- June 2002 -
[Click Message To Learn More]
This paper offers new insights into the technical and financial benefits of incorporating enhanced flash storage solutions into residential gateway products.
by Gerard O'Driscoll, DigitalMentors
The availability of new in-home interconnection technologies, combined with the explosion of non PC-based devices, is driving the demand for a single device to connect in-home appliances to the public Internet. The residential gateway, as its name implies, is a central entry and control point at the home for all voice, video and data services, as well as the cornerstone for future services. It is the true enabler of the Age of Information.
The push by network operators to deploy new revenue-generating services in the home and the widespread adoption of the Internet are contributing to the mass deployment of residential gateways. But the lack of well-defined business models, and the development of gateway products without robust, secure and flexible storage systems, is slowing down mass deployment.
This paper offers new insights into the technical and financial benefits of incorporating enhanced flash storage solutions into residential gateway products.
In today's rapidly changing communications industry, the boundaries have begun to blur between PC networks, Internet-based systems, gaming consoles and broadcast television. The evolution of new services, as well as the convergence of some pre-existing services, has led to the development of a new category of products, residential gateways. In simple terms, the residential gateway (RG) is a device that sits between the broadband access network (WAN) and the in-house network (LAN), serving as the core of the home network. It enables bi-directional communication and data transfer among networked appliances in the home and across the Internet. This gateway - also known as a service gateway, media gateway or home gateway - is the key ingredient to delivering ubiquitous, high-speed Internet access to consumers around the globe. It provides a strategic platform and convergence point for integrating different broadband access types and several in-home networking solutions such as HomePNA, HomeRF, wireless LANs or IEEE 1394. Also important, the RG serves as an access platform through which service providers can remotely deploy services to the home from the Internet.
- Increased availability of new home networking technologies
- Increased number of homes that include some form of intelligent, electronic control system
- Network operators expanding their service offerings
- Increased use of the Internet
- Increased demand for non PC-based appliances
- International standards are being defined
- Increased availability of broadband connections
- Availability of new entertainment options, including digital television
Inside a Typical Gateway
Since broadband technology is relatively new and continuously evolving, the functionality of an RG must also evolve, and its design must remain relevant for the next 5 to 10 years. This means building it in a modular, future-proof fashion - both where software and hardware are concerned. To better understand the RG, it is essential to understand its key components:
- A processor
- Memory - volatile (RAM) and non-volatile (flash memory)
- A digital modem
- Home networking chipsets
- Relevant software
The processor (short for microprocessor and also often called the CPU or central processing unit) is the central component of the RG, responsible for almost every task the RG performs. It determines which operating systems (OSs) can be used, which software packages the RG can run, how much energy the RG uses, and how stable the system is, among other characteristics. The processor is also a major determinant of overall system cost. The newer, the faster and the more powerful the processor, the more expensive the RG - but also the more future-proof. Some of today's leading processors are actually Systems-on-a-Chip (SoC), all-in-one solutions that integrate the networking components and the CPU into one cost-effective chip. Such solutions are widely available from Broadcom, Conexant and National Semiconductor.
Just as a computer needs memory to function, a gateway also requires memory to store and manipulate instructions issued by the subscriber or the operator. It acts, so to speak, as a staging post between the disk and the processor. The more data available in the memory, the faster the gateway runs. The gateway uses two main categories of memory - RAM and flash memory. RAM is volatile, and thus loses its contents when power is turned off or fails. Flash memory is non-volatile, safeguarding in the absence of power the gateway's operating system and customizable features, as well as user data, network data and many other elements. It is the keeper of the crown jewels, if you will, and should be reliable, enduring, and offer hardware and software protection for the code and data stored within.
Much of the configuration flexibility is made possible by the gateway's support for a range of modems. These modems provide connectivity to the following types of broadband access networks:
- Direct Broadcast Satellite
- Fixed wireless broadband
- Two-way cable networks
- Power-line area networks
On the home networking side of the RG, chipsets provide the interface to the particular technology running on the home network. In addition to the various types of broadband and home networking chipsets, all gateways contain computing resources that support the software required to operate the device.
