标签归档:data center

400G Data Center Deployment Challenges and Solutions

As technology advances, specific industry applications such as video streaming, AI, and data analytics are increasingly pushing for increased data speeds and massive bandwidth demands. 400G technology, with its next-gen optical transceivers, brings a new user experience with innovative services that allow for faster and more data processing at a time.

Large data centers and enterprises struggling with data traffic issues embrace 400G solutions to improve operational workflows and ensure better economics. Below is a quick overview of the rise of 400G, the challenges of deploying this technology, and the possible solutions.

The Rise of 400G Data Centers

The rapid transition to 400G in several data centers is changing how networks are designed and built. Some of the key drivers of this next-gen technology are cloud computing, video streaming, AI, and 5G, which have driven the demand for high-speed, high-bandwidth, and highly scalable solutions. The large amount of data generated by smart devices, the Internet of Things, social media, and other As-a-Service models are also accelerating this 400G transformation.

The major benefits of upgrading to a 400G data center are the increased data capacity and network capabilities required for high-end deployments. This technology also delivers more power, efficiency, speed, and cost savings. A single 400G port is considerably cheaper than four individual 100G ports. Similarly, the increased data speeds allow for convenient scale-up and scale-out by providing high-density, reliable, and low-cost-per-bit deployments.

How 400G Works

Before we look at the deployment challenges and solutions, let’s first understand how 400G works. First, the actual line rate or data transmission speed of a 400G Ethernet link is 425 Gbps. The extra 25 bits establish a forward error connection (FEC) procedure, which detects and corrects transmission errors.

400G adopts the 4-level pulse amplitude modulation (PAM4) to combine higher signal and baud rates. This increases the data rates four-fold over the current Non-Return to Zero (NRZ) signaling. With PAM4, operators can implement four lanes of 100G or eight lanes of 50G for different form factors (i.e., OSFP and QSFP-DD). This optical transceiver architecture supports transmission of up to 400 Gbit/s over either parallel fibers or multiwavelength.

PM4
PAM4

Deployment Challenges & Solutions

Interoperability Between Devices

The PAM4 signaling introduced with 400G deployments creates interoperability issues between the 400G ports and legacy networking gear. That is, the existing NRZ switch ports and transceivers aren’t interoperable with PAM4. This challenge is widely experienced when deploying network breakout connections between servers, storage, and other appliances in the network.

400G transceiver transmits and receives with 4 lanes of 100G or 8 lanes of 50G with PAM4 signaling on both the electrical and optical interfaces. However, the legacy 100G transceivers are designed on 4 lanes of 25G NRZ signaling on the electrical and optical sides. These two are simply not interoperable and call for a transceiver-based solution.

One such solution is the 100G transceivers that support 100G PAM4 on the optical side and 4X25G NRZ on the electrical side. This transceiver performs the re-timing between the NRZ and PAM4 modulation within the transceiver gearbox. Examples of these transceivers are the QSFP28 DR and FR, which are fully interoperable with legacy 100G network gear, and QSFP-DD DR4 & DR4+ breakout transceivers. The latter are parallel series modules that accept an MPO-12 connector with breakouts to LC connectors to interface FR or DR transceivers.

NRZ & PM4
Interoperability Between Devices

Excessive Link Flaps

Link flaps are faults that occur during data transmission due to a series of errors or failures on the optical connection. When this occurs, both transceivers must perform auto-negotiation and link training (AN-LT) before data can flow again. If link flaps frequently occur, i.e., several times per minute, it can negatively affect throughput.

And while link flaps are rare with mature optical technologies, they still occur and are often caused by configuration errors, a bad cable, or defective transceivers. With 400GbE, link flaps may occur due to heat and design issues with transceiver modules or switches. Properly selecting transceivers, switches, and cables can help solve this link flaps problem.

Transceiver Reliability

Some optical transceiver manufacturers face challenges staying within the devices’ power budget. This results in heat issues, which causes fiber alignment challenges, packet loss, and optical distortions. Transceiver reliability problems often occur when old QSFP transceiver form factors designed for 40GbE are used at 400GbE.

