Larry: Welcome to this E-Forex Equinix presentation. Today, the subject is FX in the Cloud. We’re pleased to welcome our three speakers: Robin Manicom Director Financial Services Equinix, Javier Paz Senior Analyst at Aite Group, and Jubin Pejman Managing Director at FCM360.
欢迎来到本次E-Forex Equinix讲座。今天的主题是云端外汇。我们很高兴邀请到三位嘉宾：Equinix金融服务总监Robin Manicom，Aite集团高级分析Javier Paz，FCM360常务董事Jubin Pejman。
Today’s webinar seeks to help institutional sell-side and buy-side FX trading firms better understand cloud computing technologies that increasingly optimize high volume, latency sensitive trading operations. It looks at how cloud enterprise technology, infrastructure location, low latency, and multiple connective options can help firms reduce costs, capture new business opportunities, improve trading performance, and meet the challenges of regulatory reform in FX.
To give a brief bio, Robin Manicom has over 20 years’ experience working in the financial services sector, providing high performance solutions within the quantitative and algorithmic electronically traded markets. Robin joined Equinix in 2009 to manage the EMEA Financial Services business strategy for platform Equinix. Before Equinix, Robin spent three years at Markit as Head of Business Strategy, Sales and Marketing where he developed the solutions business for real-time feeds. Prior to this, he held various management and sales positions at IBM’s STP business and also at Sybase. Robin holds a Bachelor of Engineering from Bath University and received his MBA from Surrey University.
Javier Paz is a Senior Analyst at Aite Group within the Wealth Management practice. He examines retail brokerage trends in business, regulations, and technology across asset classes with particular expertise in FX markets. Prior to his current role, Javier was an Industry Consultant and also part of management at an FX brokerage firm, as well as a sell-side analyst for two major banks.
Jubin Pejman, Managing Director of FCM360. Prior to founding FCM360 to facilitate turnkey low latency trading infrastructure for hedge funds, banks, and brokerages Pejman had spent over 15 years as an experienced investment industry professional. Jubin recently left the position of VP Americas for GFI Trayport which he held for nearly four years, and was responsible for strategic growth and overall operation for the Americas. Jubin has held various positions at Long Term Capital Management, JP Morgan’s Lab Morgan, Oppenheimer & Co., and GFI Group.
FCM360的常务董事Jubin Pejman。在成立FCM360促进土耳其境内对冲基金，银行和经纪公司的低延时交易基础设施之前，Pejman具有超过15年经验的投资行业专家。Jubin最近辞去了担任了4年的GFI Trayport美洲副总之职，在这期间他负责美洲的战略增长和总体运营。Jubin在Long Term资本管理公司，JP摩根的实验室，Oppenheimer有限公司及GFI集团担任过各种职位。
We would like to welcome our speakers today and thank them for attending this exciting webinar. We will go straight into our first question which Robin Manicom from Equinix will answer.
我们对这3位嘉宾的到来表示欢迎并感谢他们参加此次研讨会。我们马上进行每一个问题的讨论，Equinix 的Robin Manicom会对此进行解答。
Robin, in what ways does the crucial need for colocation and proximity to exchanges, liquidity venues, and low latency market data, and news feeds create drivers for change in the way many buy and sell-side FX trading firms currently have their trading architectures structured?
Robin: Thanks Larry, and hello to everybody there. Yes, there has been over the past 5 to 8 years there’s definitely been architectural change in the way that you access your trading venue. We all know about colocation and proximity services which is moving your infrastructure out to where those trading venues are in various data centers around the world.
What this presents is an opportunity to, and I think you summed it up in your opening remarks Larry, it presents organizations an opportunity to be able to compete at a more high performance level, or to reduce their costs, or to look for new revenue opportunities within the ecosystems that are being created around where these trading venues are located, and where the colocation and proximity services exist. This is something we’ve seen traditionally in the equity markets and as of recent years, say the past 3 to 4 years we’ve started to see this type of opportunity arise in the FX markets as well.
Larry: Thank you Robin. I will pass the mic now to Javier Paz, over to you Javier.
Javier: Thank you Larry, it’s good to be with you. So I would say that the cloud and the colocation needs and goals are different for the different firms. We have buy-side firms looking for colocation and proximity to changes as a way to reduce latency and increase transferability.
On the other hand we have sell-side firms that have much more complex interests to contend with when doing the same. The primary driver for them in adopting cloud technology and moving this direction is to reduce capital expenditure costs and being much more precise in terms of which functionality they wish to move to the cloud. This is a time of high regulatory cost and they need to find savings from somewhere, and so they’re starting to seriously look at the hardware and software intensive functions and what they can pass through their risk management and compliance departments, and be able to put those things in the cloud. At the same time, they have to be responsive to what clients are asking for.
Larry: Thanks Javier. Now we move to Jubin Pejman of FCM360.
Jubin: Thanks Larry, it’s good to be here as well. I think it goes down to if it’s a legacy project versus new infrastructure project because most of the banks that we deal with, we’re helping them sit in the middle of connecting to different buy-side and sell-side firms, aggregators, ISVs. As the new algos are being developed, it’s forcing the traders and the brokers to collocate closer to the exchange of the ECNs.
