Exegy Market Data Blog

2012 Regulators in the Driving Seat

2012 will be subject to many political fluctuations given that the United States is in an election year. In addition, there are elections in Russia, France and Mexico. This will no doubt affect volatility of the markets and therefore data rates. Because of the heightened political discourse there will continue to be pressure to tighten market rules & regulations. In the US, in particular both SEC and CFTC are subject to the will of the people via Congress.  As such, they are keenly aware of political trends.

The current dialogue has notably shifted to the extent that even main stream Republicans seem to be in favor of tougher regulation for the Street, or at least that’s what they are saying in public. This could be quite significant for market data vendors as government possibly becomes a market data customer in its attempt to monitor markets and as financial firms continue to invest in order to remain compliant.

Not surprisingly, the impulse to regulate has been particularly strong since the September 2008 debacle. Then the May 6^th 2010 Flash Crash piled on yet more pressure for regulation and gave the rule makers a second wind.  I keep thinking that a high water mark has arrived but Dodd Frank will take a long time to roll out and the Volcker rule hasn’t yet been implemented.  Indeed, the comment period for the Volcker rule didn't close until 13^th January.

As a consequence of the emergency rule making in 2010-11 there are now regulations to ensure that there are guard-rails to contain volatile conditions, new standards to prevent unsupervised direct access to markets and rules (in-the-making) for large traders to report their positions.  In addition they are asking for a complete and robust audit trail capable of allowing the overseers to determine the cause of any incident considered worthy of investigation.  Indeed, the SEC’s call for a massive overhaul of the current order audit trail system has a lot of wind behind it. The initiative is called CATS- Consolidated Audit Trail.  It is not clear what the whole system will comprise, but the US financial industry would have to spend millions, possibly billions, to implement the vision.

On the other hand, there are signs that a high water mark has been reached after all. How so?  Well, over the Christmas break the Canadian regulator wrote to the US regulators outlining their worries that over zealous implementation of the Volcker rule would affect their capital markets. http://www.sec.gov/comments/s7-41-11/s74111-54.pdf

Being Canadian, they are scrupulously polite of course, but they are absolutely correct in pointing out that the world’s markets could be adversely affected if the big banks, currently providing liquidity, are prevented from carrying out their customary activity in international government bond markets.

I also suspect that as the government continues be in a budgetary squeeze, the more grandiose schemes for capturing every single trade and quote from every market imaginable stay on ice.

Economic Shocks=Volatility=Market Data

The financial markets have been notoriously volatile the last three years. This is a result of the convergence of a number of huge macro-economic, technological and political trends.  Together these factors have created the perfect conditions for a Big Data Tsunami in the world of financial markets and it is bound to continue into 2012 and 13.

The world economy was released from huge economic and intellectual burdens in 1990-91.  The Soviet Union fell, India and China embraced capitalism and various emerging markets such as Brazil began to bring their respective houses in order by adhering to financial discipline.  This led to a long, prosperous phase of growth in the world economy. 

As ever in periods of prosperity, many poor investments and decisions are made, but few are revealed as truly awful until the economy goes into recession. Since the classical downdraft in the economic cycle was dulled between 1987 and 2007 by various central bank interventions, the build-up of consequences was so much the greater when the dam finally broke in September 2008 with the bankruptcy of Lehman.  It was a veritable financial earthquake.

As you can see from the chart of the VIX, the volatility of the market moved to stratospheric levels.  The markets gradually calmed but the propensity for aftershocks remains and the world has adjusted to a new normal. But what is that normal? The new normal, at least for 2012 is high volatility:

• US economy held back by a permafrost of overvalued residential properties constraining growth and labor mobility

•  The Euro on the verge of break-up

•  China grappling with how to tame growth and inflation without upsetting their own 99%

• Middle East absorbing the promise and violence of a political springtime periodically unsettling world oil supplies

• Dealing with a nuclear Iran

• Other Black Swan events

Yes, this is the new normal, with no end in sight.  Hence one can assume that the volatility will continue and this will produce much more data as prices continue to gyrate up and down.

Hopefully Santa will bring you a large market data budget for next year, but knowing how things are going, he probably won’t.

Ultra low latency: faster than the speed of light?

