Information Systems On Delta Airlines

Headquartered in Atlanta, USA, Delta Airlines is by far the world’s largest airline by fleet size, destinations as well as passenger revenue. Delta airlines, founder and included in the SkyTeam airline alliance, encompasses a broad domestic and international travel network, with it unsurpassed global network. The largest operational hub of Delta airlines is the Hartsfield-Jackson Atlanta Internal Airport and the Detroit Metropolitan Wayne County Airport. Serving more than 170 million passengers every year, and counting, Delta together with its Northwest subsidiary as well as the Delta Connection carriers fly to as many as 355 destinations covering 66 countries, across 6 continents (Delta 2009) (StealingShare).

Delta accomplished its merger with the Northwest Airlines on October 29, 2008, with the main aim of forming the globe’s largest commercial carrier. Then, in February 2009, it started merging ticket counters and gates at airports at which both Delta and Northwest operated, and received permission from the Federal Aviation Administration to commence its operations under a single certificate. This consolidation was finished in February 2010. With the successful incorporation of the Northwest acquisition, investments of Delta Airlines, in newer products and network backed by continued efforts for strengthening its balance sheet, the company is favourably positioned to gain full benefit of the economic recovery. Northwest acquisition is estimated to generate about $2billion as annual revenue and price synergies by 2012(Delta 2009) (Travel Video News 2010).

Evaluation:

Ever-increasing competition throughout the airline industry is causing the development of new applications of information systems and technology. This includes a new strategic focus on electronic commerce or e-commerce at Delta Airlines. Delta’s mid-tier operation information systems has been presented as a scheme for leveraging its operational online transaction processing system (OLTP) infrastructure, in order to be an active part of the emerging world of e-commerce and enable new applications. The basic approach is to insert minimally intrusive ‘taps’ within the OLTP systems to track transactions as they occur for reproducible reply in the mid-tier operational information system (OIS).

For the existing environment, a hybrid approach can be developed and evaluated,

Conventionally, large enterprise computing at organizations like Delta is based on the usage of clusters of mainframes that run patent information systems software. A goof evidence of this statement is that Delta works depending on the cluster of IBM S/390 mainframe machines that run system TPF or Transaction Processing Facility. Such traditional OLTPs often support applications which automate most of the airline’s operational services. Further, the TPS systems architecture has shown high degree of scalability and availability, with the system operating successfully since the last 30 years and withstanding the Y2K bug scare (Delta 2009).

Technically, it is difficult to change the functionalities of these existing OLTP applications in order to accommodate a varying business. Several applications were deployed in assembly language and have evolved since then. The applications were originally designed for executing specific business models and providing little flexibility to support newer business models as well as processes. Particularly, these applications sustain ownership of fixed data sets, and their legacy data formats do not allow creation of new relationships to application data. Also, new business models and processes lead to new applications, many of which leverage the Internet, thereby resulting in exposure of legacy systems to unanticipated transaction volumes (Vasilecas et al 2006).

As a response to these drawbacks, Delta pursued a novel strategy of adding mid-tier enterprise information systems known as operation information systems, (OIS).

In essence, the pool of information in the current OLTP systems is gathered by “grabbing” strategic transactions when they occur in soft real-time. The transactions are then duplicated and consistently reproduced in the newly formed OIS. This new environment sees the mapping of data stemming from the transactions into alternative acquirable formats, which bears a correlation with initial unrelated information, together with information from sources apart from the OLTP systems. Furthermore, the immediate mutual-relation triggers events that are extracted from the transaction records. This susceptibility allows for a totally new category of real-time event based applications, which aim at radically improving the effectiveness of airline operations.

Moreover, the new mid-tier OIS, considered along with the legacy OLTP system, is said to be the foundation on which Delta generates new applications and enhances its existing business operations, such as improvement of the “Customer Experience”. The primary factor to their ever-growing success is the development of new mission-driven software and hardware infrastructures supporting these efforts (Vasilecas et al 2006).

The architecture of the operation information systems has evolved on a whole, since the scalability and availability requisites have changed. Earlier, the system represented a concept that gained instant success and was implemented far-ahead of its designed capability. The currently executed system has been technically refined to fulfil the scalability and availability requirements (HubPages 2008).

From a perspective, relative to data warehouses that generally store enormous bulk of historical information, an operational information system contains only the basic subset of information needed for day-to-day operations. While the size of the operational working set is relatively smaller, the collection of operation flows from internal as well as external sources may lead to operational data stores of terabytes in value. Maintaining such databases and the analytical processing of the data are two primary and basic tasks of the mid-tier OIS. Additional tasks are acquisition, derivation, broadcasting events having low latencies and in soft real time. Taking into account the demands of these tasks, a crucial observation indicates that the order of magnitude of the information from where applications events are obtained is possible to be reduced, mainly by emphasizing on the data required for operational decisions and actions. Hence, event latencies and throughput are improved by specifying a derivation subset termed as the Derivation Working Set. The DWS comprises of minimal amount of data required to derive the events needed for the OIS applications. Moreover, performance of data storage as well as data access for derivation of events is significantly enhanced as this working set can be executed as a main-memory database which is organized for accommodating event derivation and initial state queries (Oleson et al) (Mendelson & Brynjolfsson 1993).