When you look at how the long-term contracts have been written with the data center space and things like that, they’re starting to be forced into that especially as the traders are branching out into new asset classes. It doesn’t make sense for a bank or a hedge fund to run its own operating center in let’s say 3 or 4 different cities when they can go to let’s say a managed hosting provider or a manage services provider, and be able to have the low latency connectivity between those buildings within a region. The clients are demanding it, and it also helps attract the buy-side clients when the sell-side brokers, prime brokers, and whoever’s providing the liquidity to ISV is actually located closer to that execution point.
Larry: Thanks Jubin. The first question again we have for Javier, question number two. How might cloud applications delivered through shared data centers dramatically alter the dynamics of the FX market to reduce entry barriers, and make previously unprofitable lines of business viable for sell-side institutions?
Javier: It raises considerable complexity and adds potential cost in moving from one asset class to offering multiple asset classes. There is a natural evolution that I will see taking place from stock trading to trading in all the asset classes where pricing may not be as efficient, so fixing Kenmore foreign exchange options. And also took places where liquidity is more abundant like in FX.
Brokerage firms, some of them more reluctant to leave than others, are following this strand of adopting multiple asset classes and it’s easier said than done. It’s not just each of the different asset classes, as you need specifications and requirements that again are maybe they like the experience to be able to get up to speed with to be comparative. So they need reliable partners who can quickly and efficiently get them in or out of certain lines of business.
Larry: Thank you. Robin, we’re going to pass the mic to you now on this question please.
Robin: I think Javier mentioned most of the main points there, but I’ll just add that the cloud commercial terms it presents which allows organizations to acquire the infrastructure that they need to conduct this type of trading in co-locational proximity. To get those gains and advantages we talked about in the first question, prior to cloud has always been expensive because you need to look at a large CAPEX spend in the first instance to acquire all of that infrastructure to build it, and to be able to then begin operating and trading.
What the cloud presents is an opportunity, it does dramatically as you say it’ll change the dynamics of the FX market by allowing firms to work in a OPEX commercial model where you’re effectively ranging at far lower rates from the start than in the traditional CAPEX model which means you can build out your infrastructure and get yourself to a point to trading with a much lower amount of initial out rate. I think what it does is it renews the barriers to entry, especially for some of the smaller players who may have been unable to get a viable business place for meeting into this market in the past.
Larry: Jubin, any comments on this question?
Jubin: Yes, it’s an interesting question. But we’re in the middle of viewing with several buy-side institutions that are grappling with this question. Basically we’re seeing the sell-side brokers being able to track sub-brokers and trading customers while they’re competing with the larger banks and prime brokers, collocating the rights facility whereas before, they would never really be able to justify the cost; it didn’t matter.
But at this point being in the right facility, being near the ECNs, especially near OTSFx or the retail FX business where there are a lot of private labels of different ECNs. It’s allowing greater access to people who go in and have that sort of institutional edge, be in the same building, be within a proximity where the other brokers would never be able to do in the past.
When you look at the legislation being passed, you’re looking at equal access, you know the government wants equal access for everyone while there’s some caveats on funds being deposited. The playing field is definitely being leveled with this, and the sell-side brokers would normally not be in the hosting business could definitely attract customers and make more money this way.
Larry: Thank you Jubin. Question number three again for you Jubin. Many sell-side institutions have made huge investments in what might now be considered as legacy architecture in non-collocated environments. How does cloud technology make it more attractive for them to switch to the colocation model?
Jubin: I touched on this a little bit before, but the basic trend that we’re seeing with the customers that we’re dealing with right now, the banks and the brokers are being forced by buy-side clients from wide proximity access or at least have POPs near their matching entrance. So in particular in New York and London, a lot of the major banks that had policies of never having any infrastructure within somebody else’s building; they always had their own infrastructure, their own building, everything – that trend is definitely changing. These guys need the third-party help until the major contracts end and they can migrate over.
At this point, especially with the economy the way it is right now, a lot of banks are taking a lot longer on the approval processes. We’ve been told that a normal two month approval process is taking between 6 to 9 months in the major banks in London. You can go to a third-party provider located in a data center or a managed services provider and provide that access that your buy-side clients need, and we can call that the cloud. The buy-side provider could actually get access to different venues in Chicago, New York, Tokyo, Brazil with one, essentially one cloud provider until they get their own infrastructure in order based on business needs.
Larry: Thank you Jubin; the next question again for you to start. A wide array of options are now available to help FX trading firms connect to exchanges, ECNs, banks, and market data providers. With a wide variety of possible infrastructure setups, why is it so important that networks must be able to connect seamlessly to trading counterparties in a cost-effective way?
Jubin: The way that we resolve this issue of the seamless connectivity, cost-effective way, I mean Robin touched on this before when you turn the CAPEX into OPEX. There’s something a little bit more important here though, it’s important for the client to know that a cloud provider can actually provide them with something that is a real-world hosting, real-world telecom, real-world support, but it’s in that cloud. So wrapped around the cloud, all those things have to be there.
