Blade server vs Rack server, which one should you choose?

2017-09-30
The server that we often see, from the appearance of the type can be divided into three types, tower servers, rack servers and blade servers. Due to the limited space of the enterprise room and other factors, blade servers and rack servers are increasingly popular with users, then what are the characteristics of them, which one is better?  This article will discuss this problem with everybody.
Blade server vs Rack server, which one should you choose?
The tower server and its characteristics


Tower server is the most seen one, its shape and structure are almost the same with the ordinary PC host, but a slightly larger head, the size does not have a uniform standard.

Tower server motherboards are highly scalable, and the chassis will often be reserved for space, in order to carry out hard drives, power, and other redundant expansion. Tower servers do not require additional equipment, the installation environment is not too much demand, and has good scalability, and thus a wide range of applications, to meet the common server application needs.

But things are double-sided, tower server also has many limitations, such as the need to use multiple servers at the same time to work to meet the high demand for server applications, because of its relatively large size, take up more space, management complexity It is not suitable.

Blade server and its characteristics

The blade server is a high-availability, high-density, low-cost server platform designed specifically for a particular application industry and a high-density computer environment. Its main structure is a large main chassis with many “blades”, The blade is actually a system motherboard, similar to a separate server, they can boot from the local hard drive their own operating system. Each blade can run its own system, serving the specified different user groups, not associated with each other. Moreover, you can also use the system software to integrate these motherboards into a server cluster. In cluster mode, all blades can be connected to provide high-speed network environment, shared resources, for the same user base services. By inserting a new blade in the cluster, you can improve overall performance. And because each blade is hot-swappable, the system can be easily replaced and the maintenance time is minimized.

Depending on the server function required, the blade server is divided into blade blades, network blades, storage blades, management blades, Fiber Channel SAN blades, expansion I / O blades, and so on. Blade server recognized the characteristics of two, one is to overcome the shortcomings of the chip server cluster, has become a cluster of Terminator; the other is to achieve the cabinet optimization.

Rack server and its characteristics

Rackmount servers are servers that can be installed directly into standard 19-inch racks. Typically, such servers are somewhat similar to the size of the switch, so the rack server is actually a product under industrial standardization, and its appearance is designed according to a uniform standard, With the unified use of the cabinet to meet the enterprise server-intensive deployment needs. The main role of the rack server is to save space because more than one server can be installed in a cabinet, not only can take up less space, but also easy to manage. The height of a common cabinet is 42U (1U = 1.75 inches or 4.4 centimeters), the rack server is 19 inches in width, and most rack servers are 1U-4U high.

The advantage of rack servers is that they are small in space and easy to manage, but because of internal space constraints, scalability is more limited, for example, 1U servers are mostly only 1 to 2 PCI expansion slots. In addition, the cooling performance is also a need to pay attention to the problem, in addition to the need for cabinets and other equipment, so this server is used for a large number of servers for large enterprises to use, there are many companies using this type of server, but the server delivered to Dedicated server hosting to host, especially the current site of many servers are used in this way.

Blade server vs Rack server

The current blade server products in the multi-core, low-power technology driven by the initial pursuit of the high-density blade server, to emphasize the overall performance, high productivity of the new generation of blade products. The blade server will be the focus of development with higher density, agile deployment, and maintenance, full-scale monitoring management, high scalability, high availability, and become a mature mainstream product with rack-mounted servers. The current blade server in a number of industries have opened a breakthrough, the market demand has been stimulated, blade server market, the future of concern.

At present, as the server field of the star, this high-density blade server market prospects have been optimistic about the major server manufacturers, including of course, including IBM, HP, SUN, DELL and other international big and like Lenovo, Dawn, and other well-known brands. Let’s take a look at the advantages of blade servers and rack servers.

Latitude 1: Spatial density
Blade servers are more space-saving than rack-mounted servers, and in fact blade servers increase space density by 50 percent for rack servers. In the cabinet system configuration under the premise of the 1U rack optimization server system transplanted to the blade server, the occupied space is only the original 1/3 to 1/2. In a standard cabinet environment, the blade server processing density to increase four to five times. For example, in a high-density computing server environment that handles 1024 nodes, if you use a 1U-configured server, you need 24 racks, which do not include the rack space occupied by the Ethernet switch hub, and with eight “blades” The blade server requires only nine cabinets, but includes the cabinet space for the Ethernet switching hub.

Latitude 2: Infrastructure Cabling and management efficiency
The blade server in the Infrastructure Cabling as long as the unified network cable, power lines, blade servers do not need between the artificial wiring; and the rack server will have to each server’s network cable, power line wiring, if a 42U cabinet Install more than 1U of the server, the cabinet behind the wiring is very much, it seems messier. If we need 10 more reliable 1U rack servers and need KVM, network and management functions, then we need accessories are 20 power (redundant power), 10 KVM cables, at least 20 cables (One for each server to manage the line and network), which does not include cables that connect external storage and other devices. The cabling of the blade server solution is significantly better than the 1U rack solution.

From the TCO point of view, blade servers are easier to manage, provide more processing power in less space, and spend less. Because many components can be shared, with the 1U server is different. For example, in the above example, we use the blade server, you no longer need 10 DVD drive, 10 different remote management module, and 20 power supply.