Software running on the RG enables the smooth inter-operation of information appliances and services within the home. Software hides the complexities of the system elements from the consumer. The RG must have a super-stable OS to regulate the allocation of resources, such as memory and processor utilization. This OS must have a small footprint to be cost effective for the consumer electronics market, yet it must also allow for a flexible and modular software configuration to enable the coveted future-proof device structure for remote updates and upgrades. A refreshing example of such a software package is Jungo's OpenRG. Remote management client software components are also crucial to achieving an extended gateway life span. Wyse technologies' Rapport software is one such fine example of how remote device management, diagnostics and asset management can be achieved smoothly and efficiently.
The design of the RG is dependent on the services offered by the network operator. Traditionally, the service provider has reaped revenues only up to the doorpost of the home. RGs provide operators with a platform to deliver traditional services, plus value-added services, right into the home. These services include:
- Communications - Service providers of consumer applications who offer
shared Internet access, both for PC and Internet appliances, and who enable
networking of multiple devices in the home will be in demand. In addition,
value-added telephony services, such as voice over IP and multiple phone
connections over one line will attract more consumers.
- Security - Security systems based on an RG are replacing existing security
systems. New and advanced services such as remote monitoring and control of
the system are done through a standard Internet browser or mobile phone.
This effectively integrates current core services with security services,
offering a broader range of security to consumers.
- Home automation - Bundling home automation services with existing
residential services is attracting consumers who were not previously
interested in home automation.
- Yet unknown services - If a design is to be truly future-proof, it must allow for the swift introduction of future services yet unknown to the RG's designer and service provider. This can be achieved only through a truly modular architecture. Such a design must incorporate a clever, expandable hardware platform, as well as a robust file system allowing for OS and application upgrades and for the deployment of new services. 2Wire's RG is a good example of such a gateway.
Categories and Types
RGs are consumer-oriented devices that are used to route digital content between the Internet and a home network. Different types of gateways are currently available:
- PC-based Home Servers -The PC-based gateway, also known as a home server,
is conceptually the easiest path to the RG largely due to the vast installed
base of PCs, their processing power, the pre-existence of a digital modem
and presence of a software platform for deployment of home networking
- Digital Video Recorders (DVRs) - DVRs are loaded with software that allows
consumers to digitally record and store TV programs. They also enable
viewers to pause live TV and possibly to skip commercials. Today, DVR's
provide an interface to a broadband digital network. In the not too distant
future, the role of the DVR is expected to evolve into a gateway device that
provides consumers with a myriad of new services.
- Digital Set-top Boxes (STBs) - Digital STBs continue to be the fastest
growing digital device in the consumer electronics marketplace. The digital
STB represents a huge revenue-earning potential for content owners, service
providers, equipment manufacturers, and consumer-oriented semiconductor
vendors. The STB, once a relatively passive device, is now capable of
handling traditional computing and DVR functionality. The introduction of
DVR-integrated STBs is particularly important for cable and satellite
operators who want to differentiate their services and platforms from
- Other candidates - Gaming consoles and utility metering devices provide an imaginative and interesting platform to evolve into potential home gateways. Companies like Microsoft and Sony believe that high-speed gaming consoles could serve as the gateway or hub to the digital home of the future.
Some of the likeliest supporters of the RG concept come from the energy-utility industry. In fact, the concept of a utility company installing new gateway devices in their customers' homes to provide new services has been gaining ground since the early 1990s.
Shared Storage Components
All the digital consumer devices vying for a piece of the RG market have one thing in common: they all require storage components that are reliable, high-performance, low cost and consume low power. Storage is a major growth area for as vendors try to fit applications and data, increasingly large in size and number, into smaller packages, while keeping costs and power requirements down. Storage for RGs can be split into solid-state (based on flash semiconductors with no moving parts) and mechanical hard disk drives and come in a variety of standards, shapes, and sizes.
