• To examine the issues facing network managers today
  
• To highlight the limitations of current technology and approaches to network management
  
• To determine and test the depth of the OpenRiver feature set
• To test the “intelligence” within the software
  
• To test the RiverSoft product claims within a controlled environment – especially the accuracy of the discovery and fault detection features
  
• To test interoperability within a multi-vendor LAN environment
  
• To consider the long-term possibilities for OpenRiver and RiverSoft’s underlying technology and what it claims is the world’s first true NMOS (Network Management Operating System) concept
  
• To study the scalability and performance of OpenRiver in a live, customer environment

Do you want to hear a good joke?

There are many, many network managers out there who really believe they are actually managing their networks. No really! These are those guys who know they have some kind of SNMP management console sat somewhere – usually in the machine room or some reserved area which is largely personnel free – flashing red across the screen in synchronisation with users ringing the support desk to moan about a loss of connection. And their pagers are switched off. But their network switches and routers have RMON support so really they’re completely covered, honest. Funny, then, about all the downtime their user base has to put up with.�

The problem is that it is all too easy to pay lip service to network management; just tick all the right boxes – SNMP, yes: RMON, yes: Anti-Virus, yes… and so forth. But as anyone whose network is recovering from the Love Bug virus as we create this report will tell you, Anti-Virus products cannot exactly be described as totally effective. And much the same applies to network management in its basic form. With the combination of the Internet and switch-based LANs now dominant, the role for SNMP and RMON in their raw state is more impossible to fulfil than ever, because they were not designed with either of these scenarios in mind. Especially with respect to e-business and e-commerce,�

RiverSoft claims, rightly, that a network must have true 24 hours a day, seven days a week availability and that the answer is not simply to continually increase bandwidth. A key issue here is complexity. The more complex the network, the more complex the administration of the network management tools and the more network management personnel needed, is the argument. The company commissioned Banner Research to undertake a recent study which concluded that 92% of network managers say they are unable to guarantee service or manage their e-business network satisfactorily because of the issues of too rapid growth and resulting complexity of their network infrastructure. This is clearly not acceptable but, at the same time, is entirely understandable, given the often-severe technical limitations of “traditional” network management tools.

The key problem is in the inherent limitations of the SNMP and RMON standards on which network management systems are primarily – and in some cases, exclusively – based, despite them being all but ubiquitous across networking. Not that RiverSoft is able to work outside of these “standards”; indeed it is partially confined by them like other vendors, but it is very much a case of making the most of what you have available to work with. In the view of the NSS Group, there has been far too much “lip service” paid to SNMP and RMON support with too little effort made to provide management tools that – while supporting these standards – do move the science of management forward. However, RiverSoft is attempting to move network management forward, albeit within the aforementioned constraints, and this approach should be applauded.

Ironically, the networking industry fought for years to ensure that an industry standard for network management came about and that vendors thereafter adhere to international standards. The philosophy being quite clear – users should not be held ransom to a proprietary technology as a result of a “one off” decision.� Rather, the customer should be given the opportunity to constantly test one supplier against another, secure in the knowledge that there was always an option to switch suppliers in the event of poor quality of product or service. So SNMP (Simple Network Management Protocol) was borne out of this philosophy having won over the confidence of customers when compared with a plethora of vendor-specific proprietary management solutions and the ISO (International Standards Organisation) supported, but excessively complex, CMIP (Common Management Interface Protocol) standard.�

As a result, SNMP became the absolute, internationally accepted standard for managing network devices. And it was followed by RMON (Remote MONitoring), a subset of SNMP intended to provide “wire management” statistics and facilities, based around nine statistics groups designed to inform the network manager of the absolute status of any device.� Being entwined with SNMP gave RMON a status advantage over any other mechanism in that it was immediately acknowledged as an international standard. And when measured in terms of support, RMON has undoubtedly been successful, as pretty well every network hardware vendor incorporates it within their offerings.

This, in turn, gave the user-base the industry standard solution it wanted. However, as is so often the case with standards that are agreed between vendors with conflicting proprietary interests, RMON�– like SNMP - is a compromise solution which falls short in a number of very important ways. First, the amount of information that it provides is not complete enough for network managers and administrators who are faced with resolving complex problems at a distance, a problem the Internet explosion has exacerbated.�

Second, the mechanisms employed for data retrieval to the central management console approach typified by something like HP’s OpenView are slow and very bandwidth inefficient. And third, despite being a defined international standard, a very large percentage of network hardware vendors also do little more than pay lip service to RMON support, in the sense that they implement only a fraction of the standard. Typically they will support the first four of the nine RMON groups – sufficient to be “RMON compliant” without having to invest in the kind of hardware levels required to effect a complete RMON solution. While RMON and SNMP are here to stay – in the mid-term at least – and therefore need to be embraced, a network management system needs to offer far more than just the basic levels of support for these standards, in order to be of any real use. It is exactly these kind of true, value-add features we are looking to find in OpenRiver.

