We sometimes install our equipment, often in multiples, in tough environments that include corrosives, humidity and intermittent vibration. Our normal approach to operator interface is ruggedized industrial PCs where we store programs, some operating data, recipes, operating manuals and schematics. Now there's revived interest in doing this via thin client. In some cases, we even would have use of the customer's server. What do we really need to know to make a good decision about staying with what we know or switching?
- from May '10 Control Design
Understand the Licenses
Thin clients offer many benefits. They can be deployed as fixed devices on large or long machines as a distributed architecture, or they can be deployed on mobile handheld devices that make the operator more efficient. The latter method also can reduce hardware costs by keeping the number of required displays to a minimum.
Thin-client hardware often can be lower-horsepower PCs that can be diskless/fanless and offer more long-term reliability in harsh environments.
Evaluate how the thin clients are licensed. If the licenses for multiple thin clients are stored on the server—not on each individual client—when one goes down, nothing special has to be done to the thin client to get it or a replacement running again. Connect the power cable, connect the network cable, and you are running again.
There are other considerations. Will the HMI/SCADA software run on a current server operating system? Do the thin clients support security that is authenticated back at the server? Can each thin client view and interact with screens independent of what the server is viewing? Does it use standard Web-browser technology, or is something special needed? Is redundancy supported?
Working with the IT department is important, too. If the application will be run on a server that is most likely managed by an IT department, is the software flexible enough to adjust to the open TCP/IP ports that will be dictated by the IT policy, and not by the HMI/SCADA software?
Our thin-client solution is more optimized than the traditional terminal thin-client solution because the communication between the thin clients and the server is implemented by exception—only when values displayed on one station change in other station and the thin client is able to execute algorithm locally, resulting in a high level of flexibility and scalability to the system.
Verify how the thin client's communication is initiated. Does the communication occur all of the time to all workstations on the network, or does it happen only by exception and determined locally on thin client? This optimizes communications, reduces overall network traffic and allows the system to be very scalable.
Fabio Terezinho, Product Manager,
Safer Storage Space
The line between industrial PCs (IPCs) and thin clients is blurring, as the latest IPCs add fanless, energy-efficient CPUs, solid-state drives and embedded operating systems.
It becomes more a matter of application, rather than selecting a different PC/thin-client form factor. Today's IPCs easily exchange data to central servers if the application calls for it.
Cost-efficient, non-rotating media can replace traditional hard-disk drives (HDDs) in the IPC, eliminating any concerns about shock and vibration issues. The latest solid-state drives (SSDs) offer considerably more storage space than HDDs from just a few years ago; thus, they can handle even larger applications with ease.
Bjoern Falke, Product Marketing Lead Specialist,
Phoenix Contact, www.phoenixcontact.com
With the use of thin clients today, the network becomes more and more important. When you use thin clients, all critical tasks now run on remote servers generally in a more friendly, temperature-controlled environment. Redundancy of your network itself should be taken very seriously. Implementing rapid spanning tree protocol (RSTP), proprietary ring or chain redundancy protocol technologies should be the very least that is done to ensure smooth computing on your new thin-client installations.
Jim Toepper, Product Manager,
Industrial Ethernet infrastructure,
Consider I/O and OS
To make an informed decision, it is important to consider that there are two basic approaches to deploying thin-client solutions. The first uses a central core server or cluster of servers running a terminal services environment that allows multiple thin clients to log in to the server and start an application set. Resources are shared among all clients logged in to the server. Should one of the terminal sessions experience a fault, it potentially could affect every thin-client terminal session running on the server.
The second approach involves running multiple virtual machines on the central core server or cluster of servers. Each user logs in to a virtual operating system (OS) environment that uses a unique, dedicated set of resources. In this case, when one thin-client session experiences a fault, it is unlikely to affect any of the others attached to the server. A virtualized approach most closely represents a distributed-client-computing environment in that each operator station still runs in its own environment. However, a virtualized solution usually is more expensive and more complicated to deploy than the traditional terminal services thin-client model.
