Return to Non-Functional Requirements

• [char]Please Review

Manageability is most important during the second half of a System Lifecycle phases (i.e. operation, maintenance, support). Manageability can greatly influence the recurring costs and can increase the chances of a failure. Often a system that hard to manage is described as fragile since the smallest of changes can have dire consequences on the systems functionality.

Manageability directly influences a system's reliability, availability, security, and safety, thus being a key ingredient of system dependability.

Just like security and safety, manageability is generally hard to retrofit in complex systems—it is always easier to build it in from day one. However, in the absence of means to measure manageability and to quantify the various tradeoffs, it is difficult to get the design right. We proposed a manageability metric that combines management workloads and weightings based on real world studies with direct measurement of the number of steps involved in management tasks and their duration. ¹⁾

Return to the Top

NI (formerly National Instruments) defines four basic manageability functions ²⁾:

Return to the Top

It is no longer necessary to argue that managing enterprise computing systems is complex and time-consuming, or that the cost of managing Information Technology (IT) infrastructures far exceeds the hardware and software costs—numbers speak for themselves: IT operations account for 50%-80% of today's IT budgets, amounting to tens of billions of dollars yearly. Besides the bottom line, poor manageability also impacts reliability, availability, and security in harder-to-quantify ways. As human error becomes the dominant cause of unscheduled downtime, we desire systems that are easier, cheaper, and quicker to manage. ¹⁾

The cost of managing systems is expensive and is getting more expensive. This is particularly true when using servers that are aging. The cost of support and administration of a new serer is about $4,200/year; by year seven, the cost will rise to about $17,000/year. It is logical to try and decrease these costs through refreshing the hardware ³⁾.

Figure 1: Year-to-Tear Changes in Relative Server Costs

Although these are the costs of owning and operating physical servers, the trend is relevant to almost all IT systems including software. The costs of managing the software is analogous to maintenance of the hardware but is probably worse since application software is dependent on many more layers of hardware and software. For example, a software system sitting on multiple Operating Systems (i.e, Linux, Unix, Windows, MacOS, IOS and Android), using a DataBase Management System (DBMS), a Web Server and using Java, Python, JavaScript, HTML, CSS, and supporting multiple browser (i.e., Firefox, Chrome and Safari) soon is overwhelmed by just keeping each system up-to-date and working. Distributed systems can compound that problem because each node in the System Network needs to be managed as well.

Granted, there are products and tools that can reduce the complexity of managing the various platforms easier and this is a big step towards helping manageability. But the overall goal is always to have systems that are easier, cheaper, and quicker to manage. There is another way to think about manageability called the Bathtub Curve. Although Figure 2 is targeted at hardware, there are similar things that effect software. Probably one of the biggest problems is that as the number of components in the system increase, the number of End-of-life (EoL) or End of Sales(EoS) issues need to be addressed. For example, a system that runs on an Operating System, has to worry about the EoL of the operating system. If it uses a database, it has a similar problem. These are generally not a problem in the early stags of the a system's lifcycle and hopefully of minimum problem during its “useful Life” phase. But as the system ages, more and more EoL problems arise. The more Commercial Off-The-Shelf (COTS), Government Off-The-Shelf (GOTS), Modified Off-The-Shelf (MOTS), and NATO Off-The-Shelf (NOTS) products used by the system, the longer the system exists there is an increase in risk to the system because each subsystem, component or modular need to be managed.

Figure 2: The Bathtub Curve ¹⁾

Return to the Top

“Studies have shown the average software program lifespan over the last 20 years to be around 6-8 years. Longevity increases somewhat for larger programs, so that for extremely large complex programs (i.e., over a million Lines of Code – LOC) the average climbs as high as 12-14 years.”⁴⁾ Obviously, there is not just the lifespan of the target system, but there are independent lifespans for each version for each subsystem, module or component that is developed externally. For example, the Windows Operating System (OS) first appeared in the mid 1980s with version 1.0.⁵⁾ and the current version is 10. About every 10 years, Microsoft release another major release of Windows. ⁶⁾. IPV4 was originally available in 1883. Windows 7 was release in October 2009 and IPV4 was the dominate Internet Protocol (IP). By 2012, IPV6 gained dominance⁷⁾. Therefore, if your system was released in 20010 using IPV4 and Windows 7, by 2012 the network protocol needed to be upgraded which can have cascading maintenance effects throughout your system. By 2015, Windows 10 was released again having a cascading effect on upgrades.

Figure 3 developed by the Industrial Internet Consortium⁸⁾ illustrates some of the architectural components required in a generic Industrial Internet of Things (IIoT) system. When a system is deployed, each of these components needs to managed. Each of these components has its own unique System Lifecycle which evolves independently of the target system.

Figure 3: The Industrial Internet Consortium’s Connectivity Framework layer provides the foundation for the interoperable transfer of common structured data across systems and domains ⁷⁾

Return to the Top

The fundamental requirement for manageability of any system is the collection of data about the system. This is often done with Monitoring Software specifically designed for this task. However, the task of monitoring complex, distributed systems is often difficult and beyond the scope of any particular product. The best place to start is to think of the monitoring in terms of layers. There are considered three layers⁹⁾.