Essentially, we’re calling this the low latency financial class because there’s a very important need for the financial community to have proximity, to have access between Chicago to New York they need on the absolute fastest line. There are connections going from DC to New York and Chicago with the labor and commerce news, there are also westbound European and Asian news coming out of London. You have two major hubs of news right now, DC and London where they’re being distributed throughout the world.
If you can choose the right cloud provider, they can provide you access with all that under one master services agreement and basically get you up and running without all the headache. Again, you’re looking at market data, sources, the news sources, and their FX clients doing triangle arb. Where do you really need to be to make that happen and are you getting the data from DC or London? You’re going to have to go with the cloud solution or something that’s wrapped in a fun way all-inclusive packaged manner to be able to do that. That’s basically what the cloud is doing for the FX traders that we’re helping.
Larry: And just sort of sub question here, when you say the, “FX traders,” are you even seeing individuals, Jubin?
Jubin: Yeah, we have a client that’s trading 10 million dollars out of his house in Houston Texas and he’s paying $30,000 out of pocket for infrastructure.
Larry: Wow, okay so it’s doable?
Jubin: There are a lot of those guys.
Jubin: I mean you’d be surprised actually how many customers, Reuters, and major market data providers they have are 1 and 2 man shops.
Larry: Right. Javier, let’s move the same question over to you, the one we asked in the first place.
Javier: Sure, in terms of setup I think the keyword is seamlessly, be able to connect to the market. You as a firm put yourself in the business of integration, there are various and changing protocols to keep track of. There are new integration or additions, deletions to be made and new APIs that you’re working with.
Data come from these various market sources has to be normalized in a real-time function at speeds that have gotten increasingly closer to zero, shorter and shorter. The task of connecting to its changes goes well beyond the merit connecting and testing some API. A lot could go wrong in the integration process, and it makes a lot of economic sense to have third-party specialists provide a service that has embedded by larger and more sophisticated financial institutions.
Larry: Excellent. Robin, do you have anything to add on this question?
Robin: I think they’ve covered the major points.
Larry: Okay, move onto question five and this is for you Robin. Why does latency matter so much, and where does the real value of data center and colocation service lie by algorithmic and high-frequency FX trading firms? We’re looking for improved access to exchanges, FX execution venues, and sources of FX equity and market data.
Robin: Okay, well I’ll deal with the latency part first. Latency matters because it’s really analogist to if you think back to the open outcry exchange pits that used to operate where if you were a trader and you wanted to conduct a trade with someone else, the other side of the pit, if you could get there the fastest, if you could run there the fastest or even if they were kind of next door to your booth then you pretty much had a chance of beating the rest of the market to securing that trade.
It’s exactly the same in electronic trading in that if you want to be able to see what the market’s doing before everybody else, and therefore react to the market and put your trade onto the venue quickest, latency is the challenge for you. Latency is overcome and we think we all made it by being closer, physically closer and having a shorter network table link to that exchange platform.
You can see why colocation is so important for algorithmic and how you can see it affects trading service because being first on the exchange, seeing the market move the first and getting new orders on is absolutely key to being most effective in the market.
Now, our job really is to ensure that we bring about all those things you’ve spoken about in the past, the cost-effectiveness, the performance gains, the security to meet regulation requirements – we bring all of that to the FX trading world. We have worked with many different FX trading venues in order to bring them into our facilities around the world, our data centers around the world which ultimately provides our customers that platform dense environment in which they can come and collocate against, not just one, but several different FX trading venues which creates market.
Larry: Jubin, do you have something to add there?
Jubin: Sure, we look at this more of a buy-side dilemma than a sell-side dilemma. From the buy-side, the majority of our clients are concerned with what Robin said, getting first on the stack. But essentially we’re at the point now where we can reach the parameter of the matching engine and the lower double digits, if not single double digit below 10 Microsoft package, right?
The first one in line gets filled, right? So you look at it from the main point of the FX trader and the futures trader, the term slippage. What does slippage mean? Slippage means that it took so long for you by the time you looked at your news, your pricing, and where you want to enter your trade and that amount of latency caused you to be last in line. So again, the new location matters a lot if you’re looking at the news, there are machine readable ultra-load latency newsfeeds, there are pricing data location you have to worry about, and then finally the execution engine.
The number one thing in my mind when I look at what the customer is complaining about is that actual slippage which in fact is not getting first in line. How long, far down the run do you get dropped before you get in? And you may not be able to get filled, so if you’re doing a complex strategy where you need to get in and out you need liquidity, you need excellent execution. If you’re not getting filled properly, slippage will kill you on top of brokerage commission. You want to make sure that you have the fair chance to get your execution through.
Javier: I would just add to those great comments that it is important to look for, even though there is no official cost analysis solution being offered, I mean there’s things that come very close to it. But essentially, latency should be understood in terms of or should be tried to be understood in quantifiable ways and field links certainly is the key component for doing that. So just pitch in the fact that whatever solution we are getting from the cloud, a cloud provider or a community, try to get a sense for the tools available to be able to measure the quality of execution that the firm is getting. Some provide better solutions than others in being able to quantify and measure that kind of efficiency.
