Richard Korthammer, Lava's CEO

With its strange mix of new luxury condo buildings, nondescript offices and auto repair shops, the Morton Street area in the West Village looks like just one more New York neighborhood that is pushing the geographical envelope of gentrification. It does not look like home to a major power in equities trading.

This is where Lava, an independent subsidiary of Citigroup, which accounts for slightly over seven percent of the equities volume in the U.S. markets, is headquartered, a spot chosen some months after a successful evacuation of all 35 employees from the 83rd floor of the World Trade Center’s north tower.

But location is just one of the many unconventional aspects of Lava Trading. A market player that could be described as a direct market access provider with brains and speed, Lava was the first firm to provide access to all the U.S. market centers when it launched in late 2000.

“We help clients get best execution by accessing all the liquidity pools that are competing,” explained Richard Korhammer, CEO and co-founder of Lava.

“When you come to DMA, or aggregation, or whatever you call it these days, I think Lava timed it perfectly,” said Sang Lee, managing partner at Aite Group, a financial technology research firm in Boston. “Lava was one of the first firms to enter the institutional market, and they have done very well.”

And Lava’s role in the market is not all that makes it unique.

Lava Trading runs as a Specialized Application Service Provider (SASP), or hosted solution. And it runs entirely on Microsoft including SQL Server, and a Lava high-speed proprietary middleware. That choice was both bold and unusual back in 2000.

An electronic engineering graduate from Princeton, like co-founder and CIO Kamran Rafieyan, Korhammer brought a combination of technological and markets knowledge to trading. Lava chose to build on Microsoft initially, and has stuck with the platform. Most of the serious developers and engineers coming out of university would lean to UNIX solutions, Korhammer admitted. But by the end of the 90s, NT was becoming reliable and Intel hardware was catching up to UNIX at a much lower cost.

"With Lava, the buy side can choose which broker to trade with"

“One of the first things we had to do was decide between UNIX or Wintel. We considered that with the use of redundant PCs we could get the same or better reliability as a mainframe at a fraction of the cost and then become more competitive,” Korhammer said.

As an example of the Wintel cost advantage, Korhammer pointed out that Nasdaq spent more than $100 million on its SuperMontage technology.

“We didn’t come close to spending that kind of money. For SuperMontage, that is a very expensive cost basis,” he said. When the company started, its data center was handling 2,000 transactions per second; now it is processing 50,000. If one PC fails, another picks up the work, providing the same power and reliability as a mainframe at a fraction of the price.

“Lava rewrote the business model for executing trading,” said Kenny McBride, Microsoft’s industry manager for capital markets. “And they did it with commoditized technology. Lava showed that with business domain expertise coupled with engineering expertise, you can redefine the business model. They are an excellent example of software as a service.”

Getting the right architecture for market data is key to success, said Korhammer.

“Imagine driving a car and seeing oncoming traffic with a two-second delay. The same is true with trading; you should not go after trades that don’t exist any more. Speed is the starting point of being in the game,” he said.

In selling Lava Trading, Korhammer faced some questions in the early days about the choice of Microsoft, and the viability of a hosted solution for high speed trading.

“When we landed our first bulge bracket firm, they were very concerned about whether we could keep up with their flow. They agreed to trade through our data center in a service bureau fashion, but we had to commit to building our solution in their data center after a period of time. They felt trades wouldn’t be fast enough in a shared environment. A year or two later they said they were pleasantly surprised – our systems through our data center were faster than their local systems. On a good day, we will assist in the execution of 400 million shares, though we will actually process over a billion shares.”

Many more firms followed suit.

“Over the course of one and a half years we signed up 16 of the top investment banks,” recounted Korhammer. “We are widely accepted as the best way to be competitive in a fragmented market.” The firm’s clientele now includes most of the top investment banks, approximately 70 broker dealers, and more than 100 buy-side firms including hedge funds, he added.

"For Citi, we have provided fantastic trading technologies."