Similar challenges are also witnessed with newer modules used in 400GbE systems, such as the QSFP-DD and CFP8 form factors. A solution is to stress test transceivers before deploying them in highly demanding environments. It’s also advisable to prioritize transceiver design during the selection process.

Deploying 400G in Your Data Center

Keeping pace with the ever-increasing number of devices, users, and applications in a network calls for a faster, high-capacity, and more scalable data infrastructure. 400G meets these demands and is the optimal solution for data centers and large enterprises facing network capacity and efficiency issues. The successful deployment of 400G technology in your data center or organization depends on how well you have articulated your data and networking needs.

Upgrading your network infrastructure can help relieve bottlenecks from speed and bandwidth challenges to cost constraints. However, making the most of your network upgrades depends on the deployment procedures and processes. This could mean solving the common challenges and seeking help whenever necessary.

A rule of thumb is to enlist the professional help of an IT expert who will guide you through the 400G upgrade process. The IT expert will help you choose the best transceivers, cables, routers, and switches to use and even conduct a thorough risk analysis on your entire network. That way, you’ll upgrade appropriately based on your network needs and client demands.
Article Source: 400G Data Center Deployment Challenges and Solutions
Related Articles:

NRZ vs. PAM4 Modulation Techniques
400G Multimode Fiber: 400G SR4.2 vs 400G SR8
Importance of FEC for 400G

How to Utilize Data Center Space More Effectively?

What is data center space?

Data center space refers to the area of leased space available for servers to be stored in a data facility, including racks, cabinets, private suits, etc. It typically monitors all electrical and mechanical systems 24 hours a day, seven days a week. Nowadays, more and more companies choose data centers with larger space to meet their growing storage requirements.

However, many enterprises today encounter the challenges of limited data center space. One of the reasons is that the advancement of technology increases their demands for larger data center space, but it will cost a lot to build a new data center. Another factor is the underutilization of data center space. According to the research from an energy consortium called The Green Gird, 43 percent of respondents said they had no strategies in place to boost energy efficiency.

Therefore, it is necessary to learn some strategies to optimize the available space of data centers. Here are ten ways to make the best use of data center space.

How to utilize data center space?

  • Combine white space and gray space: Data center white space refers to the space where IT equipment and infrastructure are located, while data center gray space means the space where back-end equipment is located. By consolidating these two types of data center space, enterprises can use some technologies like cloud computing, which can save a large amount of space in data centers.
  • Refresh technologies: To improve data center space efficiency, technologies must be upgraded to minimize power consumption. For instance, new technologies like flywheels can increase the power of the machine, reducing the number of batteries required for the power supply. Besides, replacing the old and inefficient servers with new and energy-efficient servers can improve operational efficiency and reduce power consumption.
  • Use the smaller-diameter cable: Choosing the right cables is also an essential factor that should be considered. Tangled cables may cause cable congestion and then impede airflow. To prevent data center space from that problem, it’s necessary to use cables with smaller diameters, such as FS high-density fiber cables, which are more space-saving. They also allow rack space to be used to accommodate more equipment and reduce the demand for more cable management systems.smaller-diameter cable
  • Try virtualization solutions: According to the U.S. Environmental Protection Agency, most high-capacity servers are utilized at 15% or less, wasting space and power. Using virtualization technologies can reduce the number of new servers required to replace inefficient servers by sharing workloads among multiple servers, which can maximize data center space utilization.
  • Improve architecture efficiency: Data center architecture and the way that hardware is deployed have a vital impact on data center space. Terrible deployment may impede energy efficiency and lead to heating problems. Therefore, when planning a new data center, it is important to consider carefully the current design, future servers, and equipment, and how these devices will integrate with each other.
  • Optimize vertical space: Compared with horizontal data center space, making use of vertical space can increase the capacity and density of the data center without occupying floor space. Traditional racks and cabinets support from 42U to 45U of rack space, while taller racks offer up to 58U of rack space. Besides, it’s more efficient to use the space above the rack to patch the racks and cabinets.vertical space
  • Increase cabinet power density: Server racks and cabinets take up a lot of space, so it’s essential to make the best use of them. By increasing cabinet power density, the requirements for cabinets will be reduced, thus lessening the occupied floor space. Besides, this can also reduce management equipment and increase companies’ return on investment.
  • Use cooling technologies: Cooling accounts for about half of a data center’s entire energy consumption. Since computer room air conditioning (CRAC) and air handling units cannot handle the higher power densities, some companies may use liquid cooling systems, which take up a lot of valuable floor space. Using technologies like hot/cold aisle containment can save data center space to some extent while also maintaining suitable temperatures.

