Axiowave Networks - improving the economics of IP service delivery
September 6 2004		Email this page

Interview with Mr. Mukesh Chatter, Founder, President and CEO

Introduction

Axiowave Networks, based in Marlborough, Massachusetts, was founded in 2000 and has, since its formation, raised $130 million in funding. Axiowave develops IP (Internet Protocol) router solutions for the core metro network space and has to date launched one product on the market, the XCR128 Service Convergence Core/Metro Router.

This product is designed to allow the application of quality of service controls to the delivery of converged services over IP/MPLS (Internet Protocol/ MultiProtocol Label Switching) networks. The company markets its product to service provider customers worldwide and has to date announced one customer. Axiowave has equipment deployed and carrying live traffic in carrier networks.

Funding and managment

The company's funding has been raised from private equity investors, mainly Fortune 100 companies, rather than the usual venture capital sources. Investors include Gainesborough Investments, Argonaut Private Equity, Soros Private Equity and Madison Dearborn Partners. Mukesh Chatter, along with co-founder and chairman Ray Stata, is also an investor in the company. The board of directors reflects this scope of investors but also includes independent member Howard Janzen, president of Sprint Business Solutions.

Mr. Chatter said the management team, with the exception of newcomer Gary Massone, VP of sales, has a long history of working together within the telecommunications industry.

The economics of IP networks

IP = Invisible Profit

In the overall telecom market today, Mr. Chatter remarked, traffic volume continues to grow at between 50% and 100% annually, driven by factors such as consumer products, including digital cameras,

"Each time Sony releases a new camera with higher specification and more memory - now four or five megapixels resolution and 500 megabytes of storage - traffic levels can be seen to jump after a lag of a few months."

This traffic growth is seen to be continuing in a period when carriers have capped investment in existing TDM (Time Division Multiplexing), ATM (Asynchronous Transfer Mode) and Frame Relay networks. Mr. Chatter noted that AT&T was one of the first companies to cap investment in its existing 'conventional' infrastructure and rollout IP/MPLS-based networks that promised comparable levels of service, along with reduced capital and operational expenditures. Many carriers followed AT&T's lead,

"Axiowave has spoken with roughly seventy service providers around the world, the majority of which are adopting a similar strategy to AT&T in response to financial constraints. The expectation is that a shift to IP/MPLS will cut expenditure while enabling continued provision of the profitable premium services that were being delivered over ATM and Frame Relay networks."

However, Mr. Chatter stated, a look at the business models of the carriers operating IP-based networks quickly reveals that these expectations - in terms of both reduced expenditures and quality of service - have not been realised,

"One CFO that I spoke with defined IP as meaning invisible profit! The bottom line is that there is currently no money to be made in operating IP networks. I am aware of a mid-sized North American carrier that installed an IP/MPLS network in 2000, which states that it has, to date, failed to realise a profit from it. The comment was that the only sure thing was that running the network would lose the company more money tomorrow than it had lost today."

Falling capacity prices

For carriers today the dramatic fall in capacity pricing is considered the most obvious issue, compounded by the poor utilisation that can be achieved over IP networks, particularly for premium services.

Mr. Chatter noted that pricing for ATM and Frame Relay services has not fallen in line with services carried over IP networks for several reasons: one is the relative scarcity of service providers, which leads to the second factor, customer loyalty, due in part to lack of choice - it being unlikely that more than one or two service providers will be able to provide network connections to all of the locations that a customer requires. Lastly, enterprise customers demand and receive a defined quality of service and are prepared to pay for this.

Although some carriers were claiming to be operating IP networks profitably as recently as six months ago, it has transpired, Mr. Chatter said, that these claims were founded on accounting methods rather than the realities of operating an IP network. With the continuing decline in wholesale capacity pricing the situation is now seen to be far worse.

The wholesale price for transport of IP traffic continues to fall, having dropped by two-thirds over the past twelve months - from approximately $62 per megabit per month to a current price of around $20. Over the same period the retail price for capacity has declined by approximately one-third, from $500 per month for a T1 circuit to approximately $350 per month today.