It is a known fact that the life span of a hard disk is shorter than that of the average consumer electronics device. Hard disk durability is inherently limited by its internal moving parts, especially when coupled with the typically constrained environment of RG and consumer electronics device enclosures. Add to this the cost limitations of electronic devices and it becomes clear why many hardware designers are turning to solid-state flash products to meet the storage needs of next generation RGs. The total cost of ownership to maintain hard disk drives in homes is sometimes dangerously overlooked: high gateway return rates due to disk failure can seriously damage the business model of operators who are delivering electronic services to residential homes. It is worth noting that some set top box manufacturers report between 2-8% of PVR hard disk drive annual failure rates. There is no reason to expect lower rates here.
Solid-state flash solutions are generally associated with AMD, Intel, M-Systems, Toshiba and Samsung. Typical solid-state flash capacities range from 1MByte (8Mbits) to 128MBytes (1Gbit).
NAND and NOR Flash
NOR and NAND technologies dominate today's solid-state flash memory market. Both technologies have unique features and are aimed at fulfilling different market needs. The designer should carefully weigh the options when using flash memory in a next generation RG design. Further investigation of both technologies reveals a number of interesting points.
First, NOR is typically used for code storage and execution purposes only, and is thus primarily used in simple digital devices such as low-end cell phones or simple cable modems without any real gateway aspirations. NAND, on the other hand, is often used for both code and data storage in devices such as set top boxes, MP3 players, digital cameras, high-end cell phones and RGs. However, raw NAND cannot replace NOR unless coupled with a separate boot ROM and a controller to overcome NAND's inherent data integrity flaws, or special software to manage and correct data errors. (Remember, NAND was originally intended for use with media files, such as JPEGs and MP3s, where an occasionally flipped bit does not really matter).
From a performance perspective, NOR is optimized for reading data but lags behind NAND when it comes to writing and erasing data. This often disqualifies the use of NOR in devices such as residential gateways, devices that require file manipulation capabilities. Typically, NAND outperforms NOR in such operations by orders of magnitude, another fact that should not be overlooked.
By examining the physical architecture of both technologies, it becomes clear why NAND offers higher densities with more capacity on a given die size, thus making its cost structure far more attractive. The key to achieving higher capacity ranges is based on the fact that a NAND cell size is almost half the size of a NOR cell. This is true even when comparing single-level cell NAND technology with multi-level cell NOR. (Note: Multi-level cell NOR technology still suffers from some very serious bit-flipping phenomena and data errors, and should always be used with a robust error detection and correction mechanism.)
This, in combination with a simpler production process, enables manufacturers to build NAND products with a capacity range of 8MBytes to 128MBytes. Therefore, more and more manufacturers are opting in favor of NAND products to meet the BOM cost target of the providers. However, designers must then face the difficult task of "taming" NAND to perform reliably and in the super-stable fashion necessary to support the operation of such a utility-critical device as the RG. How can this be done?
RG designers use three mechanisms to control NAND functionality:
1. An external hardware controller - In addition to the NAND silicon itself, a controller is required to interact with the actual flash memory array to provide the functionality of a mechanical hard drive on a solid-state silicon chip. The performance level for this type of NAND management system is relatively high; however, the cost structure associated with its implementation is also high, and its overall endurance and reliability is questionable at best. What suits some removable storage cards sold in retail stores falls far short of the design requirements of an embedded product.
2. Management software - Typically, a NAND flash management software system uses very intricate code that runs on the RG host CPU. Although the low cost of implementation is enticing, severe performance and reliability issues plague this type of solution, making it a problematic design choice. The fact that every change in OS or flash capacity carries with its hidden costs, due in part to the need for new drivers and algorithms, makes this choice even less attractive.
3. A mixed balance of hardware and software - A hybrid design such as M-Systems' DiskOnChip uses a balanced combination of hardware, embedded into the same silicon die as the flash bank itself (with a very small overhead as compared with option #1), and a very robust file system software developed and debugged over the course of several years (which, by the way, in M-Systems' case, is distributed free of charge). Being the only NAND-based flash that enables a system boot without necessitating a separate boot ROM device, this RG building block is highly cost-effective. A hybrid approach is the preferred solution for RG designers due to its high reliability and performance levels, in combination with its very attractive price tag. Although initial costs are slightly higher than raw NAND flash (but still far more cost effective than any NOR flash component), the hidden costs associated with mishandled NAND data and reliability issues (such as truck-rolls, recalls and high return rates) makes a hybrid design far less costly in the long run.
