AWS Solution for American Airlines

Executive Summary

American Airlines are one of the famous airlines globally. Ever since a long time the airline is facing technical & maintenance issues. The airline adopted IBM cloud solution but is still in trouble due to its legacy system & incapability to analyse huge real time data and draw useful insights over the competitions. AWS provides wide range of solutions and tools not only to solve technical problems but also to improve overall business productivity and performance. AWS provides services which are highly available, scalable, secure and at a low cost. Services like EC2, S3, RDS, Glacier, ELB, Route53 and more will provide better solutions to the problems currently experienced by AA. In this report we provide possible solutions to the problems, information about selection and justification of AWS technologies, cloud environment, VPC structure and few successful stories of existing airlines adopting AWS solution for similar problems.

Table of Content 

Executive Summary

Introduction

Existing Cloud Architecture

Challenges for American Airlines

Infrastructure Issues

Technical Issues

Possible Solutions

Technologies Used

Environment & Availability Zones

Virtual Private Cloud (VPC)

Justification

Successful Stories

Conclusion

Future Scope

References

Table of Figures
 

Figure 1 Vmware HCX IBM Architecture

Figure 2 Cloud Global Market Share

Figure 3 american airlines presence in world

Figure 4 number of availability zones per aws region

Figure 5 time for data transfer across region

Figure 6 ec2 latency across region

Figure 7 date from when services are avail in regions

Figure 8 no. of days for service unavailability in a region

Figure 9 vpc environment

Introduction
 

American Airlines is a major US airline having their headquarters in Fort Worth, Texas. AMR Corporation, the parent company of American Airlines, merged AA with US Airways in the year 2013 to become the largest airlines globally in terms of revenue-which is over $40Bn as of 2018, destinations served-more than 350 destinations, fleet size-956 mainline flights and scheduled passenger-kilometres flown. AA have 3 different strategic business units: Mainline passenger-operated by the airline’s main operating unit, Regional passenger-operated by regional airline’s operating unit and Cargo segments. Over 6700 flights are operated per day serving more than 350 destinations across 50+ countries. AA has a very large domestic presence which gives the company more than 60% of their revenues and the rest is by international flights and cargo. Passenger business constitutes to the 70% revenue and 87% of this revenue is in the form of airfare.

Existing Cloud Architecture

Earlier American Airlines were always on highlights for its system failures and malfunctions. They started gaining the reputation of not being so technically advance. Thus, American Airlines started assessing cloud providers and decided to move customer site and other applications to cloud.

According to Maya Leibman[1] (Chief Information Officer at AA) said that, “The company aims to revamp portions of the aa.com website so that the customer service and booking site can take advantage of on-demand computing power offered by cloud infrastructure.”

In 2016 American Airline partnered with IBM and started using its VMware HCX solution to deploy legacy systems which include its customer-facing applications to the cloud mainly for booking system and customer service. Researches makes it evident that AA is still facing problems with its IT infrastructure, flight maintenance systems, glitches in reservation systems, website crash and many more applications impacting its reputation & revenue.

Figure 1 Vmware HCX IBM Architecture

Though the company migrated its traditional legacy system to IBM cloud still there are some issues with the existing provider. Firstly, IBM is using third party open-source products to provide PaaS Services. For example, IBM doesn’t have its own serverless feature and is dependent upon Apache OpenWhisk unlike Amazon which had Lambda & EC2 which are closed.

Secondly, IBM have a small community support over social media and Stackoverflow compared to Amazon Web Services.

Thirdly, the global market share of IBM is very small. According to article published by Nodericks Technologies[2], the global market share of AWS cloud is the highest among all the cloud providers.

Figure 2 Cloud Global Market Share

Challenges for American Airlines
 

Infrastructure Issues
 

One of the main reasons for flight cancellations/delays is problem with infrastructure which could be like lack of resources, security issues, performance, etc.
On Thursday (14 June’18) PSA Airlines[3] subsidiary of American Airlines faced huge flight cancellations due to failure in computer systems impacting crew scheduling & tracking systems. Total 700 flights got cancelled on Thursday & Friday. The issue didn’t disappear over weekend and PSA suffered about 1100 flights cancellation by Monday cumulatively. Technical glitches caused American Airlines share to fell <1 percent.

Systems affected:

      Crew Management Systems

      Tracking Systems

Technical Issues
 

In April 2015 when the flight was delayed due to an issue in the Jeppesen app. Jeppesen is a digital flight bag or an app used by pilots to observe the activities occurring on the runway. A duplicate airport runway chart occurred in the app which did not sync with the app’s version. Due to this, the app become unresponsive and the pilots could not see the runway which resulted in flight delays for almost 2 days.

One more incident took place in November 2016 with Sabre, the company which handles the airlines reservation system faced temporary outage in check-in systems. This affected not only American Airlines but also other regional carriers associated with it. The company had its backup systems in place, however these systems are connected to a single micro switch.

In February 2017 in Philadelphia, American Airlines had a planned maintenance power outage between 1 to 4 a.m. Even when the power came back on, the internet access remained down. Due to this the airlines were unable to print boarding passes.