Just when you think you’ve heard it all in the networking industry, another new phrase or acronym appears out of nowhere. In this case, the subject matter is network management and the acronym – NMOS – is intended to introduce what is being proclaimed as the world’s first Network Management Operating System. The idea is a RiverSoft invention or innovation, depending on how you view it, but it relates directly to the company’s underlying technology – the foundation of the current OpenRiver product and products to come. Its’ OpenRiver fault management product we’re evaluating in this report is its first example of the technology which underpins the NMOS concept – a new way of looking at network management. As we’ve already noted, RiverSoft’s argument is that with the combination of the Internet and switch-based LANs now dominant, the role for “classic” network management platforms is more impossible to fulfil than ever, because they were not designed with either of these scenarios in mind.�

The idea behind the RiverSoft philosophy, then, is to both increase the intelligence and reduce the complexity of the management system in order to create a platform which is capable of supporting a contemporary IP environment. Great in theory, but how does it work it practice? The company is all but obsessed – and rightly so - with what it calls an “interventionless” infrastructure – a management system which tracks and record all network changes as and when they occur, but without impacting on the network or interrupting its availability.�

Riversoft’s own claim is that no Internet based network has effective network management that can cope well with rapid change, and there are not enough skilled network management people in the world to satisfy the demand right now. A new approach to solving the change management process is needed and this will not come from the traditional vendors. This is a bold claim to make, even if there is reason enough behind it. After all, HP’s OpenView is seen as a near de facto standard network management platform, if not an operating system per se. Equally, Computer Associates with Unicenter TNG and IBM-owned Tivoli would also propose their offerings as true, vendor-independent network management platforms, as close to an underlying operating system as you can get. Even Microsoft might claim its SMS is an operating system-like management platform, but then no one would take them seriously anyway. But the RiverSoft product still needs an operating systems of sorts, such as Sun Solaris, to run on top of. So it is not a self-contained operating system environment and nor is that RiverSoft’s aim.�

What, then, is the company’s reasoning behind claiming to have created the world’s first NMOS? Again, to some extent the answer lies in the perceived – and claimed – limitations of existing products and, more importantly, the inherent limitations of the SNMP and RMON standards on which they are primarily based. Here, it is not so much a case of being the aggressor against something like HP’s OpenView, but more being seen to compliment it. In this way it adds in the very ingredients now needed to support the new wave of IP networks, such as faster polling and more accurate node discovery and root cause analysis.

Inefficiency is a key problem here and one that RiverSoft has looked to resolve with OpenRiver. Many are the examples of networks being brought to their knees, not by something like SAP, SQL Server or even e-mail, but by network management traffic itself. For example, RiverSoft itself recently dealt with a customer who had found that 70% of its network traffic was purely network management traffic. The NSS view is that IT managers have been confused and – to some extent - fooled by the seemingly all-encompassing SNMP/RMON approach. Put another way, the NSS Group believes the current crop of network and systems management tools are simply inadequate and huge amounts of cash are being wasted, spent on a supposed solution which simply does not do the job intended. Moreover it can be argued that they are limiting the customer’s ability to quickly change the network as business demands new services in order to remain competitive.

In theory, at least, the RiverSoft approach appears to make sense to our eyes. But just to make sure, we’ve spent several weeks with the system in our labs and also spent time with a live customer, Carrier1, to see how the product functions in the real world.�

Let us start by stating the obvious, because it forms the basis of our test requirements. Networks are now a critical part of most businesses. Some companies, most even, live and die by their network operations. But, and in far too many cases, networks are incredibly expensive to maintain and extremely unreliable. On this premise, the feature set of OpenRiver appears to have been designed. And on this basis the company developed its OpenRiver product to isolate only ‘root-cause’ problems and present these to support staff in a usable format in advance of users ringing with complaints about downtime – the opposite of what usually happens in the real world.�

