Key factors to understand when an enterprise chooses a 1U rack server

This article will introduce the choice of enterprise 1U Rack Server to understand the key factors, compared with the internal characteristics of a variety of 1U servers, expansion capabilities and management functions, and to provide buyers with a current mainstream 1U server specifications list.

Processor

Processors are often the primary consideration in many functional 1U servers. Today's 1U servers can typically support 1~2 CPUs, such as the Intel Xeon 5500 or Xeon 5600 series. AMD processor-based servers typically use the Opteron 4100 or Opteron 6100 series. Fujitsu Primergy RX200 S6 1U server (see table 1) uses 1 or 2 Xeon 5600 series processors, while Dell's PowerEdge R415 1U Server employs 1 or 2 Opteron 4100 series processors.

For each processor family, there are many models to choose from; Each processor has a unique mix clock rate, kernel count and cache size. The current processor typically contains 6 to 8 cores. Each 1U server can provide amazing computing power when we multiply the number of processors to the kernel, and by processing instruction optimizations (such as Intel Hyper-Threading (HT) technology). You can run multiple virtual machines (VMS) on a server as long as you manage processor resources and computing power through a virtualization Management program (hypervisor). You can also select only one processor in the procurement server and then upgrade the hardware later. Remember that typically the processor needs to match consistently to ensure that the card model and speed are the same as the manufacturing step (revision or version).

Memory

The second important factor to consider is server memory. A virtual machine usually runs out of memory (or RAM) before it runs out of processor performance. Enterprises that deploy server virtualization should pay more attention to memory than to processors. Today's 1U servers can support a large amount of memory. Take the Super Micro computer 1026t-rft+ 1U Server as an example, you can manage the 18-slot dual-column direct-insert memory Module (DIMM) as small as 2GB, up to 192GB in size. The Dell PowerEdge R415 1U Server supports up to 8 DIMM slots for maximum 128GB capacity.

It is noteworthy that this type of bulk memory typically uses registered (cached) DIMMs. It also supports non-cached DIMMs, but the total amount of memory supported is much lower-probably only half or one-third of the cached DIMMs. Note that memory is typically allocated to each processor, so 1U servers with 2 processors and 12 DIMM slots can manage 6 DIMM slots per processor. Almost all 1U servers use the current fastest DDR3 memory to match the current processor bus (front end) speed.

There are some features in memory that deserve attention. Error correction Code (ECC) is the current mainstream memory that corrects a single bit error in memory and detects a double bit error. Server administrators can also choose Non-ECC memory, but discard detection and error correction, memory problems may lead to critical application failures. Mission-critical 1U servers can also benefit from single device data correction (SDDC) memory, which extends ECC mode to check and correct memory chip-level errors.

Memory hot spare is the same as the principle of disk thermal standby. As an example of an extended memory module, it is already installed but not used. Once a memory device fails, hot spare memory will replace the failed memory operation until the failed memory is replaced and returned to normal. Memory mirroring is similar to RAID 1 (disk mirroring). The data is replicated to two different memory regions and is compared. If a memory area is detected with an error that cannot be corrected, the system obtains the data from the mirrored area.

Storage

Storage is another major feature of 1U servers, and today products can support a variety of miniaturized modules (typically 2.5-inch specification) SATA (Serial ATA) and SAS (serial Attached SCSI) disks, often with an optical drive, such as a small CD-ROM or DVD-ROM for installing software. The question of how to deploy the server and how much local storage is required is a matter for careful consideration. If the server is to run the load on a local disk, you need to provide it with sufficient capacity and performance.

SAS, for example, provides low capacity and high performance for SAS disks, and is an ideal choice for general applications. SATA provides higher capacity and lower performance and is more suitable for file and print servers. The HP Proliant DL360 G7 and the Dell PowerEdge R415 1U Server can also support solid-state drives (SSD). Although the current price of solid-state drives is extremely high, it can provide a good service to storage-performance-sensitive applications such as databases and Exchange Server 2010. The number of disk bays depends on the server model and selection, and some 1U servers do not have the disk installed by default and need to check that the server unit configuration meets the requirements before placing an order. A server that plans to use shared storage does not have to consider how local disks should be configured.

When administrators take a local disk, they need to consider using RAID technology to protect the data. Modern 1U servers support hardware-based raid, such as RAID 0, RAID 1, RAID 5, and RAID 6, as well as raid 10,raid 50 and RAID 60. If the disk controller located on the server board cannot support RAID, you need to add a separate RAID controller (and backup battery) to the server's expansion slots.

Expansion slots

1U servers typically offer a certain amount of expansion through 1 or more slots. These slots are usually prepared to install an expansion board on the motherboard, and can be installed with enhanced boards such as RAID controllers, network adapters, Fibre Channel host bus adapters, and other hardware for connecting to a storage area network (SAN). Note the distinction between PCIe (PCI Express) and PCI (pci-extended) slots--Two types of boards are not interchangeable.

In the case of a super Micro 1026t-6rft+ 1U Server, the server supports a thin PCIe x4 (4-channel) board and can support two PCIe x8 (8-channel) cards. In contrast, HP Proliant DL360 G7 supports a pcie x8 with a PCIe x16 board, but if the riser is replaced, the server can support PCIe or PCI cards. You need to make sure that the upgrade card used for the server matches the number of channels in the server's internal space and slots. Otherwise, it may not work.

Power consumption

Each server is required to transmit energy through standard AC (AC) lines, with output voltages and type matching cables, which may vary depending on the region and the country. Power problem is worth paying attention to, focusing on server power automatic regulation and fault-tolerant function.

The Super Micro 1026t-6rft+ 1U Server provides a redundant, hot-swappable 700W Power module-Each power supply module at the back of the chassis supports independent AC power supply. Other servers, such as HP Proliant DL360 G7, default to a single 460W power supply, can also support additional redundant power supplies or use a higher power 750W power supply. Servers require a higher energy supply if multiple CPUs, large amounts of memory, extended drives, and some expansion boards are configured.

Redundant configuration has great advantages. If a single power supply fails, the server can still work. Furthermore, the server can achieve good fault tolerance through two independent AC power supplies. The server can also power up to two separate UPS systems separately. Budget-intensive organizations prefer to use one power connection to the power supply, while another is connected to the UPS scheme. This saves you from purchasing another UPS cost as a redundant power supply.