10 rack server offers redundant fans.
As with its hot swapping capabilities, the redundancy available to blade servers is often located within the chassis. For instance, the chassis that support the Dell PowerEdge M830 blade server and Supermicro SBI-6129P-T3N blade server both provide redundant power supplies.
However, the Dell chassis also offers redundant cooling components, and the server itself provides redundant embedded hypervisors.
Admins must manage a server effectively to ensure its continued operation while delivering optimal performance. Most servers provide at least some management capabilities.
For example, many servers support the Intelligent Platform Management Interface (IPMI), a specification developed by Dell, Hewlett Packard, Intel and NEC to monitor and manage server systems. Not surprisingly, the servers offered by these companies, such as the Dell PowerEdge M830, HPE ProLiant DL380 Gen10, Intel Server System R2224WFQZS and NEC Express5800/B120g-h, are IPMI-compliant.
But servers are certainly not limited to IPMI capabilities. For example, the Acer Altos R380 F3 rack server comes with the Acer Smart Server Manager; the Asus RS720-E9-RS12-E rack server comes with the ASUS Control Center; and the Cisco Unified Computing System (UCS) B480 M5 blade server comes with Cisco Intersight, Cisco UCS Manager, Cisco UCS Central Software, Cisco UCS Director and Cisco UCS Performance Manager.
Blade systems usually provide some type of module to manage the individual blades. For instance, Huawei's FusionServer CH242 V5 blade system includes the Intelligent Baseboard Management System module to monitor the compute node's operating status and support remote management.
Not surprisingly, systems such as Fujitsu's BS2000 mainframes provide a variety of management capabilities. For example, each BS2000 system includes a management unit that works in conjunction with the SE Manager to offer a centralized interface from which to administer the entire server environment. And IBM's ZR1 mainframe includes the IBM Hardware Management Console (HMC) 2.14, the IBM Dynamic Partition Manager and an optimized z/OS platform for IBM Open Data Analytics.
Another important factor to consider is the server's security features. As with other features, servers can vary significantly in what they offer, with each vendor taking a different approach to securing their systems.
For example, the Lenovo ThinkSystem SN850 blade server provides an integrated Trusted Platform Module 2.0 chip to store the RSA encryption keys used for hardware authentication. The server also supports Secure Boot, Intel Execute Disable Bit (EDB) functionality and Intel Trusted Execution Technology.
Another example is the Oracle Server X7-2 rack server, which comes with the Oracle Integrated Lights Out Manager 4.x, a cloud-ready service processor for monitoring and managing system and chassis functions. On the other hand, the Huawei FusionServer CH242 V5 blade server supports the Advanced Encryption Standard -- New Instructions, as well as Intel's EDB feature and Trusted Execution Technology.
IBM's ZR1 mainframe is also strong when it comes to security. The server includes on-chip cryptographic coprocessors and the Central Processor Assist for Cryptographic Function (CPACF), which includes the new Crypto Express6S feature to enable pervasive encryption and support a secure cloud strategy. The CPACF is standard on every core. The platform also includes IBM Secure Service Containers to securely deploy container-based applications.