| Knowledge base || |
Today's data centres will be organized into different functional areas: server areas, storage areas, a central cluster of switches, routers, and high performance computers. This orderly arrangement for services requirements of growth, structural design of power supply and cooling system will have a little help.
As you can imagine, physical space is very difficult to increase, today's structure can be more expensive to support such non-urgent needs. Data centers will be in accordance with module area. When there is demand for new investment is needed, which will not hinder the effectiveness of investment by the passage of time.
Interconnection infrastructure on this issue affects line "linkage" demand (or "in-line"). Long ago, the data center is connected to the "anywhere". Many users in enterprises or wide area networks want to have access to all the services running on the server, so servers can directly access to the storage device.
Core switches is to promote joined backplane switching form, but also means that the wiring will need to provide each of these functional areas, and interconnection between each module and each core area. Maximum capacity, maximum wiring system has arrived, as the time grows, will need to add more lines and increased link speeds.
Traditional programmes, and when devices are not too much, from servers to switches and storage devices, will use the individual to directly connect the long jumper. In a small data center, there is merit in this. It has a relatively low cost but also feasible. But when the equipment and data center applications grows, this form of point-to-point direct attach jumper would slowly strangle a data center.
Everyone has the experience, as the time grows, more and more such as noodle-like jumper. Equipment change, increase becomes more and more difficult to control and management. Due to indirect storage losses, individual line performance and reliability began to decline. Finally, the available line distribution space has become crowded and blocks, seriously affecting the application of expansion in the future. Today's structured cabling design, slowly began to be accepted.
It is suited to the modularization of future data center design, increases manageability, reliability, and scalability. However,it also brings a higher initial construction costs and the need to pay attention to the wiring system generated by the added more join in extra losses.
When designing and construction of structured cable wiring system, the inevitable problems are: type of cable? There will be as many of the standard? For cable problem, the answer would be clear: OM3 (laser optimized 50um cable) can provide a lower total price (cable plus laser transmitter) can support enough equipment and link length, and bandwidth will support the 100Gbits/sec Protocol in the future.
OM4 standard also has been released. It is essentially a more high-bandwidth OM3 fiber supports more links and more. Because high speed data applications will typically use an existing link length shorter, so you must plan route structures and how to upgrade in the future.
Optical fiber quantity demand will require detailed plans and calculations. First, consider the exchange of data structures and levels in each Cabinet, end each line of Cabinet-level and collection-level. Secondly is the type and density of the device you want to use. However , this is not a simple and easy answers. Based on the type of device (such as blade servers) it is possible to design and provide the amount of fiber in each Cabinet.
There are two points to consider:
1. confirm that each row of cabinets to the total quantity of fiber, can be flexibly configured in the Configuring a day later and change the fiber usage of each Cabinet.
2. the planned fiber route and interconnect fiber optic distribution box for easy future Cabinet to increase fiber in each line. For example, the handling capacity upgrade 10Gbit links to 40G and 100G, means that each link from using 2 core fiber-optic upgrade to 8 or 20 core parallel fibers.
According to the development of equipment technology, it is hard to provide advanced from the beginning number of fibers. But with a flexible fiber-optic cabling systems to support future needs and increase the amount of fiber in the future the overall wiring system does not have much of an impact, it can be done. Suggested that at least in each Cabinet had arranged 24-core optical fiber.
For virtual computing, high-density blade servers, concentration of fibre channel over Ethernet, 10G lines, each Cabinet had arranged 48 or 96 conductor for high availability of fiber optic distribution box is a good option.
The most basic and essential components, jumper, is a system that is easy to manage, but it also may be the most complex and the most important thing. Usual problem is something like this: this jumper to the other end where you? If (jumper) is not plugged in can have a bad influence? How or where can I get new cables?
A sound labelling and filing system, is the first important thing, but most of the jumpers, particularly in regions of high density applications , often logic is beyond the user's control.
Port density will be constrained by a suitable for Fibre Channel interface type, SFP fiber optic converters, and duplex LC connectors, and in recent years has reduced the volume of fiber-optic connectors and fiber density improved twice times. In contrast, small volume of standardized connectors have cable from 3mm become 2mm and even smaller in diameter.
But now, we have too much fiber jumpers, they've struggled because from the problem is getting worse due to the bending of the scene. Lucky is that a new, flexible, bend-insensitive fiber appears, and can solve the above problem