Another issue recently faced by AA on July 2018 was when computer outage halted flight operations for about 30 minutes. The main reason behind this was OS failure and brief connectivity issue with one of its datacentres.

Systems affected:

      Maintenance and Engineering System

      Central Reservation System

      Flight Operation System

      Airport Management System

Possible Solutions

In order to overcome above challenges below solutions can be adopted by American Airlines:

1. Reliable Backup Systems
   In order to provide more redundancy in case of hardware failures, we would require reliable backup systems which can be done using AWS Database services like Glacier, Simple Storage Service(S3). With AWS, we can replicate data across multiple Availability Zones & helps in fault isolation. The AZ’s are connected using high speed links.

2. Business Analytics Service
   American Airline can move towards more data-driven solution. The sensors of aircraft have massive amount of data. This data can be used for advanced aircraft maintenance methods and improving decision making of operational systems.
   With the use of Business Intelligence & data analytics we can not only predict about the condition of various aircraft parts but also it helps in making intelligent decisions about what needs to be done at any given time – such as, to repair or replace parts.

Amazon provides business analytics framework through its AmazonQuickSight service. Through this service we get business insights from data rapidly and it helps in performing advanced analytical functions. It also improves dashboards of airlines performance metrics using data visualization.

3. Disaster Recovery
    

A disaster recover site is very necessary for any organization to keep the business activities running in case of any natural calamities, damage or cyber-attacks. Structuring a Disaster Recovery Plan (DRP) for on-premises infrastructures means creating a secondary physical site which is highly expensive. AWS has partnered with companies such as N2WS, Cloudberry lab, Commvault which deliver effective disaster recovery solutions. CPM also known as Cloud Protection manager which is sold as Amazon Machine Image in the AWS marketplace provides efficient backing up solution at much lower cost.

EBS stores incremental backups, so we only pay for the snapshots that have changed after your recent snapshot. Every snapshot contains the information required to restore the data to a new EBS volume.
 

4. Regular healthcheck of systems  
    

Even with robust backup systems applications tend to fail every time. This can be avoided by practicing real-time analysis and monitoring of systems which provide an integrated view of the system’s performance. Services such as AWS CloudWatch can be used to examine and manage an organization’s applications.

Technologies Used

Below technologies will be used in our environment design: –
 

     Elastic Compute Cloud (EC2)

     Simple Storage Service

     Relational Database System

     Glacier

     Elastic Block Storage

     Elastic Load Balancer (ELB)

     Amazon Elastic Cache

     Route 53

     Cloud Watch

     Identity Access Management

     Simple Notification Service (SNS)

     Quicksight

Environment & Availability Zones

Availability Zones (AZ):
AZ’s are geographic regions where the AWS have their data centres. Each AZ has 2 or more data centres which are independent but are located close to each other. The main function of AZ is redundancy and data replication.

American Airlines have a large customer base spread across the world. Even though their presence is worldwide, the frequency of flights operating in and around the US is more. North Virginia, Oregon, Ohio, Ireland and Singapore are the regions selected based on factors listed below.

• Based on proximity to the customer or the user: AA airlines have a large domestic presence and their headquarters is in Texas, USA. It typically makes sense to have a greater number of regions in the US and few across other continents.

Figure 3 american airlines presence in world

• Number of Availability Zones – If we are building applications with strict availability requirements, regions having more than 2 AZs is required because if one AZ is temporarily unavailable there will be two other AZs to process all the transactions which helps in better performance of the application.

Figure 4 number of availability zones per aws region

• Regions have different latencies and data transfer speeds – N. Virginia and Ohio have the best inter-region latency with 23ms. For the application deployed in N. Virginia, Ohio is the best failover region.

Figure 5 time for data transfer across region

Figure 6 ec2 latency across region

• Availability of services – As of January,2018 N. Virginia is the only region to have all the services available. Ohio, Ireland and Oregon are other regions where the service is available for most of the days.

Figure 7 date from when services are avail in regions

Figure 8 no. of days for service unavailability in a region

• Cost of data transfer between AWS regions: The cost of data transfer between the AWS regions costs the same (20$ for 1GB), except for Ohio (10$ for 1GB). Ohio can be a best option for backing up of data.

Virtual Private Cloud (VPC)

It is a customer defined, logically isolated section of AWS where the AWS resources are launched on a virtual network. The customer has control over networking environments such as IP address range, creation of subnets and config of route table and network gateways.

      VPC Environment:

It maximises the number of AZ’s to isolate the data centre outage. For disaster recovery and high availability up to 4 AZs are recommended.

There are two separate subnets for routing requirements:

1. Public subnets for External facing resources
2. Private subnets for Internal facing resources

Independent routing tables configuration for each subnet to control the flow of traffic within and outside VPC. Allocation of one routing table for all the public subnets is enough as they all use the same internet gateway.

Usageof highly available NAT gateways instead of NAT instances.

Spare capacity for additional subnets for scalability.

Justification
 

American Airlines mostly faced issues related to website crash, system outages and downtime which eventually led to flight delays, customer dissatisfaction, loss of revenue and negative impact on the reputation of the company. Similar kind of problems were faced by Porter airlines, which eventually adopted AWS cloud technologies. Today, the Porter Airlines has overcome a lot of major issues by migrating their Sales and Customer Service Engine on the cloud.