Larry: Thank you Javier. The next question is for Robin. Robin, why is it important that cloud providers must be able to emulate a traditional hosting model that visually removes all of the complex individual components and makes it easy to set up, trade, and then exit a market in a flexible way?
Robin: Well I think this is the secret to success of cloud business and why there’s so much hype around it. In that is exactly what it’s trying to do, it’s trying to make it seamless for organizations to move into the same type of infrastructure they’re used to, but without presenting the same cost and complexity problems that the traditional infrastructure in their own data centers, managing their own servers and building out their own applications has had in the past. I think it’s true that if the cloud is to be successful or cloud services is to be successful in FX trading, it needs to make it seamless for organizations to move into those cloud services, news, and PREFIX trading.
Javier: Yeah, I would just mention that to render complex individual components into an easy setup it sounds great, it sounds easy, but the fact is if we take a step back and realize that cloud technology is not just an FX thing, it’s not just a financial market’s thing, there’s many cloud providers out there. Certainly it is a complex decision that a firm certainly the larger the firm in terms of where to be positioned.
At the end of the day, the fact boils down to that within these major clouds being formed, multiple issues such as payments, transactions, training. Some cloud providers are better and more specialized for a particular area than others and that’s where high level of discernment has to be made in terms of not just going with the largest cloud, but certainly the one that offers the most relevant and like you said, easy ways to get in and out of a particular market. You don’t want to be the one client teaching a cloud provider how to, what are your needs? They should already know those. Again, there’s companies like Equinix is more better position if you will than others to provide adequate service to firms involving training.
Jubin: I think the number one thing is getting buy-in from decision makers to move into a cloud model and the familiarity to the operational process in which our companies are used to dealing with. It comes down to basic brick and mortar mentality. If they’re familiar with how you present the actual product, they know it’s going to work and it looks like something they’ve been using, although it’s wrapped up in this cloud package with tremendous advantage.
I think also when we look at customers that they’ve got their data center in their telco closet or they’re using some hosting company somewhere in Europe, if you can draw them a logical diagram and show them exactly what they’re getting. You know, this is your security, this is your network layer, this is your layer three, this is your server, these are what the discs are that make up the server, here’s your storage, these are the buildings you can connect to – they all of a sudden can wrap their mind around the cloud and it looks logically what they’re looking for, so there’s a familiarity to move in.
I think a lot of people just don’t understand what cloud is because it’s been such a widely used term in so many different industries. I mean it comes down to basics, it’s like you draw the diagram, everyone sits around and says, “It looks exactly like what we have. But it’ll let us save that money, it’ll let us open up 2 or 3 locations around the United States or maybe one in Asia, one is Europe, one in New York for the traditional cost.” That’s what I think the real value is. That’s the traditional hosting model, it’s a lot of CAPEX; you have to still maintain it. But draw the logical diagram if you want to sell cloud; make people feel comfortable.
Larry: Excellent. Next question which we’ll kick off with Robin Manicom from Equinix. Robin, what levels of investment are typically required to employ data center services, and what steps can be taken by trading firms to shorten colocation deployment timeframes?
Robin: Okay, so I think people will be pleased to hear that the level of investments to move into colocation of proximity trading maybe not be as high as some people out there think. For example, typical spaces and data centers can start at a few thousand dollars a month. But in terms of cloud services that definitely allows that initial capital expenditure expense that is needed to buy all the upfront service which runs into sometimes hundreds of thousands of dollars to really be taking away and move into an operational expense from the word “go”.
I think that approach is moving. I’ve spoken about this before, the traditional CAPEX model to new OPEX model offered by cloud services does help shorten the colocation deployment timeframes for customers who want to move into colocation trading. Also, to add that what has typically in the past always given us long lead times which was the network circuits that had to be established between broker and data centers and the data centers where those trading platforms are which often ran into several months, 2 to 3 months typical timeframe for those sorts of e times.
When inside colocation of proximity services inside a data center, the cross connects which are the connections between cabinets, all customers within the data center, those cross connects are put in place within 24 hour periods. It is quite feasible to assume that connected to the right circumstances, once a contract is signed, you will be up and running, physically connected within a 24 hour period using this colocation of proximity trading training model that exists today.
Jubin: I think Robin just did a great job explaining it. I can go into a little bit more detail based on typical trader or prop shop that approaches us, or even the smaller software company. In the straight colo model you absolutely have a lot of capital expenditure. You can start with like what Robin said, for one cabinet several thousands of dollars. On top of that, we usually see about a 50 to $100,000 expenditure on buying the hardware if you want to do it the straight colo method. You have to also factor in the support, the expertise of the cisco switch, the juniper – you hire somebody. Generally speaking, even a small shop will need what we call one and a half people or one and a half technicians to basically run their infrastructure for them.
Then you go over to the hybrid cloud model, when you talk about hybrid cloud sometimes you’re looking at a fully virtualized service with some piece of standalone like storage or some applications on the trading side, need to get that 10% that you lose when you virtualize back into the performance of the CPUs to the NIC card and things like that. You’re talking about the numbers I mentioned before minus 80% to get up and running.