All the methods mentioned above work very well on boosting data center space utilization. The key is to choose a plan that best meets your goals and needs.

Article Source: How to Utilize Data Center Space More Effectively? | FS Community

Related Articles:

Data Center White Space and Gray Space | FS Community

What Is Data Center Storage? | FS Community

Data Center Containment: Types, Benefits & Challenges

Over the past decade, data center containment has experienced a high rate of implementation by many data centers. It can greatly improve the predictability and efficiency of traditional data center cooling systems. This article will elaborate on what data center containment is, common types of it, and their benefits and challenges.

What Is Data Center Containment?

Data center containment is the separation of cold supply air from the hot exhaust air from IT equipment so as to reduce operating cost, optimize power usage effectiveness, and increase cooling capacity. Containment systems enable uniform and stable supply air temperature to the intake of IT equipment and a warmer, drier return air to cooling infrastructure.

Types of Data Center Containment

There are mainly two types of data center containment, hot aisle containment and cold aisle containment.

Hot aisle containment encloses warm exhaust air from IT equipment in data center racks and returns it back to cooling infrastructure. The air from the enclosed hot aisle is returned to cooling equipment via a ceiling plenum or duct work, and then the conditioned air enters the data center via raised floor, computer room air conditioning (CRAC) units, or duct work.

Hot aisle containment

Cold aisle containment encloses cold aisles where cold supply air is delivered to cool IT equipment. So the rest of the data center becomes a hot-air return plenum where the temperature can be high. Physical barriers such as solid metal panels, plastic curtains, or glass are used to allow for proper airflow through cold aisles.

Cold aisle containment

Hot Aisle vs. Cold Aisle

There are mixed views on whether it’s better to contain the hot aisle or the cold aisle. Both containment strategies have their own benefits as well as challenges.

Hot aisle containment benefits

  • The open areas of the data center are cool, so that visitors to the room will not think the IT equipment is not being cooled sufficiently. In addition, it allows for some low density areas to be un-contained if desired.
  • It is generally considered to be more effective. Any leakages that come from raised floor openings in the larger part of the room go into the cold space.
  • With hot aisle containment, low-density network racks and stand-alone equipment like storage cabinets can be situated outside the containment system, and they will not get too hot, because they are able to stay in the lower temperature open areas of the data center.
  • Hot aisle containment typically adjoins the ceiling where fire suppression is installed. With a well-designed space, it will not affect normal operation of a standard grid fire suppression system.

Hot aisle containment challenges

  • It is generally more expensive. A contained path is needed for air to flow from the hot aisle all the way to cooling units. Often a drop ceiling is used as return air plenum.
  • High temperatures in the hot aisle can be undesirable for data center technicians. When they need to access IT equipment and infrastructure, a contained hot aisle can be a very uncomfortable place to work. But this problem can be mitigated using temporary local cooling.

Cold aisle containment benefits

  • It is easy to implement without the need for additional architecture to contain and return exhaust air such as a drop ceiling or air plenum.
  • Cold aisle containment is less expensive to install as it only requires doors at ends of aisles and baffles or roof over the aisle.
  • Cold aisle containment is typically easier to retrofit in an existing data center. This is particularly true for data centers that have overhead obstructions such as existing duct work, lighting and power, and network distribution.

Cold aisle containment challenges

  • When utilizing a cold aisle system, the rest of the data center becomes hot, resulting in high return air temperatures. It also may create operational issues if any non-contained equipment such as low-density storage is installed in the general data center space.
  • The conditioned air that leaks from the openings under equipment like PDUs and raised floor tend to enter air paths that return to cooling units. This reduces the efficiency of the system.
  • In many cases, cold aisles have intermediate ceilings over the aisle. This may affect the overall fire protection and lighting design, especially when added to an existing data center.

How to Choose the Best Containment Option?