In many ways, even though RiverSoft doesn’t necessarily see it this way, given that it has announced true integration with OpenView, in our view OpenRiver provides a direct replacement for HP’s product. That is to say it maps and maintains the network, monitoring the state of all nodes making up that network – which sounds like classic SNMP manager material. However, RiverSoft has a number of feature areas it is claiming significant improvements in over past products from HP and others. The first point the company makes is that it maps and maintains the “true connectivity” of the network, defined as the “topography” rather than the topology of the network. The product adapts its topography based on configuration management events received from network devices, the aim being to ensure, at any time, that its’ network connectivity map will be genuinely 100% accurate.�

It achieves this using a modular distributed discovery system, linked to a directory enabled information model, and an active-object database. Between them, these tools establish device connectivity (physical and mesh) for both layer 2 and layer 3 protocols, then map this accurately as what RiverSoft terms “instantiated Active-Objects”, meaning registered as active on the database, from which point it will be tracked by the software.�

In use the product revolves around a single screen which clearly highlights alerts and alarms as they arise, including an instant check on their severity. Once an alert is raised, it is then possible to carry out a number of tests against the problematic network node from the management console itself. These include the generic “pinging” of the device, as well as a clever route trace which shows you how many and what hops are between you and the device in question. In typical SNMP manager type fashion you can drill down into the network by clicking on a network, segment, device, or sub-device to get down to the next level of detail. Just how much detail is available depends on whether the products’ SNMP MIBs are compiled – an automatic process - into the OpenRiver database, or whether it is having to use whatever information is available from the network node.�

A classic problem here with most management systems is that hardware vendors produce proprietary MIB extensions that cannot be picked up by generic SNMP management systems. To combat this, RiverSoft has worked with a number of the major vendors to directly incorporate their MIB extensions into the database so that the complete details of a product are available to the management console. This list of fully supported devices is growing constantly.

We were intrigued by what we can only term as the “intelligence” within the OpenRiver product. This is what might be termed the difference between paying lip service to de facto industry standards and actually trying to provide tools that really do the job. RiverSoft claims its’ product – and overall focus – very much concurs with the latter, rather than the former approach. The key word here is “intelligence” – in so much as providing the software with the ability to react to a given scenario in a way that makes the management of that network easier.�

For example, on discovering a problem on the network, a classic SNMP manager will identify that a problem exists (“eventually” in some well documented cases!) and then attempt to continue polling the remainder of the network devices, even if they are effectively unavailable as a result of a total device failure en route to that part of the network. Examples would be a router failure or even just a module or port within a switch. In each of these cases there might be 10’s, 100’s or 1000’s of devices which are now uncontactable because of the device failure. At this point it would make sense for the management system to understand this and not make an attempt to contact those devices – using often-vast amounts of network bandwidth – until the problem device was brought back online. Full marks then, to RiverSoft, for providing OpenRiver with a suitable degree of intelligence here. On discovering a failed device, OpenRiver effectively throttles back the network traffic, stopping polling to the failed device and beyond and informing other management applications you might have installed to do likewise until it detects that the device is back online. At this point it restarts full polling and instructs the other management applications to do likewise.

You might think this kind of built-in intelligence should be standard, but in providing it, the RiverSoft product is more the exception than the norm, sadly. And there is a very real reason why management traffic control is important. A combination of the classic “unintelligent”approach, plus the limitations of SNMP and RMON in themselves has shown us that bandwidth utilisation can easily be 20 times what it might be if some intelligence was applied to the management system. So not only does the inherent “smartness” make the network more manageable and more efficient but it might actually prevent it from grinding to a halt because of excessive management traffic and there are many well-documented cases of this happening, crazy though it may seem.

An important point to make here is that OpenRiver does not attempt to intelligently diagnose a problem, but purely identify its existence as accurately and as quickly as possible. For many years now, some vendors have attempted to integrate expert analysis of a problem into their management products, usually based around some kind of rules-based expert system. No one such product in history has yet been able to carry out this kind of analysis truly successfully however. Maybe for this reason, RiverSoft has made no attempt to provide analytical tools within OpenRiver, at least to date. But then knowing what you’re good at and what you’re not so good at is a worthy virtue to possess. Like most networking product vendors, RiverSoft makes a number of bold claims about its product, many of which have already been highlighted in this executive summary. Not least of these is the “accuracy” of all elements of the system – both in terms of its discovery and its fault-finding functions. Key to this is OpenRiver’s ability to differentiate between Layer 2 and Layer 3 devices, in order to build up the network map step-by-step and, in so doing, pin-point precisely where every node is within the network.�

During testing we were able to witness the methodology used, running several discoveries with a change of network layout between each discovery – adding and removing nodes - and it did create a very accurate topography of our test network each time. The speed of discovery was also very impressive to that extent that the whole process of starting the software, performing the discovery, creating the maps, defining the poll methods and alerts, was finished within a matter of minutes.