Figure 4: Three major monitoring layers representing sample design of monitoring solution

Return to the Top

The Data Monitoring is the use of Data Logging to collect and store data for analysis to discover trends or record the events and actions of a application, a system, network. This allows for tracking of interactions using messages. Some of the commonly used logging levels are:

Return to the Top

Project Management Software is software used for project planning, scheduling, resource allocation and change management. It allows project managers (PMs), stakeholders and users to control costs and manage budgeting, quality management and documentation and also may be used as an administration system. Project management software is also used for collaboration and communication between project stakeholders. These tools can help throughout the system lifecycle from requirements analysis through to sun-setting the system. However, in a distributed system, these tools can play a significant role in determining the health of the each of the nodes, the network and the overall system of nodes.

Although the publication on Software Metrics for Predicting Maintainability ¹¹⁾ is a bit dated, many of the ideas of capturing metrics to measure Maintainability are still relevant today. By studying these metrics and understanding the formulas and the parameters used in the formulas, a lot of insight can be provided in explaining positive and negative Manageability traits.

• Note: Many of these metrics were originally targeting for Systems (or projects) that used the Waterfall Model which has been replaced for many systems (or projects) with Agile Models and DevOps. Many of these metrics can and should be applied to each Sprint to make sure the qualities of the system are maintainable and manageable. In a Distributed Application (ĐApp or DApp), word module can be replaced with subsystem or Node.

Return to the Top

A major management issue for many projects is the avoidance of Vendor Lock-In. Vendor lock-in restricts the options available to a system (or project) because of the dependency on sole-source proprietary technology, solution or service provided by a single vendor or vendor partner. This technique can be disabling and demoralizing because customers are effectively prevented from switching to alternate sources for the technology, solution or service making the customer-vendor relationship one sided.

Vendor Lock-In reduces the ability of manage costs over the life expectancy of the system (or project) or avoid risks when a vendor of a product ceases to maintain a critical component of the system (or project) or even when the product ceases to exist.

This can be partially mitigated through the use of Open Source Software (OSS), however, remember, just because something is OSS, does not mean there is not a vendor. Additionally, if the OSS software is deprecated or evolves in a divergent way from the requirements of the target system (or project), then the responsibility for the care and maintenance of the OSS has to be covered by the system (or project). However, there are OSS solutions which also adhere to standards such as the Data Distribution Service (DDS) vendor Object Computing Incorporated (OCI) .

An example of an OSS offering that has suffered from a calamity is XeroMQ. Even though ZeroMQ is still around and being used, the ZeroMQ OSS effort was shaken by the death of its prime moving force Pieter Hintjens who died¹²⁾. There are any spinoffs and derivatives of ZeroMQ. Here are a few reported in Wikipedia¹³⁾

• In 2012, two of the original developers forked ZeroMQ as Crossroads I/O.¹⁴⁾¹⁵⁾
• Martin Sustrik has started nanomsg,¹⁶⁾ a rewrite of the ZeroMQ core library.¹⁷⁾
• In 2012, Dongmin Yu announced his pure Java conversion of ZeroMQ, JeroMQ.¹⁸⁾
• This has inspired further full-native ports of ZeroMQ, such as NetMQ for C#¹⁹⁾.
• March 2013, Pieter Hintjens announced a new draft of the ZMTP wire-level protocol bringing extensible security mechanisms to ZeroMQ²⁰⁾.
• Martin Hurton implemented the CurveZMQ authentication and encryption mechanism²¹⁾ in the core library shortly afterwards.

Not recognizing or managing the risks due this kind of unfortunate occurrence to the system (or project) might be expedient, but is in many ways irresponsible for systems (or projects) with a long lifespan. An alternative to the risks of OSS or proprietary vendor lock-in is the selection components that are standards based from a Standards Organization that offers a wider spectrum of vendors to chose from.

Return to the Top

Over the last few decades, many advances have been made in terms of Open Source Software (OSS) which has to change the way that software was developed, released and used. During this same time period, the traditional Waterfall Model of System and software development has also been largely supplanted with the Agile Model. Many of these changes also have to do with the evolution of systems (or projects) from being Greenfield to Brownfield development and from a “build the world” attitude towards “integrate and glue the world” mindset.

Successful OSS development and adoption not only has to produce products which are solid, strong and robust but also must meet the needs of a Community of Interest (CoI) that has coalesced around a single minded, purpose built, functionality (i.e., Apache Tomcat application server, PostgreSQL Database, Node.js an asynchronous event-driven JavaScript runtime, Docker containerized apps, Kubernetes orchestration engine for containers, etc.). Many of the OSS products are part of many of the successful projects today.

However, its not good enough to just write software and make it publicly available. At the heart of these successful efforts are the well governed, focused, supporting CoIs. There is a desire from almost all systems (or projects) to join the OSS trend but unfortunately, the need for strong governance and rigorous methodology is minimized or skipped in the name of expediency. Fortunately, there is an organization which can help with this called Talk Openly Develop Openly (TODO) (not to be confused with a to-do).

TODO organization, though focused on OSS, has written a series of white papers that are well worth studying and using even it your system (or project) is not OSS. One of these papers which is particularly germane to Software Manageability is Tools for managing open source programs ²²⁾. It is beyond the scope of this document to try to recreate the full content of this white paper. It does present a list of many of the tools available for managing software and how to use them. Here is the table of content from the document:

• Why you need special tools for open source program management
• How to select and plan your tools
• Elements of a basic toolset
• Tools for managing source code
• Tools for tracking project health
• Tools for communications and collaboration
• Tools for corporate-scale GitHub management

Return to the Top

Data Distribution Service (DDS) can not solve all of a systems(or projects) Maintainability issues, however, by DDS's design, it can eliminate or reduce the Manageability issues that could arrise from using DDS.