Every data center is unique. To find the most suitable option, you have to take into account a number of aspects. The first thing is to evaluate your site and calculate the Cooling Capacity Factor (CCF) of the computer room. Then observe the unique layout and architecture of each computer room to discover conditions that make hot aisle or cold aisle containment preferable. With adequate information and careful consideration, you will be able to choose the best containment option for your data center.

Article Source: Data Center Containment: Types, Benefits & Challenges

Related Articles:

What Is a Containerized Data Center: Pros and Cons

The Most Common Data Center Design Missteps

The Chip Shortage: Current Challenges, Predictions, and Potential Solutions

The COVID-19 pandemic caused several companies to shut down, and the implications were reduced production and altered supply chains. In the tech world, where silicon microchips are the heart of everything electronic, raw material shortage became a barrier to new product creation and development.

During the lockdown periods, some essential workers were required to stay home, which meant chip manufacturing was unavailable for several months. By the time lockdown was lifted and the world embraced the new normal, the rising demand for consumer and business electronics was enough to ripple up the supply chain.

Below, we’ve discussed the challenges associated with the current chip shortage, what to expect moving forward, and the possible interventions necessary to overcome the supply chain constraints.

Challenges Caused by the Current Chip Shortage

As technology and rapid innovation sweeps across industries, semiconductor chips have become an essential part of manufacturing – from devices like switches, wireless routers, computers, and automobiles to basic home appliances.

devices

To understand and quantify the impact this chip shortage has caused spanning the industry, we’ll need to look at some of the most affected sectors. Here’s a quick breakdown of how things have unfolded over the last eighteen months.

Automobile Industry

in North America and Europe had slowed or stopped production due to a lack of computer chips. Major automakers like Tesla, Ford, BMW, and General Motors have all been affected. The major implication is that the global automobile industry will manufacture 4 million fewer cars by the end of 2021 than earlier planned, and it will forfeit an average of $110 billion in revenue.

Consumer Electronics

Consumer electronics such as desktop PCs and smartphones rose in demand throughout the pandemic, thanks to the shift to virtual learning among students and the rise in remote working. At the start of the pandemic, several automakers slashed their vehicle production forecasts before abandoning open semiconductor chip orders. And while the consumer electronics industry stepped in and scooped most of those microchips, the supply couldn’t catch up with the demand.

Data Centers

Most chip fabrication companies like Samsung Foundries, Global Foundries, and TSMC prioritized high-margin orders from PC and data center customers during the pandemic. And while this has given data centers a competitive edge, it isn’t to say that data centers haven’t been affected by the global chip shortage.

data center

Some of the components data centers have struggled to source include those needed to put together their data center switching systems. These include BMC chips, capacitors, resistors, circuit boards, etc. Another challenge is the extended lead times due to wafer and substrate shortages, as well as reduced assembly capacity.

LED Lighting

LED backlights common in most display screens are powered by hard-to-find semiconductor chips. The prices of gadgets with LED lighting features are now highly-priced due to the shortage of raw materials and increased market demand. This is expected to continue up to the beginning of 2022.

Renewable Energy- Solar and Turbines

Renewable energy systems, particularly solar and turbines, rely on semiconductors and sensors to operate. The global supply chain constraints have hurt the industry and even forced some energy solutions manufacturers like Enphase Energy to

Semiconductor Trends: What to Expect Moving Forward

In response to the global chip shortage, several component manufacturers have ramped up production to help mitigate the shortages. However, top electronics and semiconductor manufacturers say the crunch will only worsen before it gets better. Most of these industry leaders speculate that the semiconductor shortage could persist into 2023.

Based on the ongoing disruption and supply chain volatility, various analysts in a recent CNBC article and Bloomberg interview echoed their views, and many are convinced that the coming year will be challenging. Here are some of the key takeaways:

Pat Gelsinger, CEO of Intel Corp., noted in April 2021 that the chip shortage would recover after a couple of years.

DigiTimes Report found that Intel and AMD server ICs and data centers have seen their lead times extend to 45 to 66 weeks.

The world’s third-largest EMS and OEM provider, Flex Ltd., expects the global semiconductor shortage to proceed into 2023.

