AppSwitch: Network Switching with Ada from Linux
Managing network performance is like keeping an ant colony orderly: the network has its own criteria for what to do next, the logic behind which humans often cannot fathom. The challenge is actually even more impossible, for in nature, ant colonies consist of the same species. In networks-as-ant-colonies, the database transmissions and e-mails are all different species. Some are compatible, some not. Most are friendly, some hostile, a few the equivalent of anti-personnel mines.
In order to control data communication networks, Top Layer Networks, a new company located in Westborough, Mass., developed the AppSwitch. The software automatically sorts network traffic according to both a user's priorities and the application that is generating the messages. It penetrates up to Layer 7 of an e-mail's headers, for example, which can mean translating the code of seven different “species” of applications and priorities.
The prototype for the AppSwitch 2000 was written on top of a Red Hat 5.1 Linux system. The Top Layer executives assessed Linux as too rich for the application, however, and chose to program the fielded system in Ada95. The GNAT compiler from Ada Core Technologies (ACT) allowed a seamless and fluid transition from Linux to an Ada95 system. The New York City company's stable of Ada compilers is all open-source software and available at http://www.gnat.com/.
Linux was chosen as the prototype platform OS because ACT targeted a compiler to it. Also, the Top Layer software engineers are familiar with UNIX-like operating systems. The operating system for the Back Engine of the AppSwitch is a modified version of the GNAT runtime system. As part of its modification, Top Layer also created an OS kernel that supports a POSIX interface, which Linux also supports.
After the simulator proved feasible, Ada95 was used for 90 percent of the AppSwitch's code, with the remaining ten percent being low-level commands implemented in C and assembly. Top Layer still uses Linux as the “front-end” console, i.e., the gateway between the AppSwitch and the software engineers' workstations as they further develop and test the application's next release. A new AppSwitch 2500 and 3000 are expected in fall 1999 and winter 2000.
Does building a network in Ada, a field which C dominates, make sense? Top Layer evaluated their options and decided “yes”.
First, the AppSwitch software is expected to have a long lifetime. The programming language must be portable among different target machines and generations of target machines, including those not yet invented. In other words, it must be a national and international standard, as Ada is. Yet it must not make their users or themselves hostages to one target or compiler. Again, Ada meets that need.
The Ada standard is tweaked through the following bureaucracy. The ISO's Working Group 9 responds to software development tool vendors' comments and arguments about the standard's interpretations of Ada. If the WG 9 votes on the side of the vendor, the change is reflected in a Validation Test Suite. If a compiler passes the majority of the tests, it receives a certification from the Ada Resource Association (ARA), which is an organization of software development tool vendors who cover over 90 percent of the Ada market. When an developer buys an ARA “Certified Ada” compiler, she knows the compiler has been tested to interpret “standard” Ada. The interpretations have been agreed upon internationally.
Top Layer's second criterion in selecting a language was that the software must work right the first time and continue to execute in the presence of faults and software reconfiguration. The company judged Ada to have the best combination of language features for the two requirements of high reliability and portability, including strong typing, object-oriented features, multitasking and exception handling.
Ada is not “another damned acronym”. The language is named after the first published computer programmer, the Countess of Lovelace, Ada Byron King. Yes, that Byron; she was the poet's only legitimate daughter. A mathematician, she worked with Charles Babbage on the Difference Engine, and prophesied the computer's many applications, such as writing music, calculating variables, etc., as well as the potential of reusing code, which the Ada language was designed to fulfill.
GNAT was chosen as the tool chain because it is based on GCC and GCC is targeted to the ARC processor. The Ada tools are hosted on Solaris and Linux and targeted to the Motorola MPC 860, the ARC and x86.
The AppSwitch is a multiprocessor system. The software architecture is designed to take advantage of Ada95's distributed system features as the implementations mature. The AppSwitch analyzes network application traffic and enforces quality of service policies automatically, i.e., with no human intervention and no changes to PCs or servers. The AppSwitch coordinates both hardware, to examine and direct traffic at wire speed, and software, to guide the switching and the network's performance. Clients' business objectives decide the AppSwitch's policy criteria, a process that occurs with quotidian emergencies and projects, yet proves evolutionary as the business's function and size change. AppSwitch immediately benefits a business by enforcing criteria that allow the network to resemble a human messaging system more than an ant's.
With its integrated hardware and software architecture, the AppSwitch lets a network administrator monitor and understand all the elements of the network. The administrator adapts the network to changes through the AppSwitch's web-based graphical user interface.
The hardware and software components that constitute flow classification technology are at the core of the AppSwitch. With this technology, the AppSwitch goes deeply into the incoming packets, up to the application's Layer 7 headers, to determine the packets' source, type and destination addresses. The pre-configured Application Definition Library (ADL) supports the AppSwitch flow classification. The ADL is built into every switch and holds the rules for flow classification and prioritization.