So your setup cost on the cloud is going to be infinitely lower, and then the actual monthly cost could be as low as 80%. But you’re turning that 50 to $100,000 into sometimes a 5 to $10,000 range to get going which is in the budget of a lot of prop shops and a lot of software companies.
Again, it’s being driven by the demand of the actual clients. So the sell-side guys are selling their software to brokers and saying, “Look, I don’t want to host this thing in my office. I don’t want to host this thing at the data center down the street. I need to be where traders want to be.” Sell-side companies generally don’t like to spend as much money as buy-side, they’re usually very lean, they’re very relationship driven. This will get them into that 5 to $10,000 range.
Larry: Excellent, thanks for that Jubin. Javier, do you have any comments to make on that?
Javier: No, no comments on that. I think it’s explained well.
Larry: Okay, then the next question is once again for you Robin. How important is it for an FX trading operation to have local technicians near the data center, and what are the basic maintenance options available for those firms operating in data centers located in more remote regions of the world?
Robin: Okay, so this is a great question. The FX markets everybody knows a 24 by 6 market, it follows the sun, it exists in all three weekends, mainly out of Tokyo, London, and New York. If you are going to move into FX trading markets and take on the proximity and trading services and all the benefits that we’ve discussed to go with that, then it’s considering how you look after your systems and service your systems is pretty key.
What we try to look at here is making first of all making it fairly seamless to the trading firms from whether they’re in New York trying to deal with a Tokyo installation, or if they’re in London trying to deal with a New York installation. We make sure that the service engineers, the providers part of the colocation proximity services, from not only us, but I know a lot of our key partners that those engineers speak the local language whether it be English. They understand about technology and there’s not those barriers that go with cross-border, cross-regional differences. Which from a communication point of view makes niching after a global network or a global installation very, very difficult when you don’t understand the local cultures, and you can’t really understand; make it kind of difficult to communicate. We go into great lengths to ensure that communication around the world is all fairly similar and that the customers can have a very similar experience wherever they decide to install.
Larry: Thanks. Jubin, go ahead.
Jubin: So we’re looking at the do-it-yourself model right now and how I would describe it is you’re going to have your initial installations, you have your support, your remote hands once in a while, you’re going to need your support staff in general whether local or remote. So that’s where we lean in a remote hand if you need that incremental non-being on the ground.
The costs are very high. When you look at regions that I would categorize as difficult regions to deal with just because of the language barriers, the nature of how the business is run; some of them are regulated by the government, some are not. But in Brazil, Hong Kong, and Singapore we see it very hard for customers to start their own setups. It’s very hard to get the accurate pricing, there’s a taxation in some of those countries that reaches 60 to 100% tariff.
The way that we would recommend doing it unless you have people on the ground there, is to go through a third-party that has personal relationships in the data centers, manages their own infrastructure and we’re talking about cloud, I think the cloud is a good way to go in the sense that Brazil, Hong Kong, Singapore, New York, London, Frankford they can all be connected pretty easily by the right company that has the infrastructure to do it. You can leverage that and you can also leverage the expertise on the ground to make sure that post-installation your systems are working well, that your cross connects are going well and you won’t have to worry about that.
We’re in the middle of a project right now where the client doesn’t have to do anything on a 10 bank installation, doing all the cross connects within an Equinix facility. They basically contracted with us with one NSA and we’re doing the entire project for them, and it’s a cloud solution where there’s a virtualized component to it, there’s security components to it, and there’s fiber running multiple carriers. We’re also connecting that to New Jersey data center, and we’re bringing it down to Brazil for their phase three.
Again, the company has a lot of smart people, they have a lot of good expertise technically; they write software and they charge a lot of money for it. However, they don’t want to bother with this whole, you know they’re not in any of the locations that I mentioned, they’re somewhere else so they don’t have any people there so they lean on a third-party to get it done.
Larry: Thank you. Next question, can smaller FX trading firms get access to data centers, and what are the implications of rack sharing with other companies? Robin, do you want to take off on that one please?
Robin: Okay, so just briefly on this. Yeah at Equinix our smallest unit is one rack so for customers that don’t, that it’s not cost-effective for them to have an entire rack when they only need a fraction of the rack, it surely work with conscious manage service providers which FCM360 is one of those key partners for us. They’re able to offer smaller installations within their own racks I think. I think Jubin would be able to talk a bit more about that now.
Jubin: Sure, like Robin said we do start off at the lower space or slices of the rack as it were. I think it’s primarily driven by the amount of people that started approaching us, they were starting hedge funds, starting prop shops, and they’ve left a large bank or a large hedge fund. They have plenty of money to trade, enough to get their account started with the prime broker; they’ve got the market data feeds.
The one thing that they really don’t want to burn their money on is spending 10 to $20,000 a month to do it themselves. We basically come in and we essentially provide them with the hardware firewalls and provide them with a tier 1 IP. We can provide them access to cross connect into different liquidity providers.
We just recently connected fiber to NASDAQ in New York. So we were starting to see more FX traders doing news trading off the same server, but they’re hitting NASDAQ for the equity markets. So they’re using the same server, they’re using the same internet, they’re reaching out and getting one extra cross connect that we’re providing to them. And they’re able to take advantage of news coming out of BC, they’re taking advantage of news coming out of London, and University of Michigan through an exclusive that the market data provider has.