In May 2021, Global Foundries, the fourth-largest contract semiconductor manufacturer, signed a $1.6 billion, 3-year silicon supply deal with AMD, and in late June, it launched its new $4 billion, 300mm-wafer facility in Singapore. Yet, the company says its production capacity will only increase component production earliest in 2023.

TMSC, one of the leading pure-play foundries in the industry, says it won’t meaningfully increase the component output until 2023. However, it’s optimistic that the company will ramp up the fabrication of automotive micro-controllers by 60% by the end of 2021.

From the industry insights above, it’s evident that despite the many efforts that major players put into resolving the global chip shortage, the bottlenecks will probably persist throughout 2022.

Additionally, some industry observers believe that the move by big tech companies such as Amazon, Microsoft, and Google to design their own chips for cloud and data center business could worsen the chip shortage crisis and other problems facing the semiconductor industry.

article, the authors hint that the entry of Microsoft, Amazon, and Google into the chip design market will be a turning point in the industry. These tech giants have the resources to design superior and cost-effective chips of their own, something most chip designers like Intel have in limited proportions.

Since these tech giants will become independent, each will be looking to create component stockpiles to endure long waits and meet production demands between inventory refreshes. Again, this will further worsen the existing chip shortage.

Possible Solutions

To stay ahead of the game, major industry players such as chip designers and manufacturers and the many affected industries have taken several steps to mitigate the impacts of the chip shortage.

For many chip makers, expanding their production capacity has been an obvious response. Other suppliers in certain regions decided to stockpile and limit exports to better respond to market volatility and political pressures.

Similarly, improving the yields or increasing the number of chips manufactured from a silicon wafer is an area that many manufacturers have invested in to boost chip supply by some given margin.

chip manufacturing

Here are the other possible solutions that companies have had to adopt:

Embracing flexibility to accommodate older chip technologies that may not be “state of the art” but are still better than nothing.

Leveraging software solutions such as smart compression and compilation to build efficient AI models to help unlock hardware capabilities.

LED Lighting

The latest global chip shortage has led to severe shocks in the semiconductor supply chain, affecting several industries from automobile, consumer electronics, data centers, LED, and renewables.

Industry thought leaders believe that shortages will persist into 2023 despite the current build-up in mitigation measures. And while full recovery will not be witnessed any time soon, some chip makers are optimistic that they will ramp up fabrication to contain the demand among their automotive customers.

That said, staying ahead of the game is an all-time struggle considering this is an issue affecting every industry player, regardless of size or market position. Expanding production capacity, accommodating older chip technologies, and leveraging software solutions to unlock hardware capabilities are some of the promising solutions.

Added

This article is being updated continuously. If you want to share any comments on FS switches, or if you are inclined to test and review our switches, please email us via media@fs.com or inform us on social media platforms. We cannot wait to hear more about your ideas on FS switches.

Article Source: The Chip Shortage: Current Challenges, Predictions, and Potential Solutions

Related Articles:

Impact of Chip Shortage on Datacenter Industry

Infographic – What Is a Data Center?

The Most Common Data Center Design Missteps

Introduction

Data center design is to provide IT equipment with a high-quality, standard, safe, and reliable operating environment, fully meeting the environmental requirements for stable and reliable operation of IT devices and prolonging the service life of computer systems. Data center design is the most important part of data center construction directly relating to the success or failure of data center long term planning, so its design should be professional, advanced, integral, flexible, safe, reliable, and practical.

9 Missteps in Data Center Design

Data center design is one of the effective solutions to overcrowded or outdated data centers, while inappropriate design results in obstacles for growing enterprises. Poor planning can lead to a waste of valuable funds and more issues, increasing operating expenses. Here are 9 mistakes to be aware of when designing a data center.

Miscalculation of Total Cost

Data center operation expense is made up of two key components: maintenance costs and operating costs. Maintenance costs refer to the costs associated with maintaining all critical facility support infrastructure, such as OEM equipment maintenance contracts, data center cleaning fees, etc. Operating costs refer to costs associated with day-to-day operations and field personnel, such as the creation of site-specific operational documentation, capacity management, and QA/QC policies and procedures. If you plan to build or expand a business-critical data center, the best approach is to focus on three basic parameters: capital expenditures, operating and maintenance expenses, and energy costs. Taking any component out of the equation, you might face the case that the model does not properly align an organization’s risk profile and business spending profile.