Overnight yesterday we saw massive coordinated intervention by the central banks that turbocharged the stock markets. The Dow finished up 490.05 points at 12,045.68.   Oddly enough, the big moves really happened to prices before the market opened. So we didn’t see a massive spike on www.marketdatapeaks.com.  However, I did look very carefully at the marketdatapeaks data to try and understand what was happening.

I noticed something that looks like someone might have broken the speed of light barrier or perhaps someone is getting official data a little quicker than anyone else?

The peak that drew my attention occurred at 9:41 am. It isn’t a breathtaking number at 4.92 million messages per second, but it does stand out compared to other parts of the morning.

I checked with Eric Hunsader from Nanex www.nanex.net.  Eric is a very keen observer of market events. He pointed out that the Chicago PMI number is released at 9:42 am to premium customers.  “At 9:41:53, Total Ask depth in the EMini futures (ES.Z11) sunk from around 10,000 to approx 3,500 contracts, then a buy order for about 1,500 contracts came in and cleared 5 levels (1.25 points). At the same time, about $70 million of SPY was bought, and the explosion of traffic resulted in trades printing ahead of quotes by a 100ms or so.

Usually, the PMI number is released to premium subscribers at 9:42:00 -- and in the past, we've seen someone get the jump on that by 1 or 2 seconds. Today it was 7 seconds.”

Thanks Eric.   That is really insightful.

Here is www.marketdatapeaks.com from yesterday and it confirms a lot of activity just before 9:42am.

 

Low Latency, Equity Options & Risk Management

With volatility and data rates so high and a derivative trading scandal rearing its ugly head again, it is not surprising that there is renewed activity in the field of risk management and technology.  Data rates are hitting 5 million messages per second several times a week and they even topped 6 million messages per second on 12^th September and on 7th October. Most of these messages come from the derivatives markets.  Just checking www.marketdatapeaks.com today I can see that OPRA hit 2.85 million messages per second at 1:15 pm, and as far as I can see there is nothing special about today, by recent standards at least. These kind of data rates are obviously a cause for concern when your systems are trying to keep up with evaluations, proprietary & customer positions as well as theoretical pricing.  Below is a screen shot of the 7th October, the day that Fitch downgraded Spain and Italy by a couple of notches.

A few weeks ago, on 13^th September I attended the Inside Market Data Conference in Chicago and realtime risk management was prominent on the agenda.  The last panel of the day discussed the topic.  The subject fascinated me and brought back memories from my trading days- I remember the first day at Scandinavian Bank in 1988 when a compliance manager was employed in the trading room.  He sat next to me! All the other traders were suspicious and didn’t even bother to disguise their contempt for my new friend Alistair. His risk reports were delivered to him on a massive print-out in the afternoon the day after all the positions were taken! In the meantime, the markets chopped and changed and all the traders reacted accordingly. The poor risk manager didn’t stand a chance. The option trader who sat on my other flank, used to make up delta numbers for the options revaluations.  He said, confidently, that he could guess the correct numbers off the top of his head because of his experience. How far we have come since then!

 So it was not without interest at IMD Chicago that I heard senior risk managers from Bluefin Trading and FT Options discuss whether to place risk managers inside trading groups, or outside. Another point they discussed was the desire for risk managers to have alternative systems to carry out evaluations, systems that are not shared by the traders who might well be tweaking Greeks for various effects in the complex book.  Things haven’t changed so much then!

 The conversation also gave me a new appreciation for the VoleraFEED available from Hanweck Associates (http://www.hanweckassoc.com/) and currently in development at Exegy. Options traders and firms increasingly demand an outside system to furnish theoretical prices and Greeks so that they can double-check their own numbers and data.  The world moves so fast and the positions are so complex, this kind of feed is a must-have. In addition, OPRA has suggested that they might well come out with a feed along these lines. OPRA is calling it the OPRA Implied Volatility & Greeks Feed.  If they can tune it to the needs of the market and deliver it economically, it could become a market standard. However, it is not easy making such a feed since many practitioners have very specific preferences as to algorithms and they are often quite particular about inputs for dividends, historic volatilities and interest rates. Nevertheless, if you are interested in taking such a feed from OPRA, you should not hesitate to check out their survey and give them feedback https://www.surveymonkey.com/s/OPRAIVandGreeksfeed

 And if you need to simply be able to handle all the data on one server from the OPRA feed, please feel free to contact Exegy to find out about the Exegy Ticker Plant as a dedicated appliance http://www.exegy.com/.  It is also possible to get OPRA data via the Exegy Data Tone.  A single connection at $5000 per month might well deliver all you need.

