Clou Leve By Kennet Cloud C possibil Mainten quickly choose Cloud co computer profession can help unpreced transpare computin is already operation performa this tech maintena What i Cloud co software, used, and confusion company term is us At least o Institute o “ p se se What doe available d Co ragin h Piety, Ph.D omputing ities in th ance (Pd realized to leverag mputing is world, als do not u their oper ented levels ncy for main g the way of available t use cloud nce? Can ma nology and nce industry? s Cloud C mputing is hardware, a rarely expla is due in pa has its own ed in so man ne organizat f Standards a Cloud Compu ool of config rvices) that rvice provide s this mean over the Inte mput g Te . and David G is openi e world M). The by com e this tech the latest but man nderstand w ation. “The of efficie tenance pro the future, o oday? How computing intenance p revolutio omputing a hot topic nd Web com ined, that it rt because t spin on what y different co ion has put a nd Technolo ting is a mo urable comp can be rapi r interaction ? In laymen rnet to anyo ing - chnol eswein, BS, ng up ma of Predict benefits panies w nology. buzzword in y mainte hat it is or Cloud” pr ncy, uptime fessionals. I r is it someth can a main to improv rofessionals nize the p ? these days pany is usin leaves many he term “clo cloud comp ntexts, it is d definition t gy (NIST): del for enab uting resour dly provision .”1 ’s terms, clo ne with a n Anot ogy? M.E. ny ive are ho the nance how it omises , and s cloud ing that tenance e plant leverage redictive . Every IT, g this buzzw non‐comput ud computin uting is and ifficult to def o the term c ling convenie ces (e.g., ne ed and rele ud computi etwork conn her B ord in its ma er profession g” can mea how it can b ine. loud comput nt, on dema tworks, serv ased with m ng is a way ection. These uzzw rketing. The als scratchin n many diffe enefit custom ing. Accordin nd network a ers, storage, inimal mana to make co resources c ord term is so w g their heads rent things. ers. Becaus g to the Na ccess to a s applications gement effo mputer reso an come in or idely . The Every e the tional hared , and rt or urces many 2 forms. The resource could be a server, a network, or software as long as the intended user can quickly and easily gain access over the Internet. Typically, these resources are provided by a vendor on a subscription or rental basis without any capital expenditures necessary on the part of the client. NIST further specifies three service models for cloud computing:2 • Infrastructure as a Service (IaaS) • Platform as a Service (PaaS) • Software as a Service (SaaS) The first two models are only used by computer professionals and are beyond the scope of this paper. For the purposes of this paper, wherever the term cloud computing is used, the SaaS model is assumed. What is Software as a Service (SaaS)? With the Software as a Service (SaaS) model, the client pays a subscription to use software. All other tasks associated with maintaining the software are handled by the provider. These tasks include software updates, data backups, server maintenance, etc. The SaaS model is the most commonly encountered by non‐computer professionals and offers the most benefits and cost savings. The term “cloud” comes from the idea that the user does not see or touch the physical equipment that powers a cloud‐computing system. So, what is a cloud‐computing system made of? Today, large server farms power many web‐based software systems. A server farm is multiple connected servers that function as a large “super computer.” The server farm can instantly scale by powering up or down servers as needed without interruption to service. Multiple software systems can run on a single farm. As more computing power is needed more servers are added to the farm without downtime. During low‐use periods like nights or weekends, many of the servers in the farm can be powered off to conserve energy. This also allows maintenance activities to take place without impact to the user. As demand increases, more servers are brought online to meet the users’ needs. Cloud computing is one of the few computer innovations to start at the consumer level and then migrate to enterprises.2 “Most IT innovations start in the enterprise and go to customer. This one is going to enterprise from consumer.” — Geoffrey Moore, TCG Advisors.2 Some of the oldest, most well‐known, cloud computing systems are consumer products. Google Gmail, Facebook, Windows Live Hotmail, and YouTube are all examples of cloud computing, more specifically SaaS, in the consumer market. Hotmail, one of the first Web‐based