When considering which method is appropriate to control and operate raw NAND flash, the separate and incompatible paths Toshiba and Samsung plan to take over the next few years become an issue. Toshiba is moving toward multi-level cell flash technology that will drive the cost of NAND down even further, while Samsung is placing its bet on more efficient and advanced single-level cell processes than those it employs today. This incompatibility further increases the value of the direct hybrid approach. Let those who truly understand flash deal with these issues, rather than wasting design resources on second-guessing the market and technology.
What about the "right capacity"? What is the winning amount of memory that a designer should provide in a RG? The market has seen many different figures, ranging from 1MByte to 64MByte in flash, to Gigabytes of hard disk drive storage. While the need for a hard disk in some designs is understandable - strictly for mass media storage purposes (cleverly coupled with reliable flash storage for the software code, of course, to prevent a mechanically "dead" gateway), it is difficult to justify or understand designers who introduce a new design nowadays with 1, 2 or even 4MBytes of flash.
This minimalist, procurement-driven 1-4MByte design carries no added marketing value whatsoever, apart from a slightly lower price (and even this is no longer the case, as these designs usually use the more expensive NOR flash when they could be enjoying higher DiskOnChip or NAND capacities). In fact, such shortsighted BOM-based designs neglect to deliver any promise for new and exciting services critical for success in attracting new consumers, or the ISPs to deploy them. Who needs a design as such? Why not settle for a standard wireless router, or a "plain vanilla" cable or DSL modem? Where is the much-needed marketing differentiation?
A true RG must provide plenty of room for the ISPs to deliver a myriad of services to the subscribers - some, as mentioned earlier, yet unknown. "Plenty of room" means a minimal flash capacity of 8MBytes to 16MBytes for today's designs, and 32MBytes to 64MBytes for 2003 designs. (Remember: 16MBytes to 32MBytes of NAND flash in single-level cell technology, and even more in multi-level cell technology, costs about the same as 4MBytes to 8MBytes of NOR flash.)
Increasingly, consumers will want plenty of versatility from their RG systems. Instead of receiving TV signals via a STB, recording a movie on their DVR or using a PC to connect to the Internet, they will expect one device to enable them to access all these services. Because the various types of gateways currently on the market use similar electronic components, their expectations may soon be met. The importance of a reliable and high capacity, low cost and fast solid-state storage subsystem for such a converged device is paramount to its efficient operation. Two competing technologies vie for the billions of dollars generated by the RG storage industry - NAND and NOR. It is becoming increasingly evident that NAND will eventually win, due to its superior cell cost structure and efficient production processes. As NAND becomes the flash storage technology of choice for RG hardware designs, overcoming raw NAND's inherent flaws to ensure a future-proof, long-lasting RG design has become the major challenge. M-Systems' DiskOnChip is currently the only viable flash storage solution capable of delivering the advantages of NAND - cost, reliability and performance - with an easy NOR-like integration effort, a smooth migration path, and a robust and reliable flash file system to secure the design.
Consumers will judge the gateways we are designing today on their fast and reliable operation, which is greatly affected by the local storage used. Thus, as we embark on the long journey towards successful deployment of RG design, it is essential to employ the fastest, most reliable, and most cost-effective local storage solution.
Gerard O'Driscoll (32) is Chairman and Chief Executive Officer of DigitalMentors, a leading provider of e-learning solutions to the global digital television and home networking industries. O'Driscoll graduated from the University of Limerick in 1992. After college O'Driscoll held senior management positions with Seagate, Monaghan Poultry Products, and Irish Productivity Center. O'Driscoll subsequently joined Chorus, an Irish cable company, as a senior technology strategist. O'Driscoll is a published author of two best selling books on topics ranging from digital set-top boxes and interactive TV to home networking technologies. Additionally he has been a featured speaker at a wide range of conferences and events on e-Learning and home networking technologies.