Numerous network problems were “created” from simple PC NIC failures to entire switches being powered down. In all cases the problems were identified within seconds and the software accordingly stopped attempting to poll the now unreachable parts of the network until the primary or root cause problem was addressed. Once the problem was cleared the whole network immediately became available for polling again.�

Interestingly, OpenRiver identified network nodes we weren’t aware of – two notebook computers we’d “temporarily” attached to the network via DHCP-sourced IP addresses and forgotten were still attached. Put this controlled-lab scenario into a customer scenario with perhaps 50,000 network nodes and this ability to provide a true “inventory” of the network is a, perhaps incidental, but valuable feature of the product. Moreover, using the route-finding function it lets you know exactly where the product is, hop by hop, just in case you don’t trust the network map, though this was always accurate during our tests.

Another area we wanted to test was that of multi-vendor compatibility. Within the NSS labs the OpenRiver system was managing a network that included layer 2 and Layer 3 switches from Extreme Networks (Summit 48 Gigabit Ethernet Switch), Foundry Networks (FastIron WorkGroup Ethernet switch) and 3Com (Superstack II Ethernet switch), as well as a Layer 5/7 Switch from ArrowPoint Communications, the CS-800, plus racks of servers and workstations from HP, running NT4 with SNMP agents loaded. Every device was identified correctly and, where MIBs were compiled into the system, the full extent of support by OpenRiver could be seen. Overall, we had no problems with any network node not being identified and managed.�

As part of the “test” process we wanted to see how OpenRiver performed in a live, customer environment.�

For this purpose we chose Carrier1, a “new wave” service provider whose network management operations are based in London’s Docklands. The company describes itself as a pan-European facilities-based provider of end-to-end Internet, voice, bandwidth, data centre and access solutions.�

Carrier1 is headquartered in Zurich, Switzerland.� The Carrier1 networkwas brought into service in September 1998 as a fully redundant, pan-European network.� Since then, multiple transatlantic trunks have been added. Carrier1 provides its clients with carrier-grade transport and network solutions as well as end-user ready value added services. Carrier1 clients then brand and market these solutions and services to their respective end-users.�

This means that in practice it an already large and ever-growing IP infrastructure which is absolutely critical to the operations of the service provider.�

Put simply, the network cannot afford to go down – it must be managed in every sense of the word. The absolute aim of Carrier1 from day one was to provide managed bandwidth, to get into what Matthew Bird, director of Internet operation for Carrier1, describes as “big IP managed services”.�

Already the company has over 80 major customers including BT, Cable & Wireless and Belgacom, for example, supporting in excess of 3.5 million subscribers online at peak times. Currently Carrier1 operates in 14 different countries across Europe as well as the US and is in every major International exchange. The service provider currently has customers connected in London, Amsterdam, Brussels, Paris, Frankfurt, Berlin, Hamburg, Munich, Dusseldorf, Vienna, Zurich, Geneva, Milan, Copenhagen, Stockholm, Manchester, and New York. Carrier1 completed its pan-German network build in July 2000 and is rolling out additional points of presence in 13 French cities and nine German cities..�

Recently it announced the construction of a 32km high-fibre-count Paris metropolitan network, due for completion in August 2000. When completed it will interconnect major tele-houses, carrier hotels, and telecoms access points in Paris and will cost less than $5 million. This new network will enable Carrier1 to cut local tail costs and significantly improve customer-provisioning time. It closely follows commencement of the 120-km city ring now under construction in Amsterdam which is scheduled for completion this summer. It is on track to complete 20 such city rings by the end of 2001.�

These networks will connect major telecom centres as well as at least 20 full-service data centre facilities, known as IT Mega-Centres being developed through its DigiPlex joint venture. The Carrier1 network currently includes almost 11 000 km of contracted optical fibre and uses a Nortel DWDM and SDH transmission layer, a Cisco Powered IP network, and Nortel voice switches. Carrier1 operates a Tier1 grade Network Operations Centre 24 hours a day/365 days a year.

