The system must be bulletproof because there's not enough IT staff to maintain it. Have you heard those words before? If you're a manufacturer looking to update your server and storage systems to accommodate plant-wide applications, you very likely have.
In fact, that's the premise upon which Greg Scheidemantel, plant manager at Sika Sarnafil Inc., had to choose the server that would run the company's warehouse management system. The Swiss-owned company relies on its Europe-based IT group to maintain the corporate SAP implementation. That centralized staff is supported by remote IT in other parts of the world, which includes one person at Scheidemantel's Canton, MA, industrial roofing plant.
That one IT person, however, is responsible for supporting 250 employees across the United States. So, as Scheidemantel began building out the company's warehouse management system earlier this year, the worry was not so much the application, but the hardware underneath.
The server would run an operation that pulls data from the plant floor in Canton and interfaces it with the SAP system located overseas. Given the far-flung IT operations, there was not much room for error. With that in mind, Scheidemantel's emphasis while server shopping was on uptime and redundancy for overall system reliability.
"We wanted this to be a lights-off system," Scheidemantel says. "We have one person for our IT infrastructure here ... and his main responsibility is the upkeep of the network in the U.S. If possible, we wanted this to have little to no impact on him."
Last year, Scheidemantel built a plant-floor data collection system that culls control information from programmable logic controllers (PLCs) and a variety of instrumentation. Using SAP AG's xMII technology, that information is analyzed and sent to the enterprise resource planning (ERP) application. This year, a warehouse management data collection application was put in place. It, too, relies on xMII to interface the data captured by a handheld scanner with SAP. Before this automated application was in place, everything was done manually.
The good news is that the xMII technology will optimize operations. The bad news is that it adds a new layer of complexity to the IT infrastructure. To mitigate risk, Scheidemantel invested in a Stratus Technologies server built specifically for complex manufacturing environments.
"We needed a system we could consider bulletproof so we didn't lose any data," Scheidemantel says. "In the research we did, Stratus came up as No. 1 in reliability."
Most manufacturers face a similar scenario. As they tap into more tools to help crunch numbers and measure key performance indicators (KPIs) that tie back to the ERP system, the need for reliability becomes paramount. To that end, sophisticated software that lets companies dig deep into the factory floor needs hardware to run it that is just as sophisticated.
While servers are getting faster and cheaper every year — Moore's Law says that processing power doubles every 18 to 24 months — hardware vendors are shifting their focus to build platforms that are not only fast, but also reliable. In addition, some server vendors, such as Stratus and IBM, are targeting manufacturers with systems that can support complex applications in new ways at the lowest cost.
Storage products, which go hand-in-hand with servers, are also evolving as vendors — for example, EMC Corp. and Hitachi Data Systems — add intelligence to what once was a dumb data domain that did nothing more than house the data. Concepts such as virtualization — blending a pool of servers and storage into a single array of processing power — and new pay-as-you-go models are emerging.
The ability to intermingle hardware systems is partially the result of smaller footprints that pack just as much punch — or more — as the old mainframe or refrigerator-sized storage units. But it also requires new layers of integration software to keep everything interoperable.
"In computing, everything over time gets physically smaller. That's inevitable," says John Webster, principal IT adviser at Illuminata, an analyst firm. But now the trend is to move servers and storage into a tightly integrated architecture, he says.
For example, blade servers let multiple processing boards run in the same physical server box. And virtualization technology from VMware Inc., an EMC company, is an abstraction layer that decouples the operating system (OS) from the hardware, thereby allowing multiple virtual machines with heterogeneous OSs to run side-by-side in the same machine.
The benefits of such technology come in the form of energy savings and easier administration for the IT team, according to Webster. However there's still a bit of unease for users grappling with whether to move down the path of mix-and-match hardware.
"The big area of concern for users is how much mix and match you can actually do," says Webster, who notes that there's a word fraught with emotion in the storage industry: "interoperability."
Control Out-of-the-Box
To alleviate integration headaches, Stratus and IBM are doing the work for their customers. Stratus has formed application partnerships with SAP and most of the major automation control vendors, including Emerson Process Management, GE Fanuc Automation, Rockwell Automation, Siemens AG, and Yokogawa, which test and validate their software on the Stratus platform.
The importance of validating software to run on hardware can't be underestimated, especially in mission-critical environments — for example, manufacturing.
"You would think that if you bought either a Dell box with standard Windows running or a Hewlett-Packard [server], that all the applications would run, no problem. But that's not really true," says Frank Hill, director of manufacturing sales and the SAP alliance at Stratus.
Within every standard PC environment there's always special code running the vendors' hardware platforms. So a Dell box and an HP box are not exactly the same, he says. "They're close," Hill explains, "and most of the time, things will run straight across [the different servers], but for critical applications that run the plant, people want a fully tested, fully supported system."
