TMCnet - World's Largest Communications and Technology Community



Unified Communications Magazine May 2008
Volume 1 / Number 6
Unified Communications Magazine
Richard Grigonis

Centralized vs. Decentralized Systems

One of this industry’s eternal controversies centers on the benefits and drawbacks of implementing business communications as either centralized (Centrex-like, hosted services that now also encompass voicemail and audio conferencing) or a more traditional node-by-node distributed architecture, or even an ad hoc peer-topeer set of IP phones. Other approaches are “semi-distributed”, such as call centers with teleworker extensions, or systems organized along regional, organizational and functional lines.

By Richard “Zippy” Grigonis

IP communications itself lends itself to a distributed or decentralized paradigm, since the world of computer software-powered systems enables communications clients to be decoupled form their switching platforms. Other elements — the major four logical functions being IP telephony clients, communication/call servers, media gateways and application servers — can be centralized or distributed across an IP network. The degree of centralization or decentralization implemented depends on how well a particular approach satisfies your customer service, price/performance, feature/functionality, reliability, scalability, management and business continuity requirements.

Originally, the enterprise applications were all centralized. Take for example messaging. Centralized messaging systems consisted (and in most cases still do) of a large data center that hosts all server resources, including such things as Microsoft’s extremely popular Active Directory, global catalog servers, domain controllers, and Microsoft Exchange servers. Data is hosted and managed in a centralized location regardless of whether the users are connected remotely. This contrasts with the distributed model, where users have local access to mailboxes but server administration is more complex. The advantages of centralized systems, of course, are that software upgrades can implemented from a centralized location, and that a central site can be made extremely reliable.

Today, however, there are opportunities for developers to make their mark designing cutting-edge distributed or semi-distributed systems.

For example, Recursion Software provides middleware software development solutions for the newest crop of application developers. Recursion’s middleware is used in developing intelligent applications using mobile agent technology and other complementary software solutions that offer advanced capabilities in terms of both interoperability and communication. Recursion Software has 18 patents issued related to distributed computing and agent technology, and 30 that are pending. Recently they expanded the functionality of their Voyager Edge, which leverages intelligent mobile agent technology along with interoperability between mobile agents in Java and .NET. Recursion’s products have been downloaded by thousands of customers in such diverse industries as defense, finance, healthcare, telecom, energy, and IT.

Recursion Software’s CTO, Bob DeAnna, says, “The growth of distributed systems is a direct result of the proliferation of edge devices. One of our products that has been in the market for about ten years is Voyager. We’ve positioned Voyager for what we consider to be next-gen applications. For us, next-gen applications are very much decentralized. Will such applications periodically communicate with centralized servers? Absolutely. But there is also the need in these next-gen apps where devices that come within proximity for some transient period of time — and not even physical proximity — can benefit from the sharing of information and participate in some type of shared process or transaction or objective. We see this next generation of applications coming, and the idea is getting a lot of validation from our customers. You can see such applications appearing in areas ranging from future combat systems in the Department of Defense, to Homeland Security environments, to traveling consulting teams from very large consulting organizations, to literally the shopping mall.”

“These devices that share information when in close proximity are not homogeneous in nature, although they can be,” says DeAnna. “Often, the devices are heterogeneous; you might have a mixture of servers,laptops, PDAs, smart phones, embedded devices, RFID readers, and so forth. So we’ve positioned Voyager to be the software platform that enables those types of next-gen apps. In a nutshell, we feel that the next slew of killer apps will be very decentralized. We turn every node on the network into both an intelligent client and intelligent server that can communicate with any other node or dynamically assembled groups of nodes. In our world a ‘node’ is, as I’ve said, anything from a server down to a smart phone, and now even smart tags and sensors. With Voyager, a node, such as a server or a smartphone, can turn data into knowledge, then intelligently pass it on to only the nodes that need that information creating real-time situational awareness.”

“Our platform runs on enterprise operating systems, desktop OSes, smart phones and about 16 embedded operating systems,” says DeAnna. “So that’s our vision of the next generation of applications. All of these nodes that need to share information or participate in some kind of common task or process or goal, are able to communicate amongst themselves in a peer-to-peer or peer-to-group mode, and also intermittently communicate with one or more enterprises. That generic use case is what’s defining Voyager’s roadmap.”