Examples of Carrier1 services include managed bandwidth solutions as well as a wholesale pan-European voice service called GlobalVoice that meets the long-distance and international voice requirements of established and/or emerging telecommunications operators, multinational corporations, mobile operators, and ISPs looking to provide value added voice services to their customers. Another example is a portfolio of global Internet transit and IP based connection services to help established ISPs, non-incumbent operators and new entrants into the Internet market to offer high-capacity Internet connectivity. Features include:

• a dedicated one-to-one connectivity to selected parties via main Internet Exchanges (Europe and US);
  
• access to POPs in London, Frankfurt, New York, Amsterdam, Paris, Vienna, Geneva, and Zurich;
  
• dedicated capacity on IP tunnelling, or, if desired on SDH and access speeds from 2Mb to 155Mb;�
  
• customised Service Level Agreements are available and online reporting and service statistics form part of the package.

Currently the Carrier1 network is based around 106 major routers, primarily Cisco 7500 and 7200 series. Bird describes the network as being “very busy” with a lot of traffic at all times of the day and therefore a true test of any network management system.�

Recently the company took OpenRiver from RiverSoft to manage its entire network. Carrier1's aims in implementing OpenRiver were many and various but at the heart of these was the need to automate a significant percentage of the company’s network management operations. The automation process has two obvious benefits, the first is that it result is in quantifiable service-level improvements to pass on to customers and the second is that it frees up technical staff from having to perform expensive, time-consuming basic housekeeping tasks.�

Originally Carrier1 had two products in place for providing the network topology mapping and fault detection functions. It replaced both of these products with just the one in OpenRiver. This in itself was a significant benefit in terms of reduced support and training overhead but there was another huge benefit in the move to OpenRiver.�

Key to the success of OpenRiver within Carrier1 is its ability to – following the original network discovery - continue to search for any changes to the network, instantly incorporating them into the map in order to maintain an up-to-date model of Carrier1's infrastructure. Unlike conventional discovery/mapping tools, including the previous products Carrier1 had installed, with RiverSoft this process demands no network operator intervention. Whereas any network topology changes had to be added manually into both of the previous systems, now OpenRiver goes out and automatically discovers the changes or new devices, queries the network and revises the network topography accordingly. Moreover, while the implementation of the original products was described as forever ongoing, so never finished, with OpenRiver the implementation was all but immediately effective.�

It has been estimated that the implementation of OpenRiver will save Carrier1 at least 78 days a year in network topography configuration alone. The time the company saves on identifying faults on the network is crucial to ensure its network operators can focus completely on customer service and fixing problems before outages occur and customers are actually affected. Bird claims that RiverSoft eliminates most of the time and cost associated with operation support systems, management and fault identification.

"RiverSoft's interventionless approach also reduces training costs. Together this minimises the impact of network management on the business as a whole," he explained.

RiverSoft’s entire approach to network management should be applauded.

The OpenRiver product, while a complete and successful product in its own right, is clearly just a part of the long-term strategy of RiverSoft. The focus appears to be on providing a true network management platform for third-party developers to adopt, much in the same way that HP’s OpenView has been adopted to date, regardless of that product’s limited capabilities in the real world.�

RiverSoft is clearly not taking Hewlett-Packard head on, given a recent integration policy announced between the two companies, but the NSS believes the product gives RiverSoft exactly this opportunity. This would be a bold – and perhaps foolish - step to take, hence the company’s positioning of itself and OpenRiver but, regardless, it should be supported because RiverSoft has got the basics absolutely right. Network management is all about accuracy, speed and efficiency and here OpenRiver scores on all counts. We like the heart of the product and the idea of an “interventionless” infrastructure – a management system which tracks and record all network changes as and when they occur, but without impacting on the network or interrupting its availability. And here is the real value of OpenRiver as it stands today. It appears to be truly efficient in operation and this immediately marks it down as outstanding within its genre; an area of networking where 9-12 month timescales – at best - for basic implementation only are not unusual. OpenRiver, in contrast, works from the moment you set it running, a claim confirmed by the visit to Carrier1.

Overall, then, the OpenRiver product has more than lived up to its expectations within the NSS labs and those expectations were not trivial to begin with. As such it is highly recommended.

“The bubble is about to burst”.