Lava offers access to multiple brokers through a single trading workstation.

That structure allows buy-side firms to direct trades to the broker of their choice for best execution, or to direct order flow in return for research.

“You can do business with 20 different brokers without having 20 different front ends,” Korhammer explained. “Some DMA platforms have commission allocation programs – but that’s not multi-broker technology. It is simply a way to distribute payments.” Typically those systems work through one brokerage, which then passes on the commission to another broker as directed by the buy-side client.

“If you’re a broker who gets a commission check from a brokerage firm, don’t you think you’d prefer a direct relationship with the buy-side firm? And besides, the intervening broker gets some of the spread. With Lava, the buy side can choose which broker to trade with. That broker gets the volume credit and he has a direct commercial relationship with the buy-side firm,” Korhammer said.

When Lava started, the problem was that a sell-side firm might have trading workstations from Nasdaq, Instinet and maybe one other ECN, Korhammer said. Integration consisted of traders whirling around on their swivel chairs to enter their orders on three separate dedicated terminals, a solution that might be described as sub-optimal.

With Lava, a trader can hit any pool of liquidity from a single screen and full depth of book.

A recent survey by Elkins/McSherry, which was published in Institutional Investor, ranked Lava tops in execution costs for NYSE trading and second for Nasdaq securities in a comprehensive evaluation that looked at commissions, fees, and market impact.

“A lot of our success is based on speed and intelligent order types of best execution.” And intelligent order may combine orders at the bid and the offer to get a blended price and fill the entire order. The system will also understand reserves, which are often hidden, and provide ways to draw out the reserves without driving away the price.

“We have those technologies to help you manage adding liquidity and removing it simultaneously, but help you choose the venue for your bid. If you are one of many at an ECN at the price, you might be in a line to have your bid hit. We will help you pick the venue to place your order and determine how much liquidity to add or remove.” Firms can use Lava as their front end or connect their own trading algorithm to it.

When Citigroup bought Lava two years ago, some users were concerned that Lava would lose its independence. Korhammer said the acquisition has been good for both firms, and for Lava’s clients.

“Over time we demonstrated we have complete Chinese walls,” he said. “We showed we are operating and building solutions for all our clients no differently than before. For Citi, we have provided fantastic trading technologies, we have a good technology vision and a profitable line of business. Lava has benefited from their distribution force.” Citi has relationships with almost all the buy-side firms in the country, and Lava’s buy-side growth has reflected that, Korhammer said. Lava has hired more than 100 people in the last year, and its staff is over 300 to date. Citi has the ability to help fund initiatives faster than Lava could do as a private company. 

Citi has also helped the company expand internationally since it has capital and a trading presence in more than 70 countries.

“I can go to any country and learn about their trading and their local laws,” Korhammer said.

Japanese investment banks and hedge funds can use Lava technology to trade in Japan now, and the company is looking to expand its presence in Asia.

“We launched DMA in Japan last year and this year we are working towards lighting up three or four additional Asian countries,” said Korhammer. Lava would support domestic trading – Asian banks hitting their domestic exchanges – although much of the volume may be coming into the domestic market from foreign investors.

Meanwhile, Lava has continued to innovate in the U.S. market as well.

“Clients were looking for innovative ways to process trades and manage orders,” Korhammer said.

Lava responded with ColorPalette, a sell-side order management system that it has sold to 25 clients. Sell-side order management is complex because it has to handle everything from pink sheet equities to the NYSE, he said. Lava competes with Brass and royalblue in this arena.

Harrell Smith at Celent, manager of Celent's securities and investments practice, praised the way Lava has moved beyond connectivity and aggregation to develop front-end trading capabilities.

The Network Operations Center, fondly known as NOC, is where Lava staff monitor the network to ensure high performance.

“Their success has been borne out of the mid-tier sell-side firms which have moved entirely onto Lava and have moved away from standalone order management systems like SunGard’s Brass. With the ability to provide a front- to back-end solution, they have succeeded beyond their original business model.”