"I am aware of peering arrangements signed one year ago where the end user is paying $20 per megabyte per month to a service provider that is in turn paying $40 per month for the capacity. This situation is having an impact throughout the industry now - many of the small and mid-tier carriers are battling with this problem. The bottom line is that there is no money being made from running IP networks."

Further illustrating the present difficulties, Mr. Chatter noted that roughly half the wholesale long-distance voice traffic now generates one half of one cent per minute in revenue, whereas three years ago this figure would have been ten cents per minute.

In addition, customers for data services typically sign two-year contracts which, as they come up for renewal, will now be priced two-thirds lower, meaning two-thirds less revenue for the carrier. Mr. Chatter pointed out that this situation is now being reflected in carrier's quarterly results.

The decline in pricing for IP services can be contrasted with the case for Frame Relay, where the time from deployment to a one-third drop in pricing was ten years, as compared to one year for wholesale, and two years for retail, IP traffic.

The true situation

Highlighting the economics of operating IP networks, Mr. Chatter said that it was only after meeting with carriers' financial, marketing and sales officers that a true realisation of the situation was reached.

He noted that personnel in different roles within a carrier tend to have different views of operations, especially as related to the economic viability of the IP business model correlated with expenditure on the network itself.

Three factors were commonly cited as affecting the profitability of IP networks:

1. The capital and operational expenditure involved in deploying a profusion of overlay networks. Many carriers are running up to sixteen overlay networks to serve different customers and provision different services, while adding a further network to roll out IP/MPLS-based services.

2. The low utilisation of IP networks, with average sustained port utilisation maintained at 30% or less for best-effort and a maximum of 20% for IP/MPLS services. Premium services, such as VoIP or video, might cut this utilisation level to single digits. Overall network utilisation levels are seen to decline further as the proportion allotted to premium services increases.

3. A third factor is lack of service differentiation - all service providers worldwide offer similar SLAs (Service Level Agreements) for IP networks.

Utilisation levels for ATM and Frame Relay networks are considerably higher, particularly with respect to premium services, making the operation of such networks inherently more profitable. Further enhancing this situation is the fact that Frame Relay networks tend to be oversubscribed, resulting in significant revenue amplification.

Flawed IP/MPLS strategy

In addition to the lower than expected revenues derived from IP networks, Mr. Chatter said carriers have discovered that they are not able, as they had intended, to transplant their business models from ATM to IP networks due to technical issues.

Within the respective network utilisation levels for ATM and IP networks, further factors have a detrimental effect on the operation of IP networks. Mr. Chatter explained that provisioning of ATM Constant Bit Rate (CBR) and Frame Relay services over a network is mutually independent - within the total available utilisation level the relative proportion allocated to each is irrelevant.

In contrast, with IP/MPLS networks, altering bandwidth allocated to one service will have an impact on other services. For example, increasing the level of best-effort traffic may adversely affect premium services. Mr. Chatter remarked that this disparity in operation of IP and ATM networks effectively changes the definition of convergence as applied to IP networks.

Discussing the differing performance of ATM and IP networks, Mr. Chatter pinpointed the main reason behind this: the fact that high capacity IP routers were generally architected prior to 2000 and were designed as large bandwidth boxes optimized for best effort traffic. At this time the industry did not envision a capping of investment in ATM and Frame Relay infrastructure and, as a result, the need to mix best-effort and premium services. He noted that Juniper's T640 product was originally designed in 1999, and Cisco's CRS-1 in 2000, although the underlying switch fabric was developed in 1998.

As carriers retrenched and focused on costs above all else with the collapse of the market, one effect was reduced investment in ATM and Frame Relay networks and a shift of interest to IP. In response, the system vendors began adding MPLS to their best-effort IP products in the belief, or hope, that the technology would boost performance to levels delivered by ATM systems and so allow a straightforward transfer of business across to the new IP-based networks. Mr. Chatter observed that this has proved to be a deeply flawed strategy.

