Web-acceleration company Zeus Technology is in growth mode — the Cambridge-based company has just opened a US office, and is expanding its staff and distribution channel. But unlike most of its peers, Zeus has seen it all before.
Founded in 1996, the company was one of the darlings of the dot-com boom, making it practically ancient in IT-industry terms. Yet it has come back from the lows of the crash to achieve a significant place among the fast-growing number of companies specialising in high-performance web infrastructure.
Now, the dramatic increase in number and complexity of web applications, the explosion in broadband connectivity, branch office server centralisation projects and the growth of complicated technologies such as blades and virtualisation in the data centre have all created a market seemingly tailor-made for Zeus.
Web-application acceleration sits in a grey area between the expertise of networking companies such as Cisco and giants such as Microsoft, Oracle and IBM, which focus on enterprise applications. It's a diverse area that everyone seems to define a different way, but whatever the measure, it is growing rapidly. Gartner estimated the market for what it calls application delivery controllers (ADCs) to be worth more than $1.5bn at the end of 2006, a 23 percent increase over 2005.
"Our definition of it is much broader than Gartner's," says Damian Reeves, Zeus co-founder and chief technology officer. "It's an absolutely huge market." Gartner splits the market into ADCs — generally appliances that sit in front of web applications and handle application speed and reliability — and WAN Optimisation Controllers (WOCs), which control network performance.
Zeus's main product, Zeus Extensible Traffic Manager (ZXTM), is different from the ADC appliances sold by competitors such as F5 Networks, Citrix NetScaler and Riverbed, in that it can take the form of either an appliance or software that runs on a traditional server, blade or inside a virtual machine. "We're walking toward the network, from the application side. That's something our classic competitors, who come from the hardware networking side, can't do."
Besides conventional acceleration appliance vendors, big software players are getting into the acceleration game — middleware from the likes of Oracle and BEA and even high-performance features being built into Apache. IBM's WebSphere business unit is a significant player in the market via its DataPower XML and SOA acceleration appliances.
And the multitude of start-ups in the space even includes a company co-founded by Zeus co-founder Adam Twiss in 2002 — CacheLogic, which focuses on high-performance media delivery using peer-to-peer technology.
Nevertheless, Zeus has managed to continue growing its own niche, with its infrastructure powering more than one million sites for more than 800 customers, including the likes of BT, China Telecom, Nasa and Play.com.
Ovum analyst Gary Barnett sees Zeus's classic market as the top 1 percent of companies with very exacting performance and availability requirements, comparing them to a supplier of parts to a Grand Prix team. "Grand Prix teams will buy injector jets from a two-man band if they think they can get a 100th-of-a-second advantage," Barnett says. "Zeus's ongoing challenge is to stay at that cutting edge."
It was once all so much simpler. Zeus's genesis was at Cambridge University, where co-founders Twiss and Reeves were students at Churchill College. The idea was inspired by the networking situation that existed at Cambridge at the time, where students had access to extreme bandwidth, but were stuck with servers that would fall apart under the slightest pressure.
"It was clear that while that was an anomalous situation in 1993, it would soon become the norm," Reeves says. "The growth of internet traffic was outpacing Moore's Law, and soon everybody would have that problem. The performance and scalability problems were so acute that we could see they needed an entirely different kind of solution."
The co-founders' answer was technology that could handle a large number of simultaneous internet transactions, something not envisioned by the original creators of the internet's underlying protocols.
"The internet is a relatively simple system," says Reeves. "When there's a small amount of traffic coming in, all you need is a simple program to deal with it when it comes in, and send responses back to the end users. The challenge is when the volume and importance of that traffic increases."
When the dot-com boom came along, there was suddenly a huge influx of...
...richly funded start-ups whose business models depended on being able to reliably carry out large numbers of internet transactions. "That fit very nicely with the benefits of our technology," Reeves says.
Early customers for Zeus Web Server included companies that either generated a large amount of traffic themselves, such as eBay, or aggregated large amounts of traffic, such as Worldcom's web-hosting business.