When the first packets of a flow arrive, the AppSwitch uses a connection-setup-process subarchitecture to select the appropriate “tailored” profile in the ADL. It also establishes pertinent information on the new flow, which is captured in the session database. After that, the AppSwitch routes remaining packets in the flow through the coordination of the session data and the selected profile. The state data captured by each flow in the session database can intelligently adjust the quality of the service as needed to upgrade or downgrade priority. On the output side, thousands of queues are available to expand the number of flows and priorities among those flows.
The AppSwitch works behind a router or in a LAN environment. Incoming packets are subjected to proprietary “triage” by the media-access-control function (MOM) components to broadly categorize the packets. It also puts the packets into a canonical form that is data-link-independent. From there, the “categorized” packets flow over the TopWire to the Forwarding Engine (FE Chip + QM Chip) where the application control is performed.
Ada was applied in an uncomplicated manner to both the Forwarding Engine architecture and the Background Engine. While meeting deadlines is important, the AppSwitch is not strictly a hard real-time application. Top Layer's engineers eliminated some restrictions on the software from a traditional real-time application. They therefore permitted greater flexibility in the software architecture to accommodate the dynamic nature of data communications.
Ada contributed to Top Layer's success in meeting an aggressive nine-month development schedule. That is fortunate, for the decision to program in Ada met with resistance from the predominantly C shop. The engineers' learning curve steepened and grew icicles as they wondered how “Ada” will read to future employers looking for the letter “C”.
Linux users everywhere will sympathize with management's challenge to make Ada not only palatable but also preferable to the reigning zeitgeist of Microsoft, which includes C and C++ as well as the Windows OS.
They succeeded. Once Top Layer corporate executives decided Ada was the best choice, they stuck with it. Their commitment to Ada was top down and unwavering. They hired a full-time Ada expert, Mike Kamrad, to provide intensive internal training in and advocacy of Ada. The result was a personal education for all the software engineers. Finally, Top Layer took extra steps to make Ada work technically on the target systems, thereby providing a trusted and well-performing tool chain and avoiding further resistance and frustration to learning.
Like the AppSwitch itself, the engineers found that Ada speaks a sensitive, sophisticated and robust language. Ada enabled Top Layer to create an Esperanto for network administrators. The AppSwitch allows them to interpret the e-mails and transmissions' unique signals into the international and human-friendly language of graphical interfaces. The goal is to direct the multitudes of unique ant-like bytes into the tunnel-like channels that will most efficiently build and strengthen a business. The AppSwitch makes that goal possible, but only because Top Layer, in using Linux and Ada, bushwhacked their way off the worn trails of network switching systems.
Free Webinar: Hadoop
How to Build an Optimal Hadoop Cluster to Store and Maintain Unlimited Amounts of Data Using Microservers
Realizing the promise of Apache® Hadoop® requires the effective deployment of compute, memory, storage and networking to achieve optimal results. With its flexibility and multitude of options, it is easy to over or under provision the server infrastructure, resulting in poor performance and high TCO. Join us for an in depth, technical discussion with industry experts from leading Hadoop and server companies who will provide insights into the key considerations for designing and deploying an optimal Hadoop cluster.
Some of key questions to be discussed are:
- What is the “typical” Hadoop cluster and what should be installed on the different machine types?
- Why should you consider the typical workload patterns when making your hardware decisions?
- Are all microservers created equal for Hadoop deployments?
- How do I plan for expansion if I require more compute, memory, storage or networking?
|Dynamic DNS—an Object Lesson in Problem Solving||May 21, 2013|
|Using Salt Stack and Vagrant for Drupal Development||May 20, 2013|
|Making Linux and Android Get Along (It's Not as Hard as It Sounds)||May 16, 2013|
|Drupal Is a Framework: Why Everyone Needs to Understand This||May 15, 2013|
|Home, My Backup Data Center||May 13, 2013|
|Non-Linux FOSS: Seashore||May 10, 2013|
- RSS Feeds
- Making Linux and Android Get Along (It's Not as Hard as It Sounds)
- Using Salt Stack and Vagrant for Drupal Development
- New Products
- Validate an E-Mail Address with PHP, the Right Way
- Drupal Is a Framework: Why Everyone Needs to Understand This
- A Topic for Discussion - Open Source Feature-Richness?
- Download the Free Red Hat White Paper "Using an Open Source Framework to Catch the Bad Guy"
- Tech Tip: Really Simple HTTP Server with Python
- Home, My Backup Data Center
- Please correct the URL for Salt Stack's web site
1 hour 42 min ago
- Android is Linux -- why no better inter-operation
3 hours 57 min ago
- Connecting Android device to desktop Linux via USB
4 hours 26 min ago
- Find new cell phone and tablet pc
5 hours 24 min ago
6 hours 53 min ago
- Automatically updating Guest Additions
8 hours 1 min ago
- I like your topic on android
8 hours 48 min ago
- This is the easiest tutorial
15 hours 23 min ago
- Ahh, the Koolaid.
21 hours 2 min ago
- git-annex assistant
1 day 3 hours ago