email systems, officially launched July 4, 1996, symbolizing freedom from client‐ISP based email and the ability to access a user’s inbox from anywhere in the world.6 By February of 2009, Hotmail reported over 30 million users worldwide.6 Today, both private and public institutions are making a similar switch. The City of Los Angeles now uses Google Apps, an integrated email and document system by Google, for its 30,000 employees.4 Los Angeles joins the ranks of many other notable institutions using Google Apps for their email, including the University of Notre Dame, Motorola Mobile Devices Division, and Arizona State University. A host of cloud‐based business systems are available on the market. There are cloud‐based alternatives to most 3 any traditional client‐server software systems. Salesforce.com, a Customer Relationship Management (CRM) system using the SaaS model, has become one of the most popular CRM systems available today. According to Gartner, a leading IT research firm, 30% of new customer service and support application investments will be through the SaaS model.3 It is likely that other business systems will follow the same evolution. Advantages of Cloud Computing The most noticeable and immediate benefit of cloud computing is the elimination of capital costs. Without on‐site servers and expanded IT infrastructure, the up‐front capital investment is often zero dollars. This is a huge advantage when making a business case for a cloud‐based system. In addition, the ongoing operating costs are also reduced greatly. There are a wide range of savings estimates, some of which are much more optimistic than others, spanning from 39% to 90% operating savings.4 There are many well‐documented cases demonstrating a combined implementation and operating cost savings in the 20% to 40% range.7 An analysis by the City of Los Angeles in its move to Google Apps found five‐year costs were reduced by 23.6%, saving over $5.4 million.4 These operating cost reductions are a direct result of increased efficiencies. In many companies, servers run at 15% to 20% of their capacity.5 A cloud‐provider can easily consolidate the usage of a system from many companies to a set of servers that is sized to match the workload. A cloud provider can also control the load on their servers even as demand fluctuates. By sharing fewer computing resources among a larger group of users, operational efficiency is increased. In addition, the manpower to maintain the systems is utilized more efficiently. With traditional client‐server systems, maintenance and support activities may require 5% to 10% of one IT person’s time. A cloud provider can utilize almost 100% of their IT person’s time, which creates an increase in manpower utilization and efficiency. Granted, with a traditional system, the IT person would have other tasks to perform when they are not supporting the system. With cloud computing, multi‐tasking is reduced and the IT department is free to focus on more important activities. In addition to the obvious and direct cost savings, there are other indirect savings and benefits associated with cloud computing. For example, most cloud computing vendors perform maintenance upgrades routinely with zero downtime. This means that users receive the benefits of new features and enhancements without interruption to operations. In addition, cloud providers have many fail‐safes in place for power, network, and servers. Because of this, many providers guarantee a level of uptime that is difficult to achieve with a single, client‐owned server. All of this adds up to more productivity for the users. Another benefit of cloud computing is ubiquity. A cloud system is available from anywhere that has an Internet connection. Some systems are even available via mobile phone or iPod. Because the same, high security standards are imposed no matter where the users are connecting from, there is little risk of exposing the system and data to attack or theft. Other risks are also minimized by using the cloud. Costs and performance are predictable because pricing and Service Level Agreements (SLA’s) are written into the contract. There will not be any unforeseen costs associated with hardware failure or data recovery services. The vendor bears the burden of these costs and is better protected against these hazards in general. Even if a hard drive fails, the system will failover to a redundant data drive. The cloud provider would then hot‐swap a new hard‐ 4 drive without impact to the customer. All of this is driven by the vendor’s need to meet their contractual obligations. How Cloud Computing Applies to the Predictive