In other functional areas, ColorBook is a direct market access system that includes aggregated depth of book from almost all liquidity pools, including ECNs and exchanges, and intelligent order types for executing across market centers. ColorMaker builds market-making capabilities on top of ColorBook by integrating price discovery, direct access and quote management into a single application.

To uncover dark liquidity, prices that are not posted for traders to see, the firm has DarkBook to help firms achieve faster fills and fewer partial fills while still minimizing or avoiding market impact. LavaPI offers tools for price improvement, while ColorPing and Continuous Ping provide yet additional tools to uncover hidden liquidity. They let the trader send IOC orders to one or more execution routes and ping for hidden liquidity before ColorBook executes a special order that sweeps across multiple sources of liquidity. ColorData Feed provides fast real-time aggregated market data across multiple liquidity pools, while ColorData Archive stores the market data for the firm’s uses, and ColorData Server powers Lava’s quote montage and offers it to users on demand.

For its competitors, Lava Trading presents a fast moving target. Last summer Lava announced that traders would be able to access Pipeline, the alternative trading system for large blocks of stocks.

“I think our users are very interested in getting price improvement or size improvement,” said Korhammer. “Pipeline, with its ability to anonymously trade large blocks, fits the bill for those Lava users.”

Firms well known for their trading expertise have partnered with Lava for execution of their trades. Earlier this year Susquehanna Financial Group became part of Lava’s Sponsored Broker Program, which allows it to sublicense Lava technology to its clients so they can route orders to all major liquidity sources.

“This allows us to provide our DMA and algorithmic services via an easy-to-use and well established application that has wide appeal among many types of trading customers with different trading styles,” said David Margulies, head of program and algorithmic sales at Susquehanna. “Customers can benefit from Susquehanna's cutting-edge technology and innovative trading strategies without interrupting their current work flow or adding additional cumbersome applications to already crowded desktops.”

Korhammer said Susquehanna’s highly advanced algorithmic trading strategies and transaction cost analysis complemented Lava’s high performance trading.

More than a year ago Lava began applying its technology to foreign exchange. Russell Investment Group, which began using LavaFX last autumn, has become a market maker for its clients. Ian Battye, the firm’s head of currency management, said Russell clients tend to be price takers rather than price makers.

LavaFX represents a different FX trading model than that used by most buy-side managers, but Russell anticipates this model to become more prevalent as electronic trading evolves. LavaFX will let Russell offer its clients additional liquidity, trading anonymity, and information they can use to evaluate best execution. The LavaFX model takes streaming FX price data from banks and also aggregates the prices on multiple FX portals. Users can request quotes from their banks and the system will check other sources of FX and highlight any better prices that are available. Like its equities platform, Lava’s FX allows traders to offer their trades in tranches to attract counterparties without giving away the size of their order and moving the price.

Within the next few months Lava also plans to launch an options order routing system.

---

32-bit Windows Offered Great Transaction Speed from the Beginning

In 1999 when Lava Trading was getting started, Microsoft was not the obvious choice for technology on Wall Street. Sun was probably the most widely used technology in front and middle offices as traders learned that engineering workstations provided the power and flexibility they needed. The echoes of that belief still reverberate around the canyons of Manhattan.

“In processing data, the 32-bit Windows operating system was well designed and a great match for Lava’s business.” –Kamran Rafieyan, Lava’s CIO

But even then, some recognized the processing power of the Windows 32-bit platform, said Kamran Rafieyan, the company’s chief information officer.

Maybe it was because his technology team was made up of experienced UNIX programmers who could evaluate Windows without fear or prejudice. Or maybe the economic reasons swayed the company in the Microsoft direction. After all, this was a startup without a lot of money, and Microsoft offered a cost-effective platform and a great development environment, from tools to Web sites with the sort of answers developers need.

“I and a number of other folks were using Microsoft technology throughout the late 90s,” said Rafieyan. By the end of the decade they understood its capabilities. And in processing data, the 32-bit Windows operating system was well designed and a great match for Lava’s business.