Absolute Quality of Service

Mr. Chatter believes there is a great deal of confusion about the capabilities of MPLS having come to be regarded as a 'cure-all' for the technical shortcomings inherent to IP routers. He claimed that MPLS is primarily a signalling and reservation protocol - serving as a method to request an allocation of bandwidth over a network but neither intended nor possessing the capability to actually guarantee that bandwidth. The only way to implement and enforce reserved bandwidth, Mr. Chatter said, is via hardware, as is the case with ATM networks. Illustrating this point he remarked,

"The situation with MPLS could be compared to a government passing a new law - unless or until the police force actually go out and enforce that law on the ground it may carry little or no weight, and will certainly have no effect on anyone intent on violating it."

Mr. Chatter commented that the MPLS protocol allows for the provision of 'Relative Classes of Service' but cannot ensure 'Absolute Quality of Service', adding that it is the latter, which is critical in making ATM networks effective and ultimately viable.

"To give another analogy, when booking a flight, there are three classes of service available - first, business and economy. In choosing and paying for one of these one expects a certain level of service. When checking in at the airport the actual quality of service is totally dependent on check-in staff and the number of desks that are open for each class. The ATM router could be compared to the check-in area and staff. I believe that somehow along the way this fundamental distinction between Relative Class and Absolute Quality of Service was misinterpreted along the way."

In addition, Mr. Chatter stated, the MPLS protocol cannot enforce. For example, it will inform the system that voice traffic should be transmitted with lower latency and jitter, however this may not be preserved when mixed with large amounts of background Internet traffic. The distinction is seen to be between Relative and Absolute Quality of Service - current IP/MPLS networks are simply unable to deliver Absolute Quality of Service.

As a point of interest, Mr. Chatter noted the fact that during the 9/11 disaster, despite the largely non-functional VoIP services in certain areas of the U.S., no IP SLAs were violated because carriers generally average service levels over their entire network and over a time period, so ironing out any localised or short-term service problems. This anecdote serves to highlight the difference between Relative and Absolute Quality of Service. This very issue of "when you needed it most, you had it least" remains unsolved even today with wider and accelerated VoIP deployments.

Service pricing

Axiowave compared the pricing for a number of services in the U.S., taking best-effort IP as the baseline. IP VPN services were found to be priced at approximately one and a half times, Frame Relay at three and a half times and ATM CBR four times best-effort.

These figures were then multiplied by the over-subscription rate for each service, which ranged from seven times for Frame Relay, down to no oversubscrption for ATM CBR, and utilisation rate, to arrive at a relative cost for each. These calculations showed that IP/MPLS generates the lowest amount of revenue, roughly one quarter that of ATM and Frame Relay. This result questions the expectations that it would be possible to earn similar revenues at reduced capital and operational expenditures from IP/MPLS networks.

"I believe that this state of affairs will lead to serious questioning of the prevailing IP/MPLS business model based on best-only routers optimized for best effort. Carriers are now faced with a serious business challenge of not making money on their IP investments, while system vendors, which effectively means Cisco and Juniper, are realising gross margins of around 65% on their IP router products."

Acknowledging the problem

Industry analysts Paul Sagawa of Sanford Bernstein, and Ravi Suria of Lehmann Brothers, foresaw the collapse of the telecom market in 2000. However, the results were even worse than their predictions believes Mr. Chatter. He feels that carriers relying on delivery of data services over IP networks for a substantial portion of their business could face major problems and that concern within the investment community and the industry as a whole is growing as the effects of this issue are reflected in the carriers' quarterly results.

Some carriers, Mr. Chatter claimed, have identified the economic issues relating to IP networks, while others have not, but whatever the case, companies are unlikely to suddenly acknowledge that they have invested huge sums of money in infrastructure that is not generating a reasonable return.

Axiowave's solution

Company strategy

Positioning Axiowave in this picture, Mr. Chatter said that the company offers a partial solution to this crisis by enabling improved economics for IP service delivery, primarily through increased network utilisation rates - thus increasing revenue for the service provider. He added that by employing Axiowave's system, revenue-generating opportunities are improved, even beyond that possible with ATM and Frame Relay networks. The fundamental problem with IP/MPLS networks, Mr. Chatter explained, is not the MPLS protocol per se but the fact that system vendors opted to develop products optimised for delivery of low value, best-effort services.