Again, they’re not actually getting more servers and I had this conversation yesterday with a client, they’re actually starting to trade NASDAQ and BATS, and those are two public exchanges and they’re not going to have to spend much more money, just the cross connect fee.
Again, I think you’re taking advantage of economy to scale when you go with third-party. The money’s been invested and often times when you need the cream of the crop latency between major cities like Chicago, New York, London, Frankford, the third-party will very likely be able to cost-effectively give that to you as opposed to you entering into a three year agreement for a 10 gig pipe with a carrier. I mean it’ll take you between 6 to 12 months to get that infrastructure set up yourself all in.
Larry: Sounds like a big saving obviously. Next question here once again for Robin. Robin, some buy-side institutions seem reluctant to embrace cloud technology. How can a premium cloud solution go some distance to allay their concerns?
Robin: Okay, so if I can be bold enough to put some mutes on the phones here. Everybody has been talking about cloud for what seems years now without much sort of tangibility to what cloud actually is. Well, in financial services it’s different to cloud in retail space. So you’re a developer who’s developing iPhone apps for example and using Amazon net services in the cloud for that, that’s one thing. That may not apply and probably doesn’t to the financial markets.
The public cloud is not a secure environment, its shared infrastructure, it’s unknown in what location that CAD infrastructure exists in terms of the data center. Is it next to the trading venues to take advantage of the low latency side of the proximity and colocation trading? These are all unknowns.
I think cloud in the financial services industry is going to be a lot more defined and a lot more specific. It’s going to make goals of security. We may be talking about cages of cloud of infrastructure which is dedicated to say one broker in particular or one hedge fund and that would become a private cloud. We may be talking about cages of cloud’s infrastructure provided by an exchange for a number of trading companies to use in what is called a community cloud for example. Then there may be the actual location of that equipment which used really be next to where the FX trading venues are to make use of the performance and also the cost benefits of having cross connects versus network connections.
I think once the cloud providers and banks really start to define what is needed in cloud services financial institutions, we will start to see buy-side institutions understanding it better, not fearing it so much or maybe understanding that it has real benefit to them and it’s more tangible. I definitely see we will be getting more of the buy-side coming into trading FX in the cloud in the future, if we get that recipe right.
Larry: Jubin, do you have anything to add to that?
Jubin: Well yeah, I mean I could tell you that a lot of financial companies approach us regarding this topic. What we wind up having a conversation about is often times a public cloud, some cost per computing. What we quickly say to them is, “How secure is that? What’s the latency? What’s the geography? What will happen to your data?” You’d be surprised that so many people at a high level in these institutions say, “We never really thought about that. We didn’t know that paying for CPU cycles, cloud or something like that is really not secure.” And you ask them the logical question, “Why would it be secure? You’re paying them 25 cents a month if you don’t use it to maintain an infrastructure, so what is the advantage of them to really give you what you really need?”
So we move into again, the logical diagram: what’s the security? What’s the latency? Well we can draw a diagram and say, “This is where our low latency financial cloud is located,” and it turns into a private cloud solution or conversation. So then once you look at the location of a lot of the cloud providers, you realize that they’re nowhere near where a trader wants to be. And only through trial and error, will a trader find out what the latency is. You’ve got a massive shared infrastructure right now on the so-called discount cloud.
And then the question goes the premium cloud will be a private cloud, it will be fully licensed software, it will be using really modern architecture on the servers, the SANs will be very expensive and everything. But you only pay, since the CAPEX has already been spent by the hosting company, you could get a slice of that.
The last thing I’ll add with this is there are a lot of cloud computing companies that are targeting the financial vertical and they have no clue where they need to be to sell. Once a large company approaches them and they cannot disclose the location or they do, they wind up losing the business and then they go back to: how do we give ourselves a premium solution? Go back to an Equinix, go back to the carriers and figure out how to connect this all together. Using the cloud, you can tie it all into one piece.
Larry: Excellent there. Javier, do you have anything to add to this question?
Javier: Well if anything in addition to what has been said, I would just say that even as much as we use logic, there’s always going to be some laggers. A significant portion of the buy-side still is in that category where they’re either concerned for safety and security is so high they are not going to necessarily adopt the cloud technology head on.
What I think will happen as Robin indicated, we’re going to see the emergence of kind of a mixed ecosystem with hybrid, public, and private clouds coexisting or creating within compartments in this environment at the highest level, allows the sharing of information efficiency, but again, with less of a security concern.
At the other end, you’ll have more specific clients, those dealing with either risk management or proprietary training algos being caged and protected in a very efficient way. But again, this adoption will not happen overnight, but more along the lines of gradually as they understand and warm up to the technology and the adoption rates that continue to rise and rise.
Larry: Thank you Javier. Next question for Jubin. Please give some practical applications or examples of why a cloud solution would be preferable for EFX deployment, as opposed to normal colocation deployment and what factors might influence this choice?