Maintenance Industry Drawing on the NIST definition, a fourth model can be defined specifically for the Predictive Maintenance (PdM) industry: Expertise as a Service (EaaS). In this model, expert diagnostic knowledge is delivered via the cloud. For the PdM industry, those reaping the most cloud computing benefits are combining EaaS with SaaS. To do this, vendors are providing the hardware and software needed for their solution via the Internet (SaaS) and then leveraging their own setup to also deliver diagnostic expertise (EaaS) on a much larger scale than ever possible. This combination is enabling vendors to maximize the benefit and efficiency of their cloud based PdM program by delivering highly‐skilled diagnostic knowledge to a wide audience without any physical travel by analysts. This combination of models lets maintenance staff and analysts access the PdM software over the Web and then perform whatever tasks are relevant to them. While the analyst accesses the software to review data to detect faults, the plant manager accesses it to see the analyst’s recommendations and assess the health of his entire fleet based on the latest data. The success of any PdM program rests on the ability of an analyst to make the correct interpretation of volumes of vibration data. In general, it takes about two years for a new analyst to hone his or her skills to be reasonably proficient at diagnosing machine faults. It is common for in‐house analyst positions to have high turnover as proficient analyst are promoted or seek better job opportunities. As a replacement is trained, there is a two‐year cycle in which the diagnostic accuracy of the predictive maintenance program suffers. In addition, it is well documented that today’s highly skilled analysts are retiring at a steady rate and there are few replacements coming up the ranks from the next generation. Cloud‐based PdM programs enable plants to access highly skilled analysts provided by the vendor. It brings the data to the analyst rather than the analyst to the machine. Plants need only train their maintenance staff to collect vibration data and upload it to the vendor’s secure cloud‐based system. While there is a slight learning curve, anyone familiar with the machines can be taught to collect and upload data with just a few days or weeks of training. The analyst can then view this data via the Internet, thus eliminating the need for costly travel. It also enables one analyst to serve multiple customers since all customer data can be reviewed securely from a centralized Web‐based system. Further, the analyst can review and diagnose from anywhere in the world, as long as there is an Internet connection. Another huge benefit to the EaaS model relates to delivering the right skills for the problem at hand. An in‐house analyst’s skills are limited to the types of equipment and problems with which he or she has had experience. By delivering expertise via the cloud, a vendor with a stable of analysts can deliver the person with the expertise that matches the customer’s need. Revolutionizing PdM The predictive maintenance industry stands to gain more from the cloud revolution than most any other industry. To date, PdM systems have been primarily isolated to individual laptop and desktop 5 computers. PdM was completely left behind in the client‐server boom of the 90s. By moving to the cloud, PdM systems will reap all of the cloud advantages previously outlined in addition to all of the client‐server advantages that these systems have yet to achieve. The move to the cloud will also be easier than for other industries since PdM has not made a large investment in traditional client‐server systems. Besides the general benefits outlined earlier, moving PdM to the cloud will revolutionize the industry in several critical areas: • Centralization. Data is leveraged more efficiently. If a person needs to determine which plant in a fleet is at the highest risk for downtime, it can be done quickly because the data is on a single system. The person does not need to talk to several other people or access many different systems to know the health of the fleet. • Collaboration. Multiple people can share the same information. Several experts can be called upon to examine a potential problem. By working together and drawing upon a larger set of experience, these experts can determine the best possible course of action for their customers. • Communication. Because email alerts, text messages, and even automated phone calls are possible, the profile of PdM within an organization is raised. • Transparency. More people in your organization are exposed