The starting point for Axiowave's product was to build an IP router combining performance equal or superior to that of an ATM switch, with lower capital and operational expenditure for the carriers. Mr. Chatter emphasised that this aim has been met.

Improving network economics

The key to meeting performance goals is an ability to allocate capacity for a specific service without affecting other services, as is possible over ATM networks. System utilisation over an IP/MPLS network is thus increased to 90% or higher while SLAs are maintained. For the service provider this higher utilisation level translates as a reduced demand for line cards, WDM ports and wavelengths, and fundamentally alters the economics of running an IP/MPLS network. Mr. Chatter stated,

"Within the 90% utilisation level delivered by Axiowave's product, all traffic can be premium if required, compared to 6% or 7% in a conventional IP/MPLS network, and between 50% and 60% over an ATM network. I have customer references stating that 90% network utilisation has been attained with premium traffic alongside oversubscribed background best-effort traffic. To give an idea of what this means for a carrier, if they were to install Axiowave's product it would remove the need for deployment of a separate overlay network for the planned voice over IP rollout."

A further example of how Axiowave's solution can assist service providers to improve the economics of operating IP networks is in the handling of bursty traffic. Over conventional IP networks such traffic can disrupt premium services. Therefore to avoid this, carriers attempt to drop such traffic at the network edge, so losing potential revenue. As noted previously, Axiowave's system is able to ensure that capacity allocated to premium services is maintained, irrespective of fluctuations in best-effort traffic.

Axiowave's product also gives carrier sales teams a competitive edge, enabling the offering of ATM-grade SLAs over IP/MPLS networks. Mr. Chatter cited the case of one customer that has announced IP VPN SLAs superior to ATM CBR and Frame Relay agreements.

Further, the Axiowave solution is able to address the problem of poor quality mobile calls over IP networks. To address this issue, publicly announced customer PowerNet Global is installing a point of presence based on Axiowave's product in the Far East, which promises to enable provision of land-line quality calls over a mobile phone between the Far East and the U.S.

Regarding system deployment, Mr. Chatter explained that this is completely 'phased-in', meaning that it is not necessary to make changes to other equipment on the network or to the network itself. He added that the product represents an evolutionary play in that it can gradually be rolled out into existing networks. Expanding on this topic, he said that a service provider does not necessarily need to deploy the system throughout its network to gain the benefits - installation at major points of presence will enable the company to offer ATM-grade SLAs at key customer locations.

Product overview

Axiowave's XCR128 Service Convergence Core/Metro Router system is based on a switching and queueing architecture that makes possible the delivery of true Quality of Service over IP/MPLS-based networks. As with solutions from Cisco Systems and Juniper Networks, the product supports standard interfaces ranging from OC-3 to OC-192, and also, as does Juniper's T640 product, supports up to thirty-two OC-192 interfaces. Mr. Chatter noted that Axiowave's XCR128 system is interoperating with both Cisco and Juniper systems in live networks.

The Axiowave system is able to dynamically allocate bandwidth in response to conditions on the network. For example, if 25% of capacity is allocated to video services and at times all or part of this capacity is not required, the system can assign that spare capacity to other services. However, if demand for transport of video services suddenly increases, the capacity will be restored within four microseconds.

Mr. Chatter said this capability is one illustration of the innovation designed into the Axiowave product, allowing the network operator to use an IP tunnel as a robust circuit-switched connection to ensure delivery of premium services, but also enabling IP statistical multiplexing when spare capacity is available. He added that premium traffic is protected even in the presence of 400% over-subscription with best-effort traffic.

When specifying IP router systems, Mr. Chatter noted that vendors tend to cite average or best-case figures for system capabilities, never worst-case numbers as are given for ATM equipment. He pointed out that Axiowave's product offers latency and jitter specifications lower than ATM systems.

Axiowave is presently exploring possible routes to market for its product. Potential partners have been identified, including companies with no existing relationships in place and those with non-exclusive agreements with system vendors.