“What we build are high speed messaging applications,” Rafieyan explained. “They are not CPU-intensive, so for us efficient I/O (input/output) is everything. When WinNT and Win32 first came out, we thought they did a good job of factoring in multi-threading and asynchronous networking from the start.”

At that time, he admitted, a lot of people still thought of Windows as a desktop-only environment and mostly knew the limits of the 16-bit Windows running heavy applications like Microsoft Office. They overlooked what Microsoft was doing with 32-bit Windows on servers.

“And the 32-bit Windows Server was pretty good. A huge amount of our architecture is designed around I/O completion ports, which they built into Windows 32. It is a very efficient mechanism for doing scalable asynchronous networking,” Rafieyan said.

Much of Lava’s work boils down to messaging, he added. “Win32 provided a very efficient way to have an application process large amounts of I/O tasks. The I/O completion ports lets you set up multiple threads, keep the optimal number busy, and minimize context switching. It scales very nicely because it was very much geared toward multi-processor machines.”

To support trading operations, Lava has the standard features of a mission critical installation – redundancy in computers, communications, and facilities.

“There are only a couple of approaches to proper software fault tolerance and reliability,” explained Rafieyan, “redundancy, shared-state, and shadowing.”

Lava uses all of these mechanisms. With shadowing, the same data is processed by more than one application, so that if one instance goes down, the other can immediately take over. It is complex to build, but very fast.

In trading, speed is important, and Lava ranks at the top in its speed of trading.

Achieving speed depends on a lot of basic computer science fundamentals, he said. Lava leverages the I/O completion ports in 32-bit Windows and takes full advantage of its multi-threading.

The company uses C++ for most of its programming.

“It’s still the fastest for true systems work,” Rafieyan added. The company employs software optimization techniques and has built its own middleware for high performance.

Lava uses SQL Server 2000 in a high availability mode and is testing SQL 2005, he added. Rafieyan professes to be a conservative when it comes to new software; the company only recently moved to Windows Server 2003. And Lava achieved a big gain in performance by installing a storage area network (SAN).

“We found SQL Server was scaling fine; the disk was the bottleneck, so we installed an EMC solution. If you want speed in storage you need to move to SAN,” he said.

Speed, intelligence and reliability are the mantra at Lava.

“With regards to intelligence, that comes from the combined business knowledge of everyone here,” Rafieyan added. “It requires an understanding of the markets, choosing the solution the market needs, and obviously then having the technology to implement it. We are known for our intelligent order types, our ways of approaching the market structure.”

That market intelligence is now being applying to Reg NMS, the SEC rule requiring best price for trading that takes effect by mid-year, unless it is delayed.

“NMS will have tremendous impact,” said Rafieyan. “It is one of the top priorities on every financial institution’s plate. We have been out to the majority of our customers doing presentations on it and what we are doing about it. This is probably the biggest change since the order handling rules in 1997. There’s a certain amount you have to do just to be compliant. The market intelligence comes in how to interpret it to provide competitive advantages. We also think it is going to open up volumes because at the heart of the rule is a large bias toward automated electronic trading. The regionals all have plans; all the major market systems are updating their systems.”

Analysts say Lava is wise to take notice. When Reg NMS was first announced, it had the potential to hurt DMA providers, said Aite’s Sang Lee, but as the rule has evolved, it looks as if it will benefit them.

“As Reg NMS is in the slow motion of being implemented, we are seeing a lot more market fragmentation, and for a DMA player, that is positive. The last thing a small firm wants to do is connect with all these venues that are popping up.”

Lava is watching closely, and preparing its systems for the new market world. Nothing new about that, though.

“Financial services is an industry that really uses technology as a competitive advantage, rather than viewing it as just a cost,” he said. “It is an exciting place for a technology person to be, although we do sometimes think it would be nice to have a quiet year.”

http://www.lavatrading.com/