Which I/O does the operator interface need to serve its purpose? Most industrial PCs offer great expansion capabilities to accommodate standards-based and proprietary I/O. A thin client might not support the various I/O that the industrial computer system was designed to accommodate. I/O functions can move to the central server, but there are limitations. The server must physically be able to accommodate the interface cards required in the existing solution, and device drivers that are compatible with the server's OS must exist. This is especially true when implementing a virtualized thin-client approach, as many of the virtualized platforms available today do not support the non-commercial I/O often used in industrial control environments.
Which operating system is currently running on the industrial PC? When using terminal services, the OS available to the thin client is that of the server. If the current industrial PCs run a real-time operating system (RTOS), this approach might not be an option. A virtualized thin-client environment will accommodate many of today's mainstream operating systems, but RTOS installations can present challenges in a virtualized solution, as well.
How important is reliability? Reliability can be improved or degraded by moving to a thin-client approach. The centralized server-client model means that every operator station is dependent on that one server. If that server goes down, it likely will compromise the entire control network. If complete system downtime cannot be risked, a secondary server might be required. Reliability then can be built into the solution using redundant central servers designed with one of several failover methodologies—for example, active/active, active/passive. The introduction of the redundant server dramatically increases system uptime, but it does so at a substantial cost—that of another central server.
Does the size of the control system allow cost savings to be realized? In most server-client networks, the hardware and software required to implement the central server or cluster of servers usually represents the bulk of the cost. If a large number of thin clients use the solution, that cost is distributed, driving down the total cost per client considerably. However, if the same solution is deployed to a handful of clients, the per-client costs can be much higher. In a network with few operator interfaces, the existing industrial PC solution may prove to be less expensive than a thin-client approach.
Jonathan Couch, Lead Computer Design Engineer,
Keep It Modular
Thin-client architecture makes sense with about four or more client stations. At this quantity, and with a willing user IT group, you can make an easy economic justification using standard terminal services or remote desktop protocol (RDP), or better yet with the industrial shell solutions from ACP. Beyond the initial economic justification, consider the long-term benefits to your customer.
A split architecture uses a modular touchscreen display independent from the PC processor or thin client. This layout already makes plenty of sense in most industrial applications, but even more so in extreme environments.
Quality industrial touchscreens still are somewhat costly, but true industrial-grade monitors will operate for many years in the worst of environments. These are technology-stable devices and supported through several backlamp changes over 10 years or more without obsolescence. Unfortunately, PC processors have a more checkered quality history and clearly lack technology longevity. The more harsh the conditions, the more troublesome the PC becomes. Considering it's the low dollar side of the equation anyway, best to keep it independent of the more reliable display. Thin clients allow the processor engine to be further reduced with only the most simple operating shell and basic CPU left in the harsh environment, and in particular no rotating hard disk is needed.
The actual economic value of the PC or thin-client hardware really is only hundreds of dollars. Keeping it to a minimum and modular allows for quick and inexpensive changes, hence leaving the more costly and durable display technology to soldier on. What a sin if a customer must discard a costly integrated display/PC system just due to hard drive failure or technical obsolecence of the CPU.
HMI Business Manager,
Safe and Secure
There are a number of possible solutions when considering a thin-client interface. The most-often-used solution is a Web-browser interface that accesses a Web-based application hosted in a secure location. The chief advantage of a thin-client interface is that it will allow you to store your data in a secure and safe location while providing remote access to it. If your application involves several systems that must be deployed, thin-client access has the additional advantage of allowing multi-user access to a centrally managed data repository. Data updates become much easier to implement.
The disadvantage of a thin-client system is that a reliable connection to the server is a must. If you lose connectivity to the host, then you have no access to the data. Another possible disadvantage is that the technical skills necessary to implement and maintain a centralized, thin-client-based system, vs. one stand-alone system, are quite different. The cost benefit of thin client really depends on the value of the advantages to your organization. If you can use a customer's hardware server, you could achieve a significant cost reduction because the computational power of the thin-client hardware is minimal.
Tony Martin, Director of R&D,
Sealevel Systems, www.sealevel.com
Our Embedded Intelligence columnist Jeremy Pollard wrote about the application of thin clients in an industrial environment in a recent column. Read what he had to say at www.ControlDesign.com/thinclients.