Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
CHAPTER 11 The Integrated System The SSA process is the range of technological, managerial and operational components that the agency uses in carrying out its mission. The process primarily involves establishing client eligibility and determining appro- priate payments. The SSA's data management function is fundamentally one of collecting, retrieving, selectively changing or modifying, and re-storing a large amount of information and retaining it over a long period, while providing timely and accurate access to the information by widely dispersed district offices. Information about current wage earners and potential beneficiaries is needed to determine claims entitlement and payment procedures. Once client eligibility has been established, the information data base is used to determine the appropriate payment to be transmitted to benefici- aries by the Treasury Department. Information about clients or beneficiaries must be available only to authorized individuals and proper access safeguards must be in effect at all times. In a typical claim procedure, records should be available while the client waits at the SSA office, which is probably a matter of minutes. During the interview with the client, the information exchange with the data base should be accomplished in a matter of seconds. After the interview is completed, new and updated information can be sent in a less demanding time requirement, perhaps hours, to the master file to modify the record. The fundamental requirement is an open and dynamic system with record retrieval in minutes, interaction in seconds, and modification in minutes to hours. When coupled with increased efficiencies in data processing and new capabilities in data storage, a system meeting this requirement should remove the need for any significant increase in the number of employees, despite any increased workload imposed on the SSA. The SSA data base now represents slightly more than 1 trillion "bytes" or characters of information, much of which should be accessible in real time with a high degree of reliability and respon- siveness. The panel estimates that between 10,000 to 30,000 terminal access points will be in use by the mid-1980's, when the data base undoubtedly will be larger. The future data base can be designed to accommodate changes made to increase operational efficiency, as well as 13
14 those resulting from new legislation or other considerations. The achievement of such a data base appears to be within the capability of the technological state-of-the-art. To store the massive amount and variety of client information, several storage architectures are available, ranging from centralized through regional to decentralized or local. To meet the wide spread of timeliness criteria, storage media at any location can be available in a range of capabilities that provide different response times to queries. The array of processing alternatives runs from large cen- tralized mainframes to local distributed processors. SYSTEM ALTERNATIVES The OAS has made an initial evaluation of the cost of alternatives to provide future system capabilities. The alternatives included both central system and regional system configurations. The OAS calculated that personnel costs associated with the various alternative configura- tions greatly outweighed equipment costs. Only 6 percent of operating costs in 1980 are likely to be for data processing and communications. Because of the availability of the Baltimore headquarters facilities and personnel, OAS concluded that the most desirable system would be cen- tralized in Baltimore and some distributed processing would occur at local SSA offices. The panel has examined this conclusion by evaluating conceptually three generic options, schematically represented in Figures 1 to 3, page 15. The centralized system (Figure 1) consists of a group of large comp ~ files in a single location. A nationwide SSA communications network provides access to this center. Individual terminals may have access directly to this center or through hierarchies of local computerse Clusters of terminals and printers can be arranged in configurations for temporary storage and readout of client records. The OAS chose this configuration initially because of several attractive features. Currently, the SSA process is operated at a single center in Baltimore, thus easing the transition to the proposed system. Moreover, a major construction program is underway in Baltimore to house additional data processing equipment and provide greater security. Such a massive data base can be most efficiently managed by a technical staff in one location. The major disadvantage to this approach is the lack of a backup file. A major disaster or interruption of service at the central facility could have a drastic socioeconomic impact. another drawback is that this concept involves the shutdown of the present six regional centers, where many of the 13,500 employees have useful SSA experience. The regional system (Figure 2) consists of several regional cen- ters, each with processing and storage capabilities. As in the central- ized system, terminals and local processors are interconnected in a compatible way by a communications network. The major advantage of regional centers is that they provide redundancy and support. The physical separation of processing and storage elements, together with
15 SINGLE CENTER ......... - . Com u nication~ t : :::: t::::::::::- u ' F . . . . . .. - - ...... ; ; ~ ; ;'2 dY .^ FIGURE 1 REGIONAL CENTERS r ~ .. : .... ..~ .......... :. .~ ~ ~ L ~ ' Communications Network. .] .:::: :.:; .~ sy.~ ·:-:-:-:-:-:-t 1 , ~~ t.~2~., In, :-:-:-:-:-: e :::::::::::' ::::::::::: :::::::::::' ::::::::::: :.:.:.:.:::' a ~--r--~ I ~ I. J FIGURE 2 FULLY DISTRIBUTED Ad. _ ._.~, , . ,~,.~ ~ ~.~ ~ .... .. .. ... ..... .,_.~ ~ ~ i Communications Network : ,` L ~ ~ ....... " : ~ At, a ~ pi = ~ ] d b ~ ~ I~T''''-I---~-- i: day ~ d,tb day pub FIGURE 3 1 11111 Large Computer · Distributed Computer Terminal