“Centralized servers will not go away,” says DeAnna, “but we will see the increased role of device communities on the edge that are more dynamic, more ad hoc and which collaborate in real time. Recursion Software’s intelligent, next-gen software platform, Voyager, is positioned as a great platform upon which to build these de-centralized apps.”

Even in a distributed system, however, certain functions, such as monitoring, probe various areas of a large system, but the command and control subsystem is centralized. For example, Spanlink’s SolutionWatch offering is a proactive service designed to minimize system downtime and can inexpensively supplement monitoring solutions at a fraction of the cost. As Spanlink CEO Brett Schockley explains in this month’s Q&A on page 30, if, say, your hard drive failed, SolutionWatch would detect the failure nearly instantaneously and create a trouble ticket that either goes to your IT or Spanlink’s technical assistance center, alerting them of the failure and letting them know exactly where the issue originated.

CentralControl, another Spanlink offering, is a web-based management framework that presents a unified, holistic and simplified interface to the disparate and often-times distributed resources that comprise a customer interaction network.

Aside from the convergence of the data and voice networks, Spanlink has also done considerable work in business virtualization, which are “mostly” centralized in natute. One Spanlink customer, Grainger, virtualized 500 of their store locations into a single virtual contact center, with thousands of agents across all of those sites. With Grainger the also pioneered a market-based routing algorithm enabling customers to call local resources and get local service even if the office handling the call is 30 miles away. Some companies may not realize that they have a true contact center until suddenly they’re using this technology to tie together many different sites and virtualize their business.

With any distributed system, it’s good to have some way of both accelerating the connections between components and being able to “visualize” what’s happening over systems with components scattered over a region or around the world. Akamai, for example, offers Application Performance Solutions that accelerate dynamic applications. Akamai’s Edge- Platform happens to be one of the world’s largest distributed computing platforms. It’s a network of about 30,000 secure servers equipped with proprietary software and deployed in 70 countries that relies on special algorithms to help prevent network congestion and vulnerability problems on the Internet. These servers reside within approximately 1000 of the world’s networks monitoring the Internet in real time — gathering information about traffic, congestion, and trouble spots.

Akamai’s intelligence system optimizes routes and replicates data dynamically to deliver content quickly. Akamai’s approach basically involves eliminating long routes whenever possible — by replicating and delivering content and applications from servers close to end users around the world instead of from centralized servers. Akamai calls this delivering from “the edges of the Internet”.

Akamai’s approach involves a comprehensive view of Internet conditions and the tools to control the movement of any type of content or application. The result is a growing set of interesting technology offerings designed to solve particular problems for specific types of users.

You can see some of Akamai’s data visualization tools at this site: www.akamai.com/html/technology/visualizing_akamai.html. These tools provide a real-time view of things such as current Internet traffic conditions, security, etc.

Centralized or Distributed? There are economic arguments for Small and Medium-Businesses (SMBs) to use hosted or managed services, which are the ultimate in centralized systems, but I’m fond of pointing out how many companies have a paranoid corporate culture (perhaps a healthy paranoid corporate culture) and fear to have their company’s precious data escape the corporate perimeter. (In these days of Fixed- Mobile Communications, of course, there is some confusion over just where the corporate perimeter happens to be.)

Invariably, there are many factors to consider, such as whether your business model incorporates many branch offices, whether most of your communications clients are fixed-line or wireless, how much you want super high availability and reliability (as afforded by centralized systems), how much bandwidth is available, what kind of Quality of Service can be maintained over long distances, and the overall expenditure of moving to a more decentralized system.

Richard Grigonis is Executive Editor of TMC’s IP Communications Group.

Unified Communications Communications Magazine Table of Contents

Today @ TMC
Upcoming Events
ITEXPO West 2012
October 2- 5, 2012
The Austin Convention Center
Austin, Texas
The World's Premier Managed Services and Cloud Computing Event
Click for Dates and Locations
Mobility Tech Conference & Expo
October 3- 5, 2012
The Austin Convention Center
Austin, Texas
Cloud Communications Summit
October 3- 5, 2012
The Austin Convention Center
Austin, Texas