Jubin: I think it comes down really to cost, there’s a start-up cost and an ongoing maintenance they say in the business the MRC, the monthly recurring cost. There are different components that you have to worry about. You have the pure data center cost if you’re going ahead to do it yourself. You have the telco cost, the ISP, you have software, all of that. I think in the OTC markets it presents more of a problem because of the disparate liquidity where to truly be effective, you would need a colocation cabinet with that entire infrastructure being reproduced in each location. Let’s say there are 3 or 4 locations, where we can go into a cloud solution.
A customer comes to mind, they actually they’re a Forex trading shop, they had been trading with some manage services outside of the primary buildings you want to be in; their strategies were making money. We moved them into a cost-effective solution, we can call the cloud solution because everything was there, we literally just turned them up. They came back to us and complained that the ECMs weren’t really giving them the response. I said, “Well you know, what we could do is we could move you into a better location.” Immediately they went from 500 microseconds to 150 microseconds. We were able to do it without additional startup fees for them, we literally just migrated them over to another building. At this point, their model’s working and they’ve been working on trying to get rid of flash latency.
So we’re looking at different 10 gig cards that are completely offloaded from the CPU, it eliminates that microsecond flash latency they’re getting. They would have never been able to afford it to do it themselves. They had already spoken to about 10 providers before they came to us. At this stage in the game, I’m happy to say they’re at about $1,800 a month total expense be 150 mics away from one menu, 50 mics away from the other one and they’re developing the triangle OB [SP 0:50:17.0] model to success and they’re further getting profitability on a directional model.
Again, quad solution if you guys, they have the money to trade the strategy, but what do I do now? That’s really what it comes down to. I think for that question, I’m not mentioning the company’s name but that’s the perfect solution for them.
Larry: I mean $1,500 a month sounds amazing. What would that have cost them to do themselves? Ballpark.
Jubin: They’d probably be at 7 to $10,000 monthly range if they were doing it to our level.
Larry: And the CAPEX?
Jubin: Well, I’d have to say that they may be in the $40,000 to $50,000 range. But again, they’re taking advantage of an infrastructure that costs exponentially more that they would never have access to from a telco specific. Again, if they’re not in the right building, they’re going to be paying between $3,000 to $5,000 a month on a 2 to 3 year term for the right telco to go to the right building.
Whereas we can piggyback off our infrastructure and get the lowest latency between Manhattan and New Jersey which is around 170 mics one way. Switch the switch 170 mics, there’s one line, there’s one provider and who’s going to pay for it? You’re not going to go and sign a three year contract, especially when every one of the customers says, “What happens if I go broke in three months, what happens then?” They’re not going to a carrier and signing a three year contract for that.
Larry: Just for the sake of, Jubin, for some people that don’t understand what is a millisecond, a microsecond, and a nanosecond. I know you refer to mics as microseconds.
Larry: Could you just give us a brief explanation of those three terms? I know that…
Larry: Most people will know, but there will be some probably won’t. For the benefit of those people, it might be helpful just to mention.
Jubin: Yeah, sure. 1,000 milliseconds is one second. It takes you 200 milliseconds approximately to blink your eye. The majority of our clients are happy trading at under a millisecond which means technically it’s 1/200th of an eye blink. Once they get into that level, they start actually getting stability and they say, “Well, what about microseconds?”
A microsecond, a 1,000 of those is a millisecond so you’re looking at; you’re removing essentially four zeros(0’s) from a second when you look at it in microsecond time, right? I don’t know if I did the math right, but the point is if you’re actually now at 150 microseconds away from a matching engine and you are on average in the metro area of New York let’s say, between 7 to 10 milliseconds – you essentially divided it by 20.
As it goes, you’re paying money for a newsfeed, you’re paying money for the market data, you’re making a decision or actually let me clarify, your algorithm or your software is making decisions to enter a long or short position in whatever market you’re in. You are going to be at that one millisecond roundtrip range probably better than 90% of the traders around the world, okay?
And then once you break it down into the 800, 900 microseconds it gets really exciting because exponentially you’re beating people. By the time you’re in the 5 to 50 or sub-100 microsecond latency, you’re really in the 1% or the half of 1% of traders around the world that even has had access to that. By the way, there are a lot of larger firms that don’t have that because they haven’t needed it traditionally because they’ve got better access, and better strategies, and things like that to make money. You’re buying yourself an edge when you go into that realm.
Larry: All right, so basically a millisecond is a 1,000 and a microsecond is 1/10,000 per second, right?
Jubin: A 1,000 microseconds equals 1 millisecond, a 1,000 so it’s a million. A million microseconds is 1 second.
Jubin: So you’re going from, yeah we do it on the spreadsheets and figure out…
Robin: I was going to say 10 to the minus 3 is a millisecond, and 10 to the minus 6 is a microsecond.
Jubin: Six 0’s, there you go.
Robin: The one I like specifically is with these markets moving so fast, that if you think at the moment you can go back and forth between London and New York six times, so that’s six times to New York and back again in the blink of an eye over the current speed of networks across the Atlantic. If you think that that’s the speed at which networks are operating, then if you’re in a data center and you are within 100 yards of say for argument of the exchange, instead of that distance between London and New York then you’re actually 100 yards away, you can imagine the speed at which you can operate your systems. I always find it easier to sort of think about it in those terms rather than the 10 to the minus 3, 10 to the minus 6 which I think I run out of fingers somewhere along the lines.