16 provisions for redundancy, could provide the capability for virtually uninterrupted system operation in the event of an interruption at any one of the centers. The number of regional centers would most likely be between two and six. The major disadvantages of this approach are 1) a probable increase in cost; and 2) the difficulties of staffing, managing and coordinating multiple sites, (although such problems may be mitigated because the six regional centers are in operation with experienced personnel). The fully distributed system (Figure 3) would locate data pro- cessing and storage at several hundred geographically dispersed locations. This configuration has the advantages of less complex hardware, fast response time, and flexibility for changes in hardware. It also would take advantage of the rapid development and declining costs of small new computer systems. It may provide the opportunity to measure costs and efficiency more accurately because each district office could be responsible for its own configuration. There are serious drawbacks, however, to this approach. It would be more diffi- cult to manage and control than the regional approach. Standardization of operations, security, and auditing would demand close management control in such a highly decentralized system. With a widely dispersed data base, summary information would be more difficult to derive for use at 'headquarters. It is probable that the communications-information network would be more complex and costly for this approach. Moreover, the software technology for interaction of a large number of dispersed data bases is still developing, and, therefore, more technical risk is associated with this approach. While the majority of the panel has concluded that the fully distributed design is not a meaningful alter- native, a minority has supported continued investigation of the concept because of the possible cost advantages of small processing systems. For reasons that are more fully developed in the Appendix, (pages 64-66), the panel has concluded that the fully distributed system should not be pursued, but that both the centralized and regional concepts deserve further consideration by the SSA. MAJOR SYSTEM CONSIDERATIONS Segmentation of the Data Base By all standards of measurement, the SSA data base in both the cur- rent and proposed process is huge. Its efficient management is critical to the stability and performance of the future System. The panel, therefore, urges that the SSA should consider carefully the functional and physical structuring of the data base. Such structuring is consid- ered vital to the flexibility of the future process in order to be adaptable and responsive to new and as yet unidentified requirements, as well as to achieve the desired service levels. Efficient structuring of the SSA data base is a key design decision. Segmentation of the SSA data base is referred to as symmetric when all the information on an individual client is maintained together in a
17 single file record. A collection of the client records would contain all of the information on a set of clients corresponding to a set of social security numbers. Such a structure organizes the data according to the "whole-person" concept (see Glossary) in which all information about any one client would be maintained within a single part of the data base. This would be an extension of the current manual process in which most of the information about a particular client is in a hard copy folder, maintained in one of SSA's program service centers. Data base segmentation is said to be asymmetric if the data base is segmented primarily by functional type of data rather than by indi- vidual client records. For example, one part might contain earnings info`~at~on for multiple clients; another might contain claims data for these same clients. The advantages and disadvantages of symmetric and asymmetric segmentation are outlined in the Appendix (pages 66-70~. The panel has concluded that a totally symmetric data base configuration does not appear to be feasible. In particular, the social security number assignment will need, in all likelihood, to be an asymmetric segment of the data base, with all of the alphabetical name identification references (alpha-ident) being made to this single partition of the data base. Insofar as it can be realized, however, the symmetric con- figuration has many advantages. The overall data base structure will need to include both symmetric and asymmetric segmentations. However, it should be as symmetric as possible. Partitioning of Data Base . . . Because the SSA may wish to evolve from a single centralized system into a regional system, the panel recommends that the data base be organized from the start to facilitate such a move. By partitioning the data base properly, the various parts can be stored either in different computer subsystems within a single center, or in physically and geo- graphically separated regional centers. Current policy could be continued, whereby a record remains at the physical location (region) where it was initially assigned. Any reassignment of records to other logical partitions would be determined on the basis of cost/performance tradeoffs. Updating the Data Base A fundamental problem is the optimum manner of updating the master files as new data enters the system. At least three methods could be used. First, the updated record can be inserted on-line in real time-- that is, as soon as the information is received at the central site, the update can be made on the master file. Second, the record modifications can be stored in a transaction file for insertion into the master file on a regularly scheduled batch cycle, typically overnight. The third is a combination of the first two--referred to as "time-available insertion of the record into the master file." If the computer system has the
18 capability, it would immediately insert a record into the master file at the time it is received at the central site from the local computer; otherwise, it would store the record at a temporary storage point for insertion later into the permanent file. The system needs to possess additional capability to hold modified records pending their insertion into the master file. This could be accomplished by moving the modified record to its primary storage loca- tion as well as to one or more alternate holding sites. With separate regional centers, the data base for each region could be stored at primary and secondary sites. Moreover, different portions of a regional center data base could be duplicated at different backup sites. The last alternative results in a more even distribution of load end j hence, better backup performance. Storage Devices The future SSA process is to be mainly on-line. While the current process is heavily dependent upon master files that are on magnetic tape, the future one will have information stored on direct-access storage devices such