Market Data Rates: August 2011 v Flash Crash v August 2006

When I started trading for an investment bank in the 1980s, I was told that nothing happens in the summer.  Crashes occur in October, and big market moves in January. Life was so predictable. These days everything happens in the summer- market crashes, government debt downgrades, big corporate mergers, European crises, Middle Eastern revolutions, even ice hockey finals are mid June.

Not surprisingly market data rates are growing at a strong rate of knots to reflect all this activity. On 25thAugust www.marketdatapeaks.com hit 5.4 million messages per second for the first time and it has been over 5 million mps several times this month. The OPRA feed has been exceptionally busy.  SIAC reported to the Financial Information Forum, that OPRA hit 3,460,860 mps on August 2^nd. This compares to a May 6^th Flash Crash high of 1,331,044 mps. Five years ago, back in August 2006, OPRA pushed out 157,442 mps on its busiest day. 

Not surprisingly this month, the options exchanges that make up OPRA revised their forecasts for future growth and OPRA alerted direct feed recipients on August 15^th that they need to provide for a potential of 9,127,000 mps in July 2012 as opposed to their previous forecast ceiling of 6,537,000 mps at that time.  This surprised a few people but if you look back at the long term trend, it is similar to growth rates we have seen before. 

CQS has also been scaling new heights at a strong clip. SIAC reported that the top of book feed peaked at 476,619 mps on August 5. This compared to 199,198 mps on May 6^th 2010. Five years previously, in August 2006, the CQS high point was 5,489 messages per second.  But I think the main news in August 2006 was the demotion of Pluto from full planet to a mere dwarf. Now that was a fine summer.

When Pluto was a planet, did we have a capacity problem?

Low Latency – Our Only Goal

We received several calls from friends about a comment we made in our recent blog, Google and Exegy.

In that piece, Rod Arbaugh, our COO, wrote “At Exegy, we build our ticker plant with one goal – to be the lowest latency ticker plant in the market across the use cases most meaningful to our clients.”
As many of you know, we are in the market presenting solutions to you that serve a wide range of needs in the enterprise – many of these uses are not particularly latency sensitive. Why then is our singular focus to build the lowest latency ticker plant?

We do this for the same reason the major auto manufacturers build F-1 race cars.  In designing to the most demanding performance standards, you learn a lot about how to produce a better product for a much broader set of users.
In building every product around our best performing ticker plant, we confirm a corporate culture that believes LATENCY MATTERS.  Sometimes, our customers value lower latency in solutions that are not wholly latency sensitive.  But, more often than not, our ability to process large amounts of data very efficiently at very low latencies supports other product features that our clients prize very highly.

A good example of this is the distribution of market data throughout the enterprise.  Exegy entered this business in the spring of 2011 and promised our clients a better way to distribute market data to hundreds of users.  The many advantages of our low latency ticker plant allows us to build enterprise solutions that have very small footprints, very efficient integrations with different distribution technologies (Multicast, Unicast, RDMA) and capacity headroom that legacy solutions cannot offer. 

Building a market leading enterprise solution on a less than market leading ticker plant is a very hard proposition.  At Exegy, we build our enterprise solution on the best performing ticker plant in the business and use the skills we learned in the low latency, hardware accelerated world to further extend the attractiveness of the enterprise solution.

I agree with Rod that our only goal is to build the lowest latency ticker plant.  This enables us to build an array of great products in the market data vertical with features and benefits that our clients value and that our competitors cannot match. Why would anyone try to build a world-class market data company on a less that world class ticker plant?  -seems like Business 101 to us.

Hardware Acceleration- Five Years of Being First

Exegy is just celebrating the fifth anniversary of signing its first hardware accelerated ticker plant customer in 2006. It takes the perspective of several years to fully appreciate how far ahead Exegy was- especially now when the mention of microseconds, low latency and FPGAs is commonplace.  Here’s my take on how Exegy came to be so far ahead.