Unspecified Planning and Infrastructure Assessment

Infrastructure assessment and clear planning are essential processes for data center construction. For example, every construction project needs to have a chain of command that clearly defines areas of responsibility and who is responsible for aspects of data center design. Those who are involved need to evaluate the potential applications of the data center infrastructure and what types of connectivity requirements they need. In general, planning involves a rack-by-rack blueprint, including network connectivity and mobile devices, power requirements, system topology, cooling facilities, virtual local and on-premises networks, third-party applications, and operational systems. For the importance of data center design, you should have a thorough understanding of the functionality before it begins. Otherwise, you’ll fall short and cost more money to maintain.

data center

Inappropriate Design Criteria

Two missteps can send enterprises into an overspending death spiral. First of all, everyone has different design ideas, but not everyone is right. Second, the actual business is mismatched with the desired vision and does not support the setting of kilowatts per square foot or rack. Over planning in design is a waste of capital. Higher-level facilities also result in higher operational and energy costs. A data center designer establishes the proper design criteria and performance characteristics and then builds capital expenditure and operating expenses around it.

Unsuitable Data Center Site

Enterprises often need to find a perfect building location when designing a data center. If you don’t get some site-critical information, it will lead to some cases. Large users are well aware of the data center and have concerns about power availability and cost, fiber optics, and irresistible factors. Baseline users often have business model shells in their core business areas that decide whether they need to build or refurbish. Hence, premature site selection or unreasonable geographic location will fail to meet the design requirements.

Pre-design Space Planning

It is also very important to plan the space capacity inside the data center. The raised floor to support ratio can be as high as 1 to 1, while the mechanical and electrical equipment needs enough space to accommodate. In addition, the planning of office and IT equipment storage areas also needed to be considered. Therefore, it is very critical to estimate and plan the space capacity during data center design. Estimation errors can make the design of a data center unsuitable for the site space, which means suspending project re-evaluation and possibly repurchasing components.

Mismatched Business Goals

Enterprises need to clearly understand their business goals when debugging a data center so that they can complete the data center design. After meeting the business goals, something should be considered, such as which specific applications the data center supports, additional computing power, and later business expansion. Additionally, enterprises need to communicate these goals to data center architects, engineers, and builders to ensure that the overall design meets business needs.

Design Limitations

The importance of modular design is well-publicized in the data center industry. Although the modular approach refers to adding extra infrastructure in an immediate mode to preserve capital, it doesn’t guarantee complete success. Modular and flexible design is the key to long-term stable operation, also meets your data center plans. On the power system, you have to take note of adding UPS (Uninterruptible Power Supply) capacity to existing modules without system disruption. Input and output distribution system design shouldn’t be overlooked, it can allow the data center to adapt to any future changes in the underlying construction standards.

Improper Data Center Power Equipment

To design a data center to maximize equipment uptime and reduce power consumption, you must choose the right power equipment based on the projected capacity. Typically, you might use redundant computing to predict triple server usage to ensure adequate power, which is a waste. Long-term power consumption trends are what you need to consider. Install automatic power-on generators and backup power sources, and choose equipment that can provide enough power to support the data center without waste.

Over-complicated Design

In many cases, redundant targets introduce some complexity. If you add multiple ways to build a modular system, things can quickly get complicated. The over-complexity of data center design means more equipment and components, and these components are the source of failure, which can cause problems such as:

  • Human error. Data statistics errors lead to system data vulnerability and increase operational risks.
  • Expensive. In addition to equipment and components, the maintenance of components failure also incurs more charges.
  • Design concept. If maintainability wasn’t considered by the data center design when the IT team has the requirements of operating or servicing, system operational normality even human security get impacts.

Conclusion

Avoid the nine missteps above to find design solutions for data center IT infrastructure and build a data center that suits your business. Data center design missteps have some impacts on enterprises, such as business expansion, infrastructure maintenance, and security risks. Hence, all infrastructure facilities and data center standards must be rigorously estimated during data center design to ensure long-term stable operation within a reasonable budget.

Article Source: The Most Common Data Center Design Missteps

Related Articles:

How to Utilize Data Center Space More Effectively?

Data Center White Space and Gray Space