as disks--or possibly mass storage for data that is accessed less frequently. The system is planned for implementation during the early 1980's, with operation anticipated through the 1990's. While it is feasible to accommodate the current data base with today's technology, the system designers must accommodate advances that will result in improved perfor- mance and decreased cost of on-line, direct access devices. Memory technology is discussed in further detail in the Appendix (pages 72-74~. Processing - In the proposed process, the balance between on-line interactive processing and batch processing should be determined on the basis of the timeliness and frequency required to provide the desired level of service to clients. Whether processing occurs in large central/regional proces- sors or in smaller distributed ones is also a function of the timeliness, frequency and processing requirements of each *transaction. Typical on- line functions include access to the data base to verify the correctness of client information, to determine a client's claim entitlement, or to update changes in client status. Even so, a significant portion of the processing will continue to be performed in a batch mode. Items such as earnings data will continue to be sent by employers or by individuals to data collection centers where they will be collected, edited, formatted, and then inserted into the system on a batch cycle. Furthermore, the payment process, once claim entitlement has been established, will prob- ably continue to be carried out in a batch cycle. While most processing will be done at central or regional centers, some portions could best be supported by on-line distributed processing at local offices. In order to understand how distributed processing might be utilized to support the SSA process, consider a claims representative at
19 a district office interrogating the on-line data base for information about a particular client to determine eligibility for benefits. Upon entering the district office, a client provides identification, which includes the social security number. While the client is waiting to see a claims representative, the receptionist calls up the client record from the central data base for temporary storage in the district office. A response time of up to several minutes at this stage would not have a major effect on the service performance. Once the client is assigned to a claims representative for an inter- view, the information on the client would be available at the district office. Interaction with the record by the office could be highly responsive--probably in one or two seconds. The claims representative could make modifications without changing the master copy that is main- tained in central storage. Once the client and claims representative agree that the information in the local copy is complete and correct, the information would be transmitted to the central/regional site responsible for the particular record and subsequently inserted into the master data base. At the same time, this transaction would be retained at the local office, thereby providing multiple copies to safeguard against errors in the record or malfunctions in the system. The system design must provide accommodation for the queueing of transactions at key points for processing either in regularly scheduled batch cycles, or on a "time available" basis. Distributed intelligence should be utilized where feasible for local processing of the info`~ua- tion maintained in the central/regional data base. The Communications Network Any future system will require an extensive communications network that will range from low-speed, low-capacity to high-speed, high capacity lines. However, networks are already in operation with as many terminals and as much traffic as predicted for the proposed system. Therefore, the size of the network and the anticipated workload are not inhibiting fac- tors. Some key decisions are pivotal for the design of the communications subsystem. These might not be made until well into the system design phase or even the development phase. The question of whether the pro- posed system will be centralized or regionalized has intentionally been deferred. This decision will have a major impact on the requirement for high-speed bulk data transmission. Questions about how much district office processing can be performed by distributed processors, how much may require on-line interaction with a remote site, and how much can be handled by scheduled batch transmis- sion, say, nightly, may not be answered until a working prototype of the new system is available. The amount of remote interactive processing could have significant repercussions on requirements for communications. The communications component of the future system should be segre- gated with strictly enforced interfaces between it and the processing centers. Segregation provides flexibility for the incorporation of changes in the communications system without major perturbations to
20 other elements of the system. It also gives the SSA an option to pro- ceed immediately to a separate sophisticated communications subsystem, or to defer that decision to a later point in the development program. The panel is reasonably certain that the communications network should provide both packet and line switching capabilities. Moreover, the communications network should permit communications among a wide variety of incompatible terminals and computers, with different proto- cols, formats, speeds and codes. The communications protocol should be modularized so that the distributed processors, terminals, and central processors could interact readily with the communications network. At least four communications paths will be required in a central design concept: 1) terminals to distributed processors, 2) terminals to central processors, 3) distributed processors to central processors, and 4) data collection centers to central processors. A fifth communi- cation path would be required among the regional centers if a regional system configuration is chosen. This regional communications link would be equivalent to the data base communications that would be required within a single central complex if intra-complex partitioning of the data base were implemented. Figure 4 (page 21) represents one possible communications network. Other networks are possible, and the graphic depiction should not be construed as suggesting where regional centers and computers could be located. The