to the PdM process and can clearly see the benefits. This raises the value of the program, making PdM a “need” instead of a “want.” • Integration. With so many people needing information from the PdM program, automated data exchange between business systems becomes a reality. • Participation. The door will be opened for others to gather data, contribute knowledge, report observations, and feedback findings to the system. Instead of a one‐way process, PdM becomes an interactive, ongoing process with many other departments feeding data back to the system. All of this may sound like predictions of a distant future. Many might question whether all of this is even possible. If so, then the future is here. Cloud‐based PdM systems have been quietly growing for more than seven years. Today, more than 50 companies (many Fortune 500) already rely on cloud‐based PdM services. In fact, there are already many success stories from migrating PdM to the cloud. Case History 1: Equipment in Danger saved by “The Cloud” A vibration analyst was asked to perform a vibration analysis on a centrifugal compressor. The compressor was alarming on the Stage 2 vibration sensor that is part of the control system. With a traditional vibration service program, the vibration analyst would travel to the site, gather the vibration data, analyze the data and then issue a report. In this case, it would have taken the analyst a minimum of eight hours to physically get to the plant. Luckily, th implemen the noisy already at to any au the data a easily acc Working t impeller, “in the cl short tim repair on and/or a w Case H A custom monitorin and alerts acquisitio the techn is a need primary a program with Inter is was not a ted using a machine him the site perf thorized use nd called in ess the same ogether, the but with its oud,” the an e after the o the wrong p rong diagno istory 2: V er at a larg g program. T the technic n for each m ician contact for immedia nalyst was aw was impleme net access (a traditional v cloud compu self and uplo orming othe r from a cen a colleague data and dia two highly s associated oi alysts were riginal reque art of the m sis would ha acationi e power pla he plant use ian to a pot achine. The t s the assigne te action. U ay on vacat nted using c nd credentia ibration mon ting system, ad it to a se r maintenanc tralized, secu for a closer l gnostic tools killed analyst l pump. Beca able to quic st. A travel achine, a day ve been very ng Analy nt was gat d data collec ential proble echnician re d vibration a sually this w ion. Luckily, a loud comput ls on the sy 6 itoring prog the custome cure data cen e activities. T re cloud‐bas ook. The coll without any s determined use the vibr kly collabora weary, lone after the p costly. st has No hering vibrat tion hardwa m with a pr ceived a seve nalyst to requ as not a pro nother vibra ing technolo stem). The b ram. Because r was alread ter. This was he data was ed system. T eague was tw travel neede that the pro ation monito te to make a analyst coul roblem first o Impact o ion data as re that autom eliminary dia re alert for est a quick r blem, excep tion analyst gies, the dat ackup analys this custom y equipped done while then availab he vibration o time zone d. blem was no ring service n accurate d have easily ccurred. Th n Operat part of its atically ana gnosis imme a very critica eview of the t this was a could be rea a was availa t was at a b er’s program to gather da the custome le via the Int analyst exam s away, but t with the St was implem diagnosis wit recommend e delay in an ions routine vibr lyzes the ma diately after l pump. Norm data in case weekend an ched. Becaus ble to any a ackyard barb was ta on r was ernet ined could age 2 ented hin a ed a alysis ation chine data ally, there d the e this nalyst ecue, but was a software. Using only not have machine w continued isolated o vacation. knowing t Case H Immed A steel m was previ than a ye processin email ale personne It was obv The pump in the reb result, da mill and f ble to borro a Web brow a problem. T as still being his data coll n the primar In the meant hat it was ac istory 3 iately for ill had just r ously equipp ar. This mon g, analysis, a rt was sent l as well as th ious that the was remove uild effort. B mage to the uture pump s w a laptop. ser, the ana o the relief established ection routin y analyst’s co ime, plant p tually proble : Cloud Faster R ebuilt a large ed with an a itoring system nd long‐term out by the e company w pump had a d from serv earing fits w shaft and ca ervice was n With cloud lyst was able of the techn . The alarm t e. Without c mputer. The