How to Choose the Right Server: Brands, Bandwidth, CPU, Memory, and More

1. Server Brands

There are many server brands used in various fields. Commonly used brands include Dell, HP, IBM, Lenovo, and Huawei. Additional popular brands are ZTE, Tsinghua Tongfang, Fujitsu, Hikvision, and others.

2. Main Parameters for Selecting Servers

2.1 Bandwidth: How many users can a 5M bandwidth server support?

5 Mbps equals 625 KB/s of exclusive bandwidth. Assuming a typical 50 KB web page, 625 KB/50 KB ≈ 12.5 concurrent users can be served without issue. In practice, because requests are sequential, a 5 M bandwidth can support many more users, especially with small pages.

2.2 CPU

The CPU (Central Processing Unit) is the core processing and control unit of a computer. Performance depends on clock frequency, cache size, instruction set, and architecture.

Clock speed : Higher MHz/GHz means faster data processing.

Cache : Larger cache improves data‑hit rates, reducing memory accesses.

Core count : Typically each core has one thread; Intel Hyper‑Threading can provide two threads per core, e.g., a dual‑core CPU can handle four threads.

Thread count : More threads increase parallelism but also power consumption.

2.3 Chipset

In x86 systems, the chipset (now CPU + PCH) provides interfaces and compatibility, allowing various CPUs to work with different motherboards.

2.4 Memory

Servers use ECC (Error‑Correcting Code) memory. Larger memory improves performance, especially for databases, proxy services, and web services. Recommended minimums: entry‑level ≥2 GB, workgroup ≥4 GB, department ≥8 GB.

2.5 Storage

Common storage interfaces include:

SATA : Serial ATA, second generation (SATA II), provides higher transfer speeds.

SCSI : Small Computer System Interface, offers high bandwidth, low CPU usage, and hot‑swap capability, used in mid‑ to high‑end servers.

SAS : Serial Attached SCSI, serial technology for higher speeds, still considered first‑generation.

SSD : Solid‑State Disk, no mechanical parts, offers low power consumption, shock resistance, high stability, and low temperature tolerance.

To increase storage capacity and reliability, department‑level and enterprise servers often use SAS RAID cards to build disk arrays, while entry‑level servers may use inexpensive SATA RAID cards.

2.6 Network Card

Servers must provide fast network communication. At least one gigabit NIC should be installed on the fastest port, and special servers (e.g., FTP, file, video‑on‑demand) may require two gigabit NICs. Gigabit NICs should be placed in a 64‑bit PCI slot.

2.7 Redundancy

Reliability is the most important metric for servers. Redundancy measures include:

Disk redundancy : RAID arrays using multiple disks prevent data loss if a single disk fails.

Component redundancy : Multiple NICs, power supplies, and fans ensure continued operation after a component failure.

Hot‑swap : Ability to replace disks or cards without powering down, essential for 24/7 operation.

2.8 Expandability

Servers are designed for scalability, offering numerous drive bays, memory slots, CPU sockets, and expansion slots, allowing upgrades of storage, memory, CPUs, and additional cards. Power modules can also be expanded when many drives are added.

Other parameters exist but are omitted for brevity.