With the proliferation of manufacturing execution systems (MES) to manage and schedule work in the plant, as well as the prevalence of SCADA tools and data historians for capturing data that enterprise systems, such as SAP, need to tap, the plant has to build complex IT infrastructures that never were needed before. That means IT departments have an additional responsibility. But as in Sika Sarnafil's case, that isn't always a practical proposition.
The Stratus server, priced from $10,000, was designed for these scenarios, as it duplicates every component in the system, Hill says. The processor, the memory, the disk drive — eventually something will fail. But Stratus servers are designed to not only duplicate components within the same box, but also interlock them so that the first unit mirrors the second. In the event of a failure in the first server, the second system takes over without any downtime and — more important — without losing data.
In addition, this self-monitoring system will "phone home" to the Stratus office via a modem or Internet connection to let the technicians know which component failed. The next day, perhaps even unbeknownst to the plant manager, the component will arrive at the manufacturer's site.
Similarly, IBM's System i, is designed for fault tolerance, and the operating system, database, middleware and security — all of which come from IBM — are fully tested, validated, and integrated (the "i" stands for "integration").
"The point of System i is to have fewer things to manage," says Ian Jarman, IBM's System iSeries product manager.
But while Stratus focuses on plant floor applications, IBM dedicates its system to the enterprise to run ERP applications from SAP, Oracle Corp., Infor, and other vendors. "We are the enterprise system that is linking all of the data in terms of financial systems, supply chains, and so on," Jarman says.
That's not to say, however, that IBM does not have a model that makes sense for manufacturers looking to leverage the full power of a System i Power5 processor at an affordable entry point. The System i Express 525 is a mid-size model that starts at $30,000 for 30 users. In April, IBM changed the pricing model for System i Express servers — designed for small and mid-size business — to charge by the user rather than by capacity. The system can scale to a 64-processor solution — which is a lot of processing power. But IBM does not charge for the computing power.
"Now we give customers the full performance of the machine, connecting the value and price with the number of users," Jarman says.
On the storage front, a similar pricing model is emerging. Hitachi Data Systems comes at storage "not just in terms of capacity, but in term of performance," says Claus Mikkelsen, Hitachi's chief scientist. The company uses a technique called "thin provisioning" that allows space to be easily assigned to servers on a just-enough and just-in-time basis. This ability to configure the storage to the application can be applied to Hitachi's Universal Storage Platform, which uses virtualization to pool storage from other systems, be they EMC, HP, IBM, or Sun, into one single large array. "It saves on software licensing and maintenance, and improves total cost of ownership by 30%," Mikkelsen says.
Prices Coming Down
Storage, like servers, is not as expensive as it used to be. Pricing on a mid-range product from Hitachi can start at $10,000, while enterprise storage starts at roughly $90,000, which, Mikkelsen says, "is not so unthinkable," considering the enormous amounts of data generated.
Indeed, there's a huge reservoir of data flowing through most organizations. According to Illuminata's Webster, data grows at a compounded rate of 60% to 70% per year. For regulatory reasons, that data often must be saved, so the amount of storage required becomes an important question. There are low-cost storage alternatives, such as tape, but for the real-time organization, data must be retrievable in real time as well.
"Sooner or later, you realize it has to come together at some focal point," Illuminata's Webster says. "Constructing that focal point, particularly in large organizations, is a very difficult thing to do."
Moreover, even with redundancy and virtualization techniques, the storage industry tends not to think in real time, Webster says. But manufacturers working with control applications or radio-frequency identification (RFID), for example, must address the idea of real-time storage.
The issue is top-of-mind, Webster says. "There's an upsurge in interest in solid-state storage, which is essentially silicon that looks like a disk drive." SSD lives in a land between disk and processor memory, he says. This setup could be the catalyst that moves storage and servers back together, whereas today, external storage capacity is the norm.
Given the evolution of both hardware platforms into a smaller, more powerful footprint, that goal is attainable. IBM, for example, is a one-stop shop for servers and storage. IBM's DS4700, which at about 6 inches tall, can expand to 112 Serial ATA (SATA) disk drives for a total capacity of up to 84 terabytes. That compares with the disk systems in the early 1980s that were as big as a refrigerator and held 5 megabytes. And while the IBM storage system can work with non-IBM servers, many manufacturers may see the value in a tightly coupled solution.
"When you look at IBM, you are looking at the whole package," says Harold Pike, IBM's midrange products offerings manager. "From what I know from my manufacturing days, this is critical. There is so much price competition and everyone is lean and mean, [but manufacturers] don't have time anymore to pick and choose different pieces from different vendors and put it all together."
Indeed, having neither the time nor the staff to maintain the hardware platform was what drove Sika Sarnafil's Scheidemantel to the Stratus server. With all efforts focused on increasing operational efficiency and growing the business, the server is the last thing that should get in the way.
"Within our company, there's been a growing trend toward transparency," Scheidemantel says. "Out of that effort has grown the ability to monitor the [factory] floor more efficiently and automatically." And the hardware systems, he says, can't get in the way of that.