Recently, Cisco announced its next-generation networking architecture, application-oriented networking (AON), which puts Web services at the very heart of the enterprise network.
The announcement was clearly a step forward for Web services. But how useful will AON really be for Web services architects? Is there a real benefit here -- and if so, what price will have to be paid? That's what we'll look at in this second part of a two-part article.
Is AON really new?
AON represents a major shift in the way that networks interact with Web services. With the architecture, "intelligent application message routing" will be embedded directly into the network hardware. So network hardware will include intelligence and be application-aware, and will work on a message or application level, rather than on a packet level, the way it currently does.
This means, of course, that the hardware will work directly with XML. In one sense this is nothing new, notes Frank Dzubek, president of Communications Network Architects, an analyst firm in Washington, D.C. XML-aware hardware has been available for some time, from vendors like Datapower, which makes XML-acceleration hardware and XML-aware network infrastructure. In fact, Dzubek contends, "all that Cisco is doing at the moment is XML acceleration and some application-specific acceleration." Jon Oltsik, senior analyst for information security at Enterprise Strategy Group, concurs, and says that "AON is evolutionary, not revolutionary."
But even if the changes aren't revolutionary, they will give network managers greater control over applications and their networks, allowing them to add Quality of Service (QoS), class of service, application acceleration, security and more. In the words of Oltsik, it "adds more intelligence higher up the network stack," which is a great leap forward for network architects and managers.
AON and applications
But AON might not be as simple as it seems at first glance, and there are a great many implementation issues that need to be resolved.
First and foremost is whether it's enough for a network to be XML-aware. Won't the network have to be aware of specific Web services applications, not just XML, if it's going to accelerate and handle applications? And if so, how can the network possibly be aware of the countless thousands of Web services applications, most of which haven't even been written yet?
Dzubek says that he expects at first the network won't be truly application-aware, simply because there's no technology that can yet give it that capability. That's one reason why the Cisco announcement included IBM, SAP AG, Actional Corp., Tibco Software Inc., and VeriSign -- Cisco hardware will work closely in concert with applications or tools from those companies.
He counters that "there are millions of applications and doing something for SAP applications has nothing to do with a financial application or telemetry applications."
But in the future, Dzubek says, he expects that AON will be application-aware, once Web services-oriented grid protocols come into widespread use, particularly Globus 4.0. Additionally, he says that AON will have to somehow make use of the many industry XML schema being developed. When all this goes into effect, he says, AON will become far more useful than it is today.
AON and Web services standards
Anyone familiar with Web services standards knows that there's a whole alphabet soup of them, and many are still evolving. It's difficult for developers and architects to keep track of them all, and write their applications accordingly.
Will AON mean that a whole new class of standards will be required to write Web services applications? Will there have to be an AON Web services standard that developers write to?
There's good news here -- no one interviewed for this piece believes that the Cisco move will spark a new generation of protocols.
However, one potential issue is whether applications will have to be written specifically to AON and the network in order to take advantage of AON features. Cisco says no, but Randy Heffner, Vice President in Forrester's Application Development & Infrastructure research group, notes that "from the application side, it's going to take a while to figure out the implications of this. The key question is, "do I have to rethink my networking messaging processing?"
Implications for Developers and Network Architects
One thing is clear about AON -- network architects and Web services developers will likely have to adapt what they do, at least to a certain extent, if they want to make use of the technology. Heffner says that "because the network will now be message-by-message, rather than packet-by-packet, network managers might initially stumble over how to use AON."
Developers and lead architects will have to take AON into account as well, for example, when working with network paths and with application modeling. And, Dzubek adds, "In all probability, in the future when you write an application, it will have to be made aware of the network in some way."
Despite all these potential issues, though, most analysts agree that AON is a step forward for the conjunction between networking and Web services standards. Dzubek concludes, "we're still in the early stages here. It's much like when we were in the first stages of crawling, like we were with Web services a while ago. And look what's happened since then."