A window scheme is used for operating the DWS, in which the content appears and disappears from the DWS based on relevance of information. Particularly, this set holds all state of current interest to be able to be rapidly accessed by relevant business process. For example, information regarding a flight’s arrival is stored in the DWS until the flight has departed, immediately after which ‘business logic’ is added to the DWS indicating that the data in regards to a certain flight section has been finished. Further, the lifetime or window of the data in the DWS depends on the business operations for a certain business domain. Such as years of experience in using flight information results in identification of a window of flight data coupled with behaviour for a number of days in the past and days in the future. Furthermore, lifetimes are different across business domains, and are not dynamic like the lifetimes of event arrivals. For example, a flight exits from a gate and starts taxiing, so the boarding process for that flight is not relevant anymore and may be discarded into the operational data store (ODS), and also to the data warehouse (Oleson et al).

Since an existing deployment crosses 10,000 machines, displaying flight status information, the greatest profile service of the OIS infrastructure is the soft real time delivery support of event information to numerous subscribing passengers. Further, real-time applications for event trigger the re-thinking of business processes and motivate to revolutionize the operations of the airlines. For example, when gate agents are supplied with alert displays which give the current view of relevant flight information, such as seat maps inside the flights they work for. The conventional request/reply approach is restricted as agents spend maximum time operating at the computer terminal, generally sending answers to customer questions. These heads up displays inform both the customers as well as the agents freeing the agents to spend their time in responding to more crucial issues, such as facilitating the boarding procedure (Travel Video News 2010) (Vasilecas et al 2006).

For achieving high scalability and lower latencies, the liberalization of the reliance of every event transmission is dependent on the application characteristics. Although some applications need tight assurances, others may run successfully under relaxed rules, known as the ‘reliability spectrum’. For exploiting this spectrum, the usage of a mixed sender and receiver-driven multicast protocol is capable of providing dramatic enhancements in the latency as well as communication scalability of an EDE.

The Event Driven Engine (EDE) is the major data provider and consumer for extra services related to the operational subsystem, like the Internet-based reservation and flight schedule and information services, which is the reservation system employed by external systems of a business to business model. Eventually, the EDE can directly distribute events to display points like the flight displays at airports, leading to the need for greater scalability in terms of amount of displays for certain event output streams emitting from the EDE.

The earlier EDE design employed a commercial relational database for internalizing the transaction records and depicting the operation working set. The initial purpose was to enable quick, flexible queries coupled with distribution of low latency event. Nonetheless, as the operational working sets are growing to Terabyte magnitudes, experts and the management of Delta instantly realized the competition among sustaining massive databases and rapid event derivation from this database. After using this deployed architecture, disk-resident relational data provided inadequate performance only to handle all of the work needed for the OIS infrastructure. Furthermore, not only the OIS should process the variable peaks of 12 million source message per day, but also must the OIS additionally derive a minimum of that many application friendly events to a deployment of approximately 10,000 workstations (Oleson et al). This desired amount of workstations is anticipated to increase dramatically in the future. The explosion of initial state queries take place as computers dynamically subscribe, which in turn require initial states. This initial state, for FIDS (Flight Information Display Systems) applications, resulting in XML result set of 5 MB exerts a massive load on the system. Worst case scenario will be all current 10,000 machines might come on-line at the same time requiring 10,000 queries. Further, this situation is worsened by the presence of additional external systems, such as passenger-booking traffic through the Internet, thereby resulting into the addition of much more information flows as well as resulting analysis tasks such as small flows like automatic passenger paging services, multimedia flows, etc (Mendelson & Brynjolfsson 1993).

Therefore, Delta discontinued offering further support to the feature of analytical queries of the OIS and started to maintain a lower in-memory depiction of the working set. Again, the relational database representation was used to recover this evaluated state during failures. However, frequent failures in the system could result into businesses facing substantial downtime. For this, the time to substitute the running cache from the many terabyte RDMS is approximately 45 minutes.

Furthermore, the client connectivity for the existing system depends on a hierarchical fan out on the basis of TCP socket concentrators. Delta was able to identify that this scheme adds unnecessary moving parts and inserts latency while events traverse the hops. Additionally, Delta’s requirements along with experiences in constructing a commercially embraced OIS infrastructure have greatly prompted the existing academic research. The present scalability challenges and future scalability projections demand clean slate scheme for researching more desirable and favourable architectures for an operational information system (Oleson et al).

Several applications operate successfully during incidents of message loss and take advantage of relaxed reliability protocols. This feature does not entail that the applications will have inconsistent views of information. This feature also proves that natural alternative means exist to guarantee the application information integrity. Furthermore, the most fundamental characteristic needed here is the ability to identify event loss and the capability of re-synchronizing a client application on detection of message loss. And this functionality is offered by the FIDS application of the OIS, where if a message loss takes place, the FIDS client re-synchronizes by asking for an initial state and starts receiving events that can update that state.

In essence, the reliability/performance tradeoffs of sender- vs. receiver-initiated multicast protocols are widespread, which offer stronger vs. weaker throughput vs. reliability, wherein attributes of both kinds of protocols are utilized for gaining a compromise for demanded reliability coupled with greater throughput. The receiver, in this protocol, controls and detects lost messages via sequence number analysis, while the sender is responsible for buffering the messages to accommodate retransmission requests (Mendelson & Brynjolfsson 1993).

Toward that end, the research as well as commercial opportunities have been presented by operational information systems (OIS) along with their strategic importance to Delta Airlines. “Tapping” the legacy operational systems is an interesting approach used by the research study to developing new systems employed by Delta. Further, desired representations of operational information can be reproduced for new, mid-tier OIS. The basic idea is to build additional systems across which new business applications are developed, without threatening the existing systems and their normal operations.

The evaluation of OIS then focuses on efficient, scalable and low latency processing together with the distribution of events, by evolving the existing communication/ event infrastructures and OIS event processing as well as storage engines (StealingShare).