At a base level, it seemed obvious to the technologically gifted founders of Exegy that parallel processing of massive amounts of data would be useful in all sorts of applications in the information age. They had this realization in the late 1990s and formed the company in 2003.

In retrospect, it was exceptional good fortune that one of their doctoral students at Washington University St Louis worked in the market data industry. This personal connection led to the realization that there was a strong business case in the world of market data.

Late in 2005, Exegy began hiring engineers with years of experience in market data. As soon as the summer of 2006, Exegy was able to demonstrate the world’s first FPGA based ticker plant.

I joined later in 2006, already aware that they were doing something special at Exegy. Having watched market data rates climb after the decimalization of US equity markets in 2000, I knew that there was a dragon that needed slaying. It was clear that market data rates would skyrocket and that conventional software based systems would not be able to handle the loads economically. It was also significant that Exegy already possessed significant patents regarding hardware acceleration in general.  Indeed, Exegy now has a portfolio of fifteen issued patents and twenty-three in process. 

But in 2006 and 2007 it wasn’t obvious to direct feed recipients why they should consider an unconventional computer system sold by a little technology firm from Missouri. In fact, I distinctly recall describing many times over exactly what an FPGA is and how parallel programming works while pointing to statistics from the Financial Information Forum illustrating the inexorable rise of market data to a lot of people on Wall Street and in London. But when markets were only hitting around 150,000 messages per second on a busy day, the fact that the Exegy system could handle 1 million messages per second seemed immaterial and quite academic.

In fact, as everyone in the industry now knows, the direct feed boom would take off very fast in the US and globally.  On August 4th 2011, the markets hit well over five million messages per second during the latest financial crisis.  The volumes are immense and it is now reassuring to market data professionals to see the peaks publicly recorded in realtime by one single Exegy Ticker Plant as on www.marketdatapeaks.com.  I wonder where we will be in another five years!

Google and Exegy

The recent announcement of Google’s bid for Motorola Mobility Holdings has brought forth a lot of discussion in the press about the strategic reasons behind this merger.

At Exegy, our attention is drawn to two themes that seem to be prominent in Google’s strategy.  First, Google will tightly integrate the Android software application with the hardware platform of Motorola.  This is the tactic that has been deployed so successfully by the Apple iPhone product.  This “win” for the tight integration of a software application with a proprietary hardware platform, the appliance approach, reflects our own implementation of our solution to processing market data.

In the financial vertical, legacy software vendors supply solutions that run on commodity servers.  These systems are generally scaled horizontally to meet the demand to process the ever-rising amount of market data.  This adds management complexity and latency to the solution.  We routinely visit prospects that have their incumbent ticker plant solution running on hundreds of servers.  If market data rates were to double from here, they tell us they would have to double the size of their deployments.

The legacy vendor response is to deliver better software running on faster commodity servers and this does provide an incremental improvement.  But the limits of the commodity platform do not allow a paradigm shift in performance and is very expensive to provision and maintain.

At Exegy, we build our ticker plant with one goal – to be the lowest latency ticker plant in the market across the use cases most meaningful to our clients.  The Exegy “secret sauce” is simply to offer the best integration of the best ticker plant application on a hardware platform custom built to do one thing – process market data.  And we support our product offering with an approach to managed service that is the best in the business.

The second theme that caught our eye was the speculation that Google values the Motorola intellectual property portfolio.  We appreciate that reasoning, as well.  At Exegy, we have been actively protecting and expanding our intellectual property portfolio for more than ten years.  As of August 2011, we hold 15 patents with 23 applications pending in the United States and we routinely make similar filings in Europe and Asia.  Many of these patents address the use of FPGAs and other hardware accelerated devices in the financial markets.  We like to tell clients that the Exegy solution is built upon Exegy proprietary technology.

We won’t speculate on who will win the phone wars but we do appreciate the fact that the two leading contenders have both adopted the integrated appliance approach.  We cheer them both on, as long as they don’t decide to colocate their mobile devices at exchange trading venues to process market data.

Market Data: Beware False Dichotomies

We agree whole-heartedly with the conclusion of the Victor Anderson, Editor in Chief of Waters Technology, article on the importance of service to customers of technology vendors in the financial services industry. While service is a key driver for customers, we also see a strong dependence on technology to deliver great service. Separating these issues creates a false dichotomy.