During peak times, which usually comprise of holidays and festivals, American Airlines experiences a huge amount of traffic on its website. At other times, traffic is moderate and is the lowest from August to October. With such fluctuating volumes of traffic, it is necessary to scale up the resources whenever required and shrink them as well during non-peak times. This can be done using Auto-Scaling.

EC2: The EC2 instances support the airlines internal servers, applications and websites and increases the number of instances based on the incoming requests from users.

S3: S3 is used to store backup data which is already present in the RDS along with static images and files of front-end web applications of the airlines.

DynamoDB: It is used to track the web session data like session_id, user_id, session_data, last updated, created which can be essential in drawing insights from this data.

RDS: The RDS stores large sets of application and transactional data from the website applications which include customer profile information, flight details, booking summary etc.

Glacier: Data which is not being used frequently is usually stored In Glacier. Amazon Glacier is used to store such data for a long term.

ELB: Customers carry out various activities such as searching and tracking of flights, submitting form for user profile creation, updating their flight selection preferences and making payments and cancellations. All these tasks create multiple requests which are sent to the company’s server and may sometimes result into excess traffic. The Elastic Load Balancer distributes this traffic to several EC2 instances in one or more Availability Zones.

EBS: It is used along with EC2 instances to store the persistent data on the cloud. The infrequent data of airline like the history of last few years travellers, and other infrequent access data will be stored in EBS. The snapshot EBS is taken on a regular basis and it helps in disaster recovery.

Elastic Cache: It stores real time data such as flight tracking, baggage details, notifications about delayed /cancelled flights to provide ultrafast response.

Route53: It allows user to connect to Americanairlines.com and perform health checks of website

CloudWatch:  Every system application used by American Airlines is continuously producing large sets of data. This data is converted by CloudWatch into useful information and could be used in the form of statistics to measure the performance of the system. During peak hours, if the EC2 instance’s CPU utilization starts crossing its threshold, CloudWatch notifies the system. Also, if customer data is not getting retrieved, connectivity issue occurs within customer reservation systems, metrics are developed. These logs/metrics are then stored as a backup in S3 and can be accessed by different system users.

SNS: Simple Notification Service is used to deliver important updates or notifications to the end users. These may include flight schedule and status, PNR numbers, confirmation of booking/cancellations, delay in flights or security related messages.

Identity Access Management(IAM): It provides access to backend systems running on AWS resources to employees of AA, groups according to their roles.

Quicksight: It is the business analytical feature which helps in getting real insights of the data and perform business intelligence operations on it.

Successful Stories
 

Porter Airlines:
Porter Airline is the fastest growing regional airline with headquarter in Toronto. It operates mainly in Canada and USA.As per AWS Porter Airline casetudy, the major challenge faced by the airline company was to overcome the downtime issues and to improve disaster recovery system & make it more effective. So, the company adopted cloud solution for high availability & reliability for its website. It used Amazon EC2, S3, Glacier, RDS majorly for its environment. This way the company improved on its disaster recovery strategy and site performance without worrying about downtime.

Airlines such as AirAsia, FlyDubai, Qantas had the similar successful stories of moving to AWS cloud.
 

Conclusion

Based upon above success stories of Airlines using AWS and Stephan Treacy’s lecture notes we have used technologies like EC2, S3, DynamoDB, EBS, ELB, RDS, Glacier, CloudWatch, SNS, Route53, Elastic Cache, IAM, Quick sight in developing environment for American Airline.

References

• https://www.fool.com/investing/2018/06/19/american-airlines-suffers-the-latest-airline-it-me.aspx
• https://skift.com/2018/06/19/american-airlines-isnt-sure-when-it-will-fix-computer-issues-at-one-of-its-regional-airlines/
• https://www.cnbc.com/2018/07/29/american-airlines-flights-delayed-thanks-to-systemwide-computer-outa.html
• https://www.information-age.com/6-ways-business-intelligence-can-save-aviation-industry-123468838/
• https://docs.aws.amazon.com/aws-technical-content/latest/aws-overview/analytics-services.html
• https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/EBSSnapshots.html
• https://codeburst.io/how-to-create-a-disaster-recovery-plan-for-aws-a7efa6b24bd

[1] Nash, K. (2016). American Airlines Plans Major Cloud Move. [Blog] CIO Journal. Available at: https://blogs.wsj.com/cio/2016/09/21/american-airlines-plans-major-cloud-move/ [Accessed 11 Nov. 2018]

[2] Anon, (2018). AWS vs Azure vs Google vs IBM Cloud, which is the best for me?. [online] Available at: https://www.linkedin.com/pulse/aws-vs-azure-google-ibm-cloud-which-best-me-nodericks-technologies/ [Accessed 11 Nov. 2018].

[3]Franck, T. (2018). American Airlines flights resume after suffering brief nationwide outage. [online] CNBC. Available at: https://www.cnbc.com/2018/07/29/american-airlines-flights-delayed-thanks-to-systemwide-computer-outa.html [Accessed 14 Nov. 2018].