In metropolises worldwide, owners of downtown shops and restaurants are coming to terms with the fact that WFH (work from home) appears to be here to stay. And this is just the latest significant shift in consumer and professional lifestyles to turbocharge the movement of data, applications – even entire corporate information systems – to the cloud.
Enterprises have discovered the productivity, accessibility, and reliability benefits gained when data systems run on a cloud computing platform. Growth in cloud computing creates demand for more network bandwidth and higher data-transfer rates inside cloud computing platforms and between their users. This growth includes third-party services such as Amazon Web Services (AWS), Microsoft Azure, and private clouds. The proliferation of cloud computing technologies throughout networks and at universal endpoints, such as smartphones and PCs, has led to the coining of the term cloudification to describe the trend.
And this trend is putting unprecedented pressure on the crucial timing systems which regulate the high-frequency operation of devices such as servers, network switches, and routers. So how should computing and network equipment manufacturers respond? Which options do they have to shift their timing technology up in response to the substantial changes in operating requirements engendered by cloudification?
Companies have turned to MEMS precision timing devices to address the demands of cloudification, while also meeting the space requirements of ever-smaller equipment.
As AMD put it, precision MEMS timing products such as SiTime’s solution providing clocking for the new AMD Alveo X3 series, “bring low latency and adaptive computing onto a single platform to give customers a competitive advantage for today and tomorrow.” AMD used the clock system on a chip to supply all critical clocks. The device AMD used also replaced multiple oscillators and resonatators and made it possible to simplify its clock tree design, which helped save verification time. This accelerated its time to market.
The stepwise march of networking technology
Networking and communications technologies typically advance in large, generational leaps, each of which provides a dramatic upgrade in speed, bandwidth, and network capacity. This shows in the moves from 2G through 3G and 4G to today’s 5G wireless network technology. In the core telecoms network, operators have been able to upgrade Gigabit Ethernet equipment to 10G, 25G, 100G, and beyond.
At each shift, the carrier wave frequency increases, tightening the margin of error around the timing signals that synchronize device or network operations. And the speeds at which today’s optical communications modules or PCI Express interfaces operate are so fast that it can be challenging to meet the crucial jitter specifications for timing components. It is particularly true when they are affected by noise from external sources such as power supplies.
In timing systems, jitter is the variation of a clock from its correct position in time. When the system frequency doubles, then the allowable jitter halves. So, equipment manufacturers should not be surprised that they continually face pressure to uprate the jitter performance of the timing components. Systems as diverse as 5G radio units, network switches, network processor boards, servers, and Wi-Fi® access points face this challenge.
Manufacturers can most effectively meet the predictable performance need by building the specification of the timing system into the architecture development phase of new design projects. This shift represents an essential change from traditional practice, in which the selection of timing components is one of the last decisions made before finalizing a production design. The tighter timing margins become, the more this last-minute approach to timing harms the design process--leading manufacturers to specify expensive components, compromise design specifications, and extend development time in the search for adequate timing components.
Clock frequency and jitter performance are not the only aspects of timing provision affecting cloudification. The operating environment for timing systems is also changing as cloud-based applications find new host device types. The original home of cloud computing applications, of course, is the data center. But today, cloud computing applications run on many different end-node types. One example is a premium car that continuously connects to cloud-based diagnostics or tracking. Others include communications applications as a node in a cloud computing network or a mobile phone with which a remote worker connects to corporate data systems.
This distributed network architecture raises new questions about the implementation of timing technology. A timing component in an AWS rack-mount server does not experience the shock and vibration of a construction engineer’s phone or tablet when surveying a building site. The operating environment of a car – in the frozen north of Sweden or the baking desert of Saudi Arabia – is very different from the climate-controlled stability of a data center.
The rise of cloudification means timing components are subject to tighter performance constraints, with smaller margins for jitter, and must handle a wider range of harsh operating conditions.
How can all those in the value chain, from cloud computing service providers to networking and edge equipment manufacturers, respond to ensure that timing system designs are future-proof? And that the designs will not compromise or hamper the implementation of the latest computing technologies? There are three changes that enterprises should consider making now.
A strategic plan for timing
The first is to treat timing similarly to other functions common to most or all products in a manufacturer’s portfolio. Many networking equipment manufacturers have central design teams that develop Linux® operating systems or SERDES interfaces for use across all products. A few forward-thinking technology companies have also created a central team for developing timing systems, with benefits such as enabling time synchronization across multiple platforms. This strategic, centralized approach brings focused attention to the increasingly demanding specification of timing technology. It allows a company to build and use expertise in this functional domain across all product lines.
Dedicated timing expertise can also ensure the architecture of a product from the very beginning of its development, including the timing system specification. The result allows the architecture to fit the timing solution under development.
An open mind about new timing technology
Old habits die hard. The quartz crystal has been the electronics engineer’s only mainstream option for timing signal generation since the early 20th century. Today, however, new opportunities eliminate drawbacks associated with quartz. Crystal oscillators are notoriously prone to a predictable drift in frequency performance in varying temperatures. There are wild variations in performance between crystals from a common batch of production units. And crystals cope badly with shock and vibration.
Newer micro-electro-mechanical systems (MEMS)-based timing technology eliminates these drawbacks. MEMS timing devices provide the predictably stable and low jitter required in the latest high-speed networking and communications equipment. Stable over temperature, robust, and uniform performance from one unit to another, thanks to a 100% silicon supply chain. Since MEMS timing components can handle vibration and temperature swings, equipment manufacturers can specify the same timing device for different use cases. This flexibility reduces the number of stock-keeping units to maintain and simplifies a manufacturer’s logistics operations.
Take advantage of expertise where you find it
The third response to the growth of cloudification is to enhance the organization’s understanding of the issues involved in timing system design. In high-bandwidth systems, timing has the potential to make or break the system. Yet, it is rarely the primary focus of an equipment manufacturer’s top engineers. Their expertise is in the packet handling, interface design, and processor systems.
MEMS-based timing manufacturers have the product and technology knowledge to help customers create a platform to accelerate design completion while enhancing performance across a broad product portfolio. An organization can quickly and easily augment its in-house engineering knowledge base by taking advantage of the expertise of timing component manufacturers. And unlike quartz crystals, MEMS-based timing components are programmable and can be reused across multiple functional domains and device types. This allows equipment manufacturers, for the first time, to take a strategic approach to develop a platform for implementing timing systems.
Organizations can also enhance their approach to timing system provision by participating in industry initiatives such as the Open Compute Project’s Time Appliance Project (OCP-TAP). These initiatives invite collaboration between the ecosystem of timing system users, suppliers, researchers, and developers to accelerate the adoption of new technologies and solutions. Together they seek to meet the increasing demand for high-performance timing components.
A successful response to the demands of cloudification
Delivering on the demands of a connected, data-rich world requires a strategic approach, embracing new timing technology, and taking advantage of the expertise of third parties. By adopting these three measures, companies at the epicenter of the cloudification trend can ensure their timing systems achieve the best performance and the fastest time to market.
Deepak Tripathi is director of product marketing at SiTime and Gary Guist, PhD, is senior manager of technical marketing at SiTime.