How often have we heard that phrase recently with respect to the .com world, notably those companies looking to sell services over the Internet? E-commerce in particular can be argued as being a frail, flavour of the month, concept rather than a genuine step forward in technology–based services, even if it isn’t in practice. But the bubble-bursting scenario has been equally applied to e-business - business to business communications over the Internet - and even the Internet itself. With respect to the latter, such talk is, bluntly, ridiculous. The Internet is now a fundamental part of the world’s infrastructure, economic and otherwise, but the strength of the argument shows how fragile confidence in Internet-based business really is.�

The truth is that, in almost every case, there is an alternative to an Internet-based service, be it a shop, a bank, or whatever, in its physical form - the traditional high street branch or out-of-town shopping centre for example. What the Internet alternative does offer is 24-hours a day availability, total convenience and total coverage from a single location, usually the home or office. If you have access to the Internet then you have access to every Internet-based service in the world, security privileges allowing. These are very strong arguments in favour of computer-based commerce or e-business. As such, they validate the near obsession with companies throwing their services at and onto the Internet currently. But – and this is a big but – the benefits of e-commerce are lost the moment a web-site becomes unavailable or even if it is simply very slow in operation.�

Of course, there are tens, if not hundreds, of networking hardware vendors out there selling solutions designed to both prevent a web site from becoming unavailable and maintain an acceptable level of performance, so sustaining the appeal of e-commerce to the user of that service. The problem with networks however, is that they do go wrong. While hardware – primarily in the form of network routers, switches and servers – has got both hugely faster and more resilient in recent years, it still lacks essential intelligence. So when there is some kind of physical failure, the net result is all too often the loss of one or more services on the network before any action can be taken to remedy the problem. And if that network happens to be hosting an e-commerce service it directly translates into lost income for the service provider.�

Many are the documented cases of would-be e-commerce customers finding a target web-site down and shifting their custom to a competing service provider whose web-site was up and running. Not only that, but the chance of those customers permanently making the shift to the new supplier is extremely high. This is a market with high elasticity. So one moment’s network downtime can have a lasting effect on lost income potential. The same applies in the situation where the user finds one web-site performing too sluggishly for their patience to withstand and therefore moves on to a competing site that is performing better. While this latter scenario could be simply down to poor web-site design, equally it could be the result of a problem somewhere on the network, a problem which the IT staff need to be made aware of as quickly as possible.

How, then, does an e-commerce or e-business site avoid the problems of complete downtime or poor web site performance related to network failures? The answer lies not in purely adding ever-faster components, but in adding some intelligence to the network in the form of network management to control those components. Indeed, as the speed and complexity of networks rises at a near exponential rate and the Internet takes on ever greater importance, so the role of network management becomes ever more critical to the success of e-commerce and e-business ventures.�

Not that there is anything new to the idea of network management. Most networks have some form of network management system installed on them and have had for many years. The problem lies in the difference between genuine network management and merely paying lip service to the concept. All too often some kind of basic management is put in place and – in reality – is not used. More importantly, the basic standards and technology that underpin most network management systems are, in themselves, incapable of truly managing an e-commerce network. They were never designed with an Internet, IP-based network in mind, nor web-based e-commerce and e-business systems.

What is required then, is a new approach to network management that is designed with modern IP-based, e-commerce and e-business networks in mind. At the same time, the existing network management standards must be adhered to, ensuring compatibility with existing network components. RiverSoft believes it has the answer within its underlying technology strategy, based on the changing network scenario and the new requirements “e-networking” brings to a network management system design. If we again quote RiverSoft’s finding that 92% of network managers say they are unable to guarantee service or manage their e-business network satisfactorily because of the issues of too rapid growth and the resulting complexity of their network infrastructure, then this represents a grave problem for the “E” world. It is very clearly unacceptable in any working environment, but in the context of e-commerce it is a killer.�

The more complex the network, the more complex the administration of the network management tools and the more network management personnel needed, is the argument. The idea behind the RiverSoft technology, then, is to both increase the intelligence and reduce the complexity of the management system itself, in order that it might do the necessary job – helping to keep the network up. The result of this line of thinking is what the company claims to be the world’s first true network management operating system or NMOS for short; a true management “platform” which underpins the network operations and is designed specifically for IP-based, Internet-oriented networks such as e-commerce and e-business sites.�

Key to the concept is accuracy and speed of detecting both faults on the network and changes in its configuration. The product continually adapts its topography – the “map” of the network - based on configuration management events received from network devices, the aim being to ensure, at any time, that its’ network connectivity map will be genuinely 100% accurate. On a small network these values are praise-worthy enough; on a network which effectively comprises at least a part of the Internet they are, quite simply, essential.