Latitude 3: Server flexibility
Most blade servers are less flexible than rack servers. For example, blades in high-performance database applications, the external RAID card cannot be equipped with a disk array. In addition, if the user needs large-capacity memory database server, and do not want to build the cluster, the blade will also appear powerless. Because a rack server with 16 memory slots is easy to buy on the market, most blades support only 4-8 of the memory slots. Can be seen, the blade server flexibility is weaker, even if there is a suitable product, the price is also incredibly high.

In most cases, the blade supports a 2.5-inch hard drive, which is not only more expensive than a 3.5-inch hard drive but also less performance, thanks to the “blade pursuit of high density”. Today, a typical blade can be installed at the most 140GB hard drive, and many rack products can be equipped with 2TB (4X500GB) storage capacity.

Latitude 4: Procurement costs
In theory, the cost of purchasing a blade is lower than an equivalent number of rack servers due to the reduction of many repetitive, unnecessary components such as DVDs, power supplies, and KVM and network cables. Blade server manufacturers each have their own proprietary blade architecture, which makes IBM, HP, and Dell blades high prices because of their monopolistic nature.

It is thought that the cost of a fully equipped blade server is much less than that of the most expensive rack servers of the same manufacturer. But in fact, if the user is “gradually increasing the blade in the chassis” approach, you will find that the rack server cost-effective. Because most of the blades and related products are more expensive than the corresponding rack products, the general Blade chassis needs 4000-8000 of dollars, the blade is also higher than the 1U rack price, at least not cheap.

Latitude 5: Extensibility and standardization
The advantage of 1U rack server is that it adopts standard server design technology, standard peripherals, standard interface, with RAID function, redundant function, can run independently and undertake tasks. In terms of scalability, the rack server has a poor expansion performance in the chassis because of the small chassis space, but it can be extended very efficiently through server clusters or in an external expansion cabinet. Blade servers are innovative in scaling up and outward expansion. Adding a new server typically involves inserting a new single processor or multiprocessor blade server into the open Bay of the chassis. The blade server can be plugged into an extended infrastructure. In addition, the options module inside the chassis allows us to add functionality that can be shared once it is externally connected. The modular design of blade technology can be extended rapidly.

IBM, HP, and Dell Blades are not compatible, there is no standardization, and there are different blade chassis standards for the same manufacturer. The rack server is a unified standard, and IBM’s 2U servers can be installed in Dell’s 42U server cabinets.

Latitude 6: Reliability and Maintenance
With a common 1U rack server scheme, by a large number of power cords, network cable generated by the plug point formed a large number of potential “problem points”, replaced by Blade server chassis and Blade server solution, the original network, power supply points reduced, but also increased the reliability of the system.

All key components of the blade server can be redundant or hot-swappable, including cooling systems, power supplies, Ethernet controllers and switches, Midplane and backplane, hard drives and service processors. Removing a server for repairs simply means dragging the blade out of the chassis, as simple as removing a hot-swappable hard drive. Advanced Blade Server System provides a smart way to achieve highly sensitive maintenance, the advanced diagnostic function can guide the maintenance staff to find the defective parts directly, so as to achieve rapid and effective recovery, some blade servers will not even appear single point of failure. Rack server maintenance is more complicated than blades.

Conclusion

In short, rack servers and blade servers are functional servers that specialize in a specific application. Rack-type servers can meet the needs of user applications, blade server integration is too dense, coupled with the standard, thermal aspects have not been completely resolved, and the blade server is suitable for a large number of machines together to form a cluster for users to use, so the blade server so far, compared with the rack-style has not accounted for the absolute advantage.
RECENT BLOG POST
  • 01
    2019-10
    With the continuous development of 5G communication technology, 100G modules are gradually becoming popular. We know that there are many kinds of packages for 100G optical modules. From 2000 to now, the optical module package types have been rapidly developed. Its main package types are: GBIC, SFP, XENPAK, SNAP12, X2, XFP, SFP+, QSFP/QSFP+, CFP, CXP. In the fast-developing network era, some 100G optical modules avoid the risk of being eliminated, and upgraded and revised with the wave of the Internet, such as 100G CFP optical modules.
  • 01
    2019-10
    1. What is the CWDM SFP? The CWDM optical module is an optical module using CWDM technology to implement the connection between the existing network device and the CWDM multiplexer/demultiplexer. When used with a CWDM multiplexer/demultiplexer, CWDM optical modules can increase network capacity by transmitting multiple data channels with separate optical wavelengths (1270 nm to 1610 nm) on the same single fiber.
  • 01
    2019-10
    AOC is the abbreviation of Active Optical Cables, which is called Active Optical Cables in Chinese. AOC active optical is to encapsulate two optical modules and cable together. Because the medium of transmission in the middle is optical cable, AOC optical module, which contains laser devices, has a higher price for DAC. However, its optical aperture is not exposed, it has high reliability, and its working distance can be customized for a long distance of less than 100 meters.
  • 01
    2019-10
    Dense Wavelength Division Multiplexing (DWDM) technology is capable of transmitting data in an optical fiber using bit wavelength parallel transmission or string line transmission using the wavelength of the laser.It is widely used in different fields of communication networks, including long-distance backbone networks, metropolitan area networks (MANs), residential access networks, and local area networks (LANs).The DWDM optical module is the optical module that uses this technology, so the DWDM optical module has high bandwidth and long-distance transmission characteristics.