At Exegy, we lease ticker plants used by the financial services industry to provide critical data to both trading applications and to their enterprise distribution systems.  Many of these applications require extremely low latency, less than 15 microseconds, and very high, do-not-miss-a-tick, reliability.  With more than a hundred ticker plants in the field at colocation facilities around the world, we are committed to providing great customer service round the trading clock.

Customer service in the market data space is hard.  The volume of market data is growing rapidly, client demand for ever lower latency is never ending and new groups within our clients want low latency market data for their non-trading applications delivered in lower cost, more reliable ways.

We could not meet the demands of our clients in this challenging environment without great underlying technology.  We see competitors struggling to provide the required level of service across older technology platforms.  This is a frustrating exercise.  No serious business starts out to deliver bad service but many businesses find that delivering great service based on technology that struggles to keep up with current demands is a losing proposition.  These businesses inevitably try to “refresh” their older technology platforms to answer customer demands for better service but find that an elusive goal.  (We believe this has everything to do with not adopting a corporate culture of continuous innovation – but that is a topic for another blog).  This boom-bust cycle of technology and service is not a reliable solution for clients in the financial services industry.

Our technology platform drives our approach to customer service.  We invest a lot of time and resources to continuously improve our product in order to be able to maintain great customer service.  Not just at “refresh” points in some technology cycle, but continuously, every day.

This implies that managers in the financial services industry need to understand what drives great customer service from a vendor.  Choosing a vendor with an older, struggling platform without a corporate culture of investing in the technology every day is a recipe for a bad service outcome – regardless of the quality of their service organization.

Technology with enough headroom to credibly stay ahead of market data demands while meeting client’s performance and cost metrics is the critical foundation for building a great customer service organization.

Low Latency 2.0

Paul Rowady, TABB Group senior analyst, has a thought-provoking video interview on Tabb Forum entitled Latency 2.0 that we recommend to you.

At Exegy, we strongly agree with some of what Paul puts forth, disagree with other positions; other comments have generated a strong debate in the halls.  Here is an example of each. 

Paul states early in the interview that the world has shifted to the point that reducing latency is no longer just a competitive advantage but a competitive necessity.  We see plenty of validation for this position in our interaction with the market.  Customers are moving execution services to colocation venues to reduce execution latency and provide their customers with state of the art trading platforms.  The nature of algorithmic trading execution demands that the algorithms respond to changes in fragmented markets as indicated by an ever growing amount of market data.  The North American equities and options markets routinely generate 3,000,000 changes to their order books a second.  Robust algorithmic trading platforms must efficiently make millions of trading decisions a second in close proximity to execution venues.  Why would clients demand anything less of their agents?  Why would a broker think that a less robust algorithmic platform will remain competitive?  Large buy side firms are now demanding order routing histories for each execution.  They are using this information to score algorithmic execution platforms.  This is just one example of latency reduction becoming a strategic necessity.

On the other hand, we would challenge Paul’s contention that only a few firms can dominate at the “tip of the (latency) spear”. Certainly, the opportunity set of profitable trades is not declining in this business.   At Exegy, we tend to think of this opportunity set as those trades that generate small amounts of alpha with very short holding periods.  These periods are probably measured in seconds by today’s risk systems and require access to market data and execution capability measured in microseconds.  In the past, the cost of building the technology infrastructure to compete at these levels was expensive and dominated by a few firms.  Today, colocation hosting firms offer access to key parts of this technology for as little as $5,000 per month and an ever growing list of prime brokers bundle technologies to provide “best in class” execution systems to their clients as part of their overall offering.  Technology is rapidly reducing the cost to compete at the “tip of the spear”.

The largest trading firms do enjoy a competitive advantage from interacting with customer order flow in proprietary dark pools but that is a topic being played out in the US and European regulatory environments.  It is easy to imagine that these advantages will be narrowed in the future, expanding opportunities for a plethora of other participants in the low latency trading market.

Finally, Paul’s comments on trading signal generation from news and other asset classes and the need for technologies to adapt to the client’s search for new sources of alpha should spark discussions inside any manufacturer of low latency trading technologies.  It happens at Exegy every day, most recently in considering the interaction of foreign exchange signals with equities trading.  More on that in a later blog.