communications nodes are interconnected by wideband trunk lines. Each central/regional center, data collection center, and dis- tributed computer connects to one (or more for reliability) of the communications nodes. Terminal devices gain access through distributed computers. This configuration results in a communications system that is optimized, not for any preconceived application, but, rather, for flexibility to accommodate to changing requirements. Two fundamental types of interchange must be provided to accommodate terminal and distributed processor communication requirements to and from central and regional centers: 1) interactive on a record-by-record basis and 2) file transfer where large amounts of data are involved. The panel is of the opinion that the cost of communications will not be a major factor in the future SSA system. The panel recommends that terminals for the future SSA system be stock items from the commercial mainstream rather than specially designed items. Flexibility for introducing new terminals into the network must be maintained throughout the operating life of the process. Therefore, the terminals should be selected to assure a current capability that can be enhanced as technological advances and cost reductions come about. Because there will be between 10,000 and 30,000 terminals installed at the peak period of utilization, the deployment cycle will be long. To maintain the terminals in a cost-effective manner, either the supply- ing vendors must provide the on-going maintenance, or the SSA needs to create its own maintenance capability. Of necessity, an in-house maintenance staff would have to be extensively distributed geographically to support the nationwide network. In any event, the communications system and the terminals should have the capability of being tested remotely.
21 In / \-- / N~ Al ~ FIGURE 4 Communications Network S 0, z c ~ -a cat ~ cat ~ _ O ~ ~ _ .o ~ ~ -, E O O O D
22 The GAS planning envisages that the primary interactive terminals will have video display capability--their principal interaction being with the distributed processors and through these to central/regional processors. The panel recommends that when the interaction with a local terminal is being shifted between a local and central processor, this should be selected dynamically, either explicitly or implicitly, by the terminal operator in conjunction with the distributed processor. The need will exist for terminals to deliver hard copies of selected information, primarily for the use of clients or off-network functions such as management reporting. While the design should minimize paper- handling to the extent feasible, there will always be some requirements for hard copy. Examples include client copies of status, performance records and documents requiring signatures by the claims representative or the client to acknowledge the execution of transactions. SYSTEM DESIGN METHODOLOGY In the SSA's design of its future process and the system that implements it, the panel makes the following recommendations: o Develop an integrated process with a largely on-line or real- time capability. View the entire process as one that can be implemented by a real-time on-line system to achieve the full objectives of the whole-person concept. Develop the various subsystems with off the shelf components. In order to achieve the full objectives of the "whole-person" concept, the entire process needs to be planned as on-line or real-time from the outset, even though certain processing elements cannot be executed in real-time during the initial implementation phases. The data base must be designed to function in an on-line system, and the programs that will interact with this data base must be able to do so in real-time. The storage media initially installed may not be capable of microsecond or millisecond access to the data. Therefore, the storage design must possess the leeway to accommodate to potential delays in the processing-and-hold activity at various points until enhanced capabilities become available. The data base design should remain independent of any particular storage device to the extent possible in order to take full advantage of technological advances that may be expected during the system's life. This design concept can be illustrated by the claims process within a district office. After a claim record is removed from the data base and transmitted to the local processing point for updating, it must be returned to the central or regional file for reinsertion into the master data base. How this will be accomplished in the future process has not
23 yet been determined. It will probably not be determined until rather late in the design process. As part of the implementation of the integrated real-time concept, the design should support the queueing of the updated records at the central or regional centers, as well as the processing of such records as the system is developed. In the early stages of implementation this can be done on a nightly batch cycle. Later, this can evolve to real-time insertion as the technology and the system reach maturity. The on-line real-time aspect of the system can also have signifi- cant consequences for user interface and training. The policies and procedures that will govern the activities of the system's users can be stored and maintained at the central or regional sites in an on-line data base and made available via the same communications network and interactive terminals that support the SSA process. In order to facilitate the smooth design, development and implemen- tation of the future system, the panel recommends that it be segmented into discrete computer, communications, and applications subsystems. The major computer and communications subsystems that are readily identifiable within the system are: Central Storage Central Processing Data Communications Local Storage Local Processing Terminal Devices In addition, the future process has identifiable application sub- systems that can be utilized to segment the future process, such as: Enumeration Earnings Claims Post-Entitlement Management Information This list is representative of the types of subsystems for which well-developed interfaces should be maintained. Segmentation provides the option for concurrent design and development, as well as for the involvement of individual vendors for different segments of the system. For example, one vendor can be selected to provide a communications service and another data base management. The SSA needs to exercise caution, however, so as not to acquire an ad hoc data base structure that would be neither generally available In the industry nor device independent. Segmentation gives flexibility to the transition plan. Thus, by having a standardized design with fixed interfaces to other hardware subsystems, the communications subsystem could either be the first to be implemented or it could go into service later. Flexibility makes it more likely to take advantage of new, more cost-effective, or more