ersonnel wo m free. -based A esolution critical pum utomated m gathers da storage. A cloud‐based ho was hired problem an ice and inspe ere loose, im sing had alre ot performed 7 computing, to review th ician, this wa hresholds we loud comput analysis wo uld be very n utomate p using an u onitoring sy ta on the pu fter the rebu monitoring to monitor d that the on cted. The ph peller and c ady begun. T by this vend not only is d e data and d s a false ala re adjusted ing, the softw uld have had ervous and d System nfamiliar pu stem that ha mp and send ilt pump wa system. The the equipme set was a dir ysical inspec ase wear rin he cost of th or. ata centrali etermine tha rm. The bas accordingly, are and data to wait until distracted by Detect mp service v d been in o s it securely s installed a system ale nt for such p ect result of tion revealed gs were not e rebuild wa zed, but so t the machin eline data fo and the tech would have he returned the machine Anoma endor. This peration for to “the cloud nd started u rted mainten roblems. the rebuild e several mis secured, and s refunded t is the e did r this nician been from , not lies pump more ” for p, an ance ffort. takes as a o the 8 With traditional predictive maintenance practices, it may have been several weeks before any health data was gathered on this machine. It would have been much more difficult to find a root cause at a later date. Even worse, the pump could have failed without knowing why and be sent back to the same vendor for a second rebuild; likely with the same result. Because cloud computing was already implemented for the mill’s predictive maintenance program, the problem was identified, diagnosed, and the root cause was discovered within hours. Closing - The Future of PdM Has Arrived As illustrated in the three case histories, the cloud is already opening up many possibilities in the world of PdM. The benefits are quickly realized by companies who choose to leverage this technology. Organizations that make the move to the cloud will undoubtedly gain the same benefits and discover more possibilities on their own. Leading organizations are already using this technology to drive a wedge between them and their competition. © 2011 – Azima DLI – All rights reserved. About the authors: Kenneth Piety, Ph.D., VP of Technology Kenneth Piety is the vice president of technology at Azima DLI, a machine condition monitoring company that has fully embraced the cloud. He co‐founded Computational Systems Inc. (CSI) and was a key contributor there for nearly 20 years. Ken holds more than 30 patents related to predictive and proactive maintenance technologies. He has worked for General Electric, Technology for Energy Corporation, and the Oak Ridge National Laboratory. Ken holds a Ph.D. in Nuclear Engineering from the University of Tennessee. David Geswein, BS, M.E., Director of Portal Operations David Geswein is the director of portal operations for Azima DLI. In various roles with Azima DLI he has helped build and maintain the largest cloud‐based PdM programs in the world. Prior to joining Azima DLI, David was a Reliability and Performance Engineer for Duke Energy. He is a Vibration Institute certified Category IV Vibration analyst and holds a B.S. in Mechanical Engineering from Purdue University. References: 1 (October 7, 2009) The NIST Definition of Cloud Computing. National Institute of Standards and Technology (NIST) Web site. Retrieved from: http://csrc.nist.gov/groups/SNS/cloud‐computing/ on August 22, 2010. 9 2 (April 13, 2010) Google ready to get down to business. CloudTweaks.com – The Cloud Computing Community. Retrieved from: http://www.cloudtweaks.com/2010/04/google‐ready‐to‐get‐down‐to‐ business/ on September 17, 2010. 3 (November 20, 2008) Gartner Says That 30 Percent of New Customer Service and Support Application Investments Will Be Through the SaaS Model by 2012. Gartner Web site. Retrieved from: http://www.gartner.com/it/page.jsp?id=808112 on September 17, 2010. 4 (April 7, 2010) Saving Money Through Cloud Computing. Retrieved from http://www.brookings.edu/~/media/Files/rc/papers/2010/0407_cloud_computing_west/0407_cloud_co mputing_west.pdf on September 21, 2010. 5 (March 25, 2005) Faster Guide: Server Consolidation. TechTerms@WhatIs.com. Retrieved from http://searchdatacenter.techtarget.com/definition/server‐consolidation on September 21, 2010. 6 Wikipedia – Hotmail. Retrieved from http://en.wikipedia.org/wiki/Hotmail on October 5, 2010. 7 (April 16, 2009) Cloud Computing Savings – Real or Imaginary. Retrieved from http://blog.appirio.com/2009/04/cloud‐computing‐savings‐real‐or.html on October 5, 2010.