The competition

In the high-capacity router space, Mr. Chatter commented that there are now very few players of note - principally Cisco systems and Juniper Networks, plus Avici Systems, whose products are seen as scaled down versions of Juniper's systems. He noted that Procket Networks, recently acquired by Cisco, had not developed systems radically different to any other surviving vendor in the market. Mr. Chatter feels that these system vendors are, in all probability, aware of the issues relating to the operation of IP/MPLS networks, exclaiming,

"I anticipate Cisco and Juniper following the path that Axiowave has taken with regards to solving these issues, but Axiowave is three years ahead!"

While products similar to that from Axiowave do exist for the edge rather than the core space, Mr. Chatter remarked that scaling an edge system up to core network proportions presents numerous problems, including the system architecture,

"Building a ten or twenty gigabit capacity box is a wholly different proposition to building a 500 gigabit or one terabit product. As a general rule high capacity core systems are designed from the outset for core applications and rarely, if ever, adapted from edge solutions."

Customers

Axiowave presently counts one announced customer in PowerNet Global of Cinncinati, Ohio, and the company is currently working with a large carrier in Europe and a U.S. RBOC. Mr. Chatter added that support for additional major customers is seen as beyond Axiowave's limited resources at the present time.

PowerNet Global, which operates a profitable voice business, chose Axiowave's solution to revive its loss-making data network acquired from Swedish carrier Telia. As a result, said Mr. Chatter, PowerNet Global is now able to offer SLAs on its IP network that are superior to those of its competitors.

Axiowave's approach is to target Tier two and three carriers, where sales cycles are short, with the aim of stealing a march on the competition. In contrast it was noted that Tier one carriers generally operate very lengthy sales cycles. The intention is that this strategy will gradually raise the company's profile in the market.

Mr. Chatter believes that economic challenges in sectors of the carrier market will bring Axiowave to the forefront - as the system vendor offering a solution to the fundamental problem blighting IP networks.

Market size

According to Mr. Chatter, data from Dell'Oro estimates the edge and core router market at $9.2 billion, of which Axiowave is addressing the core router segment equating to approximately $3 billion.

Regarding the edge router segment, Mr. Chatter noted that while from a technical standpoint the product could be applied to this market, the company simply does not currently possess the resources to target both segments. Also, there are already a number of products that address this market from suppliers such as Redback Networks.

The way ahead

Axiowave Networks started out frugally four years ago, at a time when such a policy was unfashionable, taking care to build a first-class team, avoid overstaffing and generally to conserve cash. Mr. Chatter noted that funding to date, at $130 million, is low when compared to rival start-ups such as Procket Networks and Caspian Networks.

The company remained in stealth mode until June 2004, when its first customer, PowerNet Global, was announced. There was deemed to be little to be gained from going public prior to this, while the market remained in the doldrums. Present cash-burn rate is acknowledged to be approximately $2 million per month. Mr. Chatter added that Axiowave is now preparing to raise additional funding through a further insider-led financing round.

The development and success of Axiowave as a company hinges on a number of factors according to Mr. Chatter, mostly related to being small. Carriers remain reluctant to purchase from small vendors due to concerns over their longevity and ability to support customers. Mr. Chatter emphasised that the latter issue is being addressed through the formation of partnerships with larger companies, notably memoranda of understanding with a Tier two carrier in Europe and a carrier in Asia.

However, the challenge moving forward is seen to be promotion of the marketing message. Mr. Chatter feels that as discussion regarding the economics of IP networks gathers momentum, Axiowave's profile will be raised.

In summary, Mr. Chatter said the greatest challenge for Axiowave is gaining credibility in the market, given that the claims made for its product are huge and likely to be received with disbelief by potential customers. However, he asserted, the message is getting out, particularly since the signing of PowerNet Global as the first customer, and marketing and sales efforts are now being geared up to address the potentially massive market opportunity.

Optical Keyhole conducts interviews on the basis of readership interest only. They are not paid for by the participating companies, nor is there any swap for newsletter subscriptions or advertising.

This article is the copyright of Optical Keyhole. It may be freely distributed by any means in an unaltered form.