In 2000, Red Herring named Zeus as one of the top 50 most important privately held companies in the world (it remains privately held). In 2001 and 2002, nearly 3 percent of the world's websites ran on Zeus Web Server, according to Netcraft's surveys; for a time in 2003 the server was the third-most popular server after Apache and Microsoft IIS.
As plenty of other companies found out, it was too good to last. "After the dot-com crash, being in the business of providing technology to people running web presences was a tricky space," Reeves says.
The financial challenge came when Microsoft started giving away IIS with Windows NT, which added to the pressure from the freely available open-source server Apache. "Being in the web server space commercially was getting to be a more difficult prospect, when everybody around you was bundling their technologies or giving them away for free."
Zeus still sells its server, which ranks as one of the highest-performing on Unix and Linux platforms, but in 2004 the company shifted its focus in a new direction with the release of Zeus Extensible Traffic Manager, which it calls the only pure software traffic-management system on the market.
"We were moving where we placed our technology," Reeves says. "Instead of just being in the application server or web server market, we effectively moved forward a layer. We're placed in front of those Apache boxes or IIS boxes, offloading a lot of the hard work that's not their strong point. It meant we could add our value to the traffic flow without having to try and displace Apache or Microsoft."
Traffic Manager continues to add sophisticated new load-balancing features. Version 2, in 2005, saw the addition of an XML engine, designed to deal with XML-based traffic in service-oriented architecture (SOA) and web services; the feature was enough to earn the company the plaudit of "visionary" from Gartner analyst Mark Fabbi.
"We've embedded an XML engine into the routing fabric of the device itself, so that it handles XML messages as they flow through the network. We're the only vendor in this space able to do this," Reeves says.
XML handling means, for instance, that Traffic Manager can parse XML messages to see which ones are purchase orders for more than £30,000, and send those to a particular set of dedicated boxes, Reeves says. "Most of the Layer 7 technologies that the switching companies are using are completely blind to XML," he says. Version 4.1 of the product, released last year, can handle more than 9,400 SSL transactions per second, twice the volume of the previous-generation ZXTM 7000.
The basic drivers for Zeus' business are still the same, Reeves says — high performance and high availability. What has changed is the underlying technology and the user context, which has evolved dramatically in the past couple of years.
An important trend in enterprises at the moment is scrapping rich client applications in branch offices and replacing them with web-based interfaces to centralised servers, especially in industries such as finance and retail. That alone creates a huge demand for web applications to become more responsive and available, says RedMonk analyst James Governor.
"As we see more and more deployment of web applications rather than rich clients, acceleration technologies will become more important," he says. "Clearly there is room to grow in this space."
Then, outside the enterprise itself, there's the phenomenon some pundits have taken to calling "web 2.0", as though to invite a second crash. "That term means something different to everyone, but as websites become much more dynamic, there's an order-of-magnitude increase in the complexity of the things you need to do to make a site high-performance," says Ovum's Barnett. "There's been a huge growth in highly dynamic, high-volume websites."
Such changes are spurred on by the growth of broadband, where people are used to interactivity and video content, and routinely download huge files.
Finally the technology inside the data centre itself has changed radically. "From a small number of large, expensive boxes from Sun or IBM, it's gone to aggregating the power of many commodity machines, through to blades, where you're packing more power into the same space. And then there's virtualisation, where you're slicing up physical computers into 10 or 50 virtual machines, each with its own OS and stacks," Reeves says. "It's an explosion in the number of devices in the data centre."
The huge increase in data centre complexity has inspired the highest-tier players, such as Google, to come up with their own proprietary techniques for massive scalability, Barnett notes. "The rest of that top 1 percent is the classic market for the likes of Zeus," he says.
Zeus is aiming to stay ahead of the game through constant technical development, international expansion, widening its channel through new distribution agreements, and even licensing its technology to be built into third-party software and appliance products. And then there's the company's long pedigree in the space, which few can match.
RedMonk's Governor says there's no reason why, given its solid niche, Zeus shouldn't prosper. "There's always a danger that big companies might move into a market, but that doesn't necessarily mean they'll win," he says. "IBM acquired DataPower, but does that mean everybody's buying DataPower now? Not necessarily."