24 responsive technology. Provided that there are well-defined interfaces among the data base and both the applications and the processing sub- systems, the option is available for introducing new storage media or access methods without causing major perturbations to the system. By using off the shelf system components rather than custom built elements in the system, many of the same opportunities as discrete subsystems will result. In this way, the SSA has more flexibility in the selection of vendors and an option to choose second source vendors at little additional expense. The use of off the shelf components enables suppliers of various hardware and software systems to make their new components upwardly compatible. This could be a major consideration in central processing. With such tools as standard processing languages, the SSA would have the option of introducing advanced central processing units into the system without major perturbations. Another favorable aspect of off the shelf components is that technical expertise would be readily available to maintain and support the equipment e It would be more difficult to sustain a highly skilled technical staff to maintain such a system consisting of specially designed components. TRANSITION PLAN The successful operation of the SSA process during the transition period is vital to millions of Americans. Ideally, the conversion sequence should best follow the chronology of the SSA processes, in which case the order would be: enumeration, earnings, claims, and post-entitlement. This sequence provides the most natural transition to the future process and probably requires the minimal amount of bridging between the current and projected data structures. The selection of a conversion sequence is related to the data base structure itself. The information within the future data base will be keyed primarily to the social security number (SSN). Therefore the first future application to be implemented should be enumeration--that is, the management of the SSN. Because each subsequent procedure requires the SSN, and because the accuracy and authenticity of the SSN is essential to the successful execution of subsequent stages of the conversion, the panel recommends that enumeration be made operational and stabilized prior to the initiation of other functional conversions. Once the SSN management application has been stabilized, it will be possible to gather earnings data for each wage earner in the SSN file. The information in the SSN file describing the status and earnings record of the client would be essential for claims actions. Finally, after claims entitlement has been established from the earnings and identity data, the post-entitlement application could be made opera- tional, utilizing the previously converted information. Such other aspects of the new system, as the communications network, the distributed processing, and the terminals can be interwoven into this basic conversion plan. It will be necessary to bring up all the supporting functions--e."., processing, storage, and communications-- involved in an application by the time the application is introduced.
25 Therefore, the panel recommends that the communications network and terminals be in operation before the enumeration process takes place so that the district offices will have access to both the SSN identity information and to the earnings data once these processes are converted. Because distributed processing will probably be used to support the claims and, eventually, the updating of SSN and earnings data, the panel suggests that the distributed process be deployed just prior to conver- sion of the claims process. Other considerations may override the panel's view of this sequence of conversion, one of which is the need to maintain the stability and cost effective operation of the post-entitle- ment process as it currently operates. Another major consideration in the conversion effort is data base organization. If an asymmetric design is selected as the primary division of the data base, more options are available for the sequence of conversion. An asymmetric data base design would more readily support the initial conversion of the post-entitlement process. If, on the other hand, the segmentation of the data base is primarily symmetric, then the conversion sequence following the life cycle of the SSA processes would be the most desirable. Because the data is now managed in a single central site, conversion from the current to the future process can be done most efficiently within a single center. The panel recommends that the conversion be completed within a single center even if it is decided to deploy the system later to regional centers for security reasons.. The data base should be segmented from the outset. It should become operational in the central site, therefore, with geographic dispersion effected on an incremental basis. Summary . In summary, the panel recommends that SSA: Make a strong effort to standardize throughout the system, utilizing operationally tested products as they become available in order to ensure that the total system will have a long life. O Segment the subsystems--such as terminal devices, distributed intelligence, telecommunications, network protocol, central site processing hardware, and central site storage media--so that their design, development, maintenance, and evolution can be facilitated by having them as independent subsystems within the total design. Segment the applications processes for the same basic reasons. · Emphasize on-line, real-time operations. . Design the system so that it could function in physically sepa- rate centers, if this is determined to be a requirement. It
26 is recognized that the transition process will not accommodate this conveniently as the first step in the conversion. Align the conversion of applications with the basic chronology of the typical SSA client--enumeration, earnings, claims, and finally post-entitlement.