Jubin: If you don’t mind Larry, there may be some listeners that are interested because we get a lot of phone calls regarding the thing that I’m mentioning right now. There are a lot of people trying to gain accessibility to the FX markets in the New York metro region and they’re using discount hosts or they’re using maybe not so much discount hosts, but they’re still paying in the 500 to $700 a month range. They’re getting around 7 to 10 milliseconds to a lot of the venues. Keep that in mind, that if you find the right provider it’s actually not that much more expensive.
A lot of these companies have 2 or t3 locations to get the proximity. Whereas if you’re in a location let’s say in Secaucus, New Jersey, you can literally get the absolute fastest connection to an ECN, for the most part any of the ECNs, and you’re really using that one server. So it becomes a lot cheaper to have yourself in a cloud solution that already has like Robin mentioned, the cross connect feature as opposed to the telco.
The last thing I’ll say is that 99% of the data centers that you’ll go in to get a hosting, they will absolutely never allow you to connect anything other than the internet, and you’re at their mercy. Again, go on the Equinix list and you can see a lot of those providers. I mean I’ll ploy myself for one second, we’re one of them but we really do take a consultative approach to figure out what someone needs. We tell them if they’re wasting money. We can’t do that without the cloud, right? Because otherwise, we’d have to turn up each client and have a huge CAPEX.
Larry: Right, thanks for that. I think it provides a bit of a framework for some people who are not intimately familiar with these terms. Final question and we’ll kick off with you Javier. Colocation and proximity hosting are clearly geographically sensitive, especially FX. What considerations come into play in deciding where to locate, and are multi-continent locations necessary to follow the market around the clock?
Javier: Sure, well thank you, yes. I think yes, there is that colocation and proximity geographic requirement where if you are truly operating in, if your algos or if your clients are located in multiple geographies, you can have the full advantage three in particular there’s London, New York, and Tokyo where you have to be in. If you want to take it a step further, Sydney. But again, it’s very limited, much more limited if you will, group that you’d be influencing.
As we mentioned before, it’s kind of following the sun. And I would mention in a couple of different ways. One is the physical sun, but in other ways it’s also following the proximity too, just like in French have their French God, their sun God so in here there’s the bigger boys to follow, one of the bigger boys and to be at, and try to be in close proximity to them.
That’s really, if anything that you remember from this webinar, basically try to be present in those three locations and try to be in close proximity as much as you can to the bigger firms providing most liquidity to the market, particularly in the interview and market. Try to do business and be collocated with firms that understand your business across asset classes and make it easy to get in and out of asset classes and markets, and do so with again cost efficiency and again customized approach to doing business.
Larry: Robin, any final words on that last question?
Robin: I think Javier covered most of the details on that.
Larry: Okay and Jubin, would you like a final say here?
Jubin: Yeah, I’ll give you without mentioning client’s names a lot of the problems that are similar to clients across board. So we’ve got a possibility of a multi-continent solution, but it may be hybrid depending on what the execution venue is, what pricing feed or pricing engine you’re using, and if you’re using the low latency news or machine readable feed. I’m breaking this down into kind of like strategy without talking about it, right?
But you got New York, London, Tokyo. The major players are definitely consolidating their centers, major players being the banks, the ECMs, the brokers and I think it makes sense sometimes to just look at: is there a tradeoff? If you’re trading news coming out of London, you don’t want your execution, your server to be somewhere if you’re trading an ECM located in New York. So likewise, if you are looking at market data coming out of let’s say the US and you’re trading one of the, there’s several large execution venues in London but they’re not the same, you don’t necessarily want to have two servers. You may actually just find a place to just have that.
I think it’s important to understand you are embarking on a project, you have ‘x’ amount of capital to spend on a month or a monthly budget, what is the best blend? Do you really need to have three locations or can you live with 1 or 2? Because if the market data’s coming from one and the execution venue is in a completely different place, that matters. Again, the whole concept of the cloud and FX, the pricing engines often times are coming from the same players that are known as liquidity providers, and there are a handful of prime brokers and ECNs providing that feed. Look very carefully to make sure that you choose the right feed because you may be able to be in one spot, and yet the best of all worlds even if you’re taking data in or news events from another country.
Larry: Well gentlemen, speakers and people who’ve logged on today to attend – thanks very much. I know there have been a number of questions posted, but unfortunately because we’re right on the hour here, we’ll be unable to take those today because we’ve arrived on our one hour timeline here. But you can and do please feel free to contact any of the speakers shown on the list here. I’ll read out their contact details in a second so that you can contact them directly by telephone or by email, and of course have a good look around the very interesting websites that they have and the services that they provide.
Finally, if you’ve missed any part of this or you wish to listen to parts of it again because there’s a lot to absorb and it’s very interesting, or you wish to pass the data onto a friend or a colleague, the webinar will be available in the next few days online for you to listen to for a period of time.
Thank you to our speakers and thank you to our attendees today for attending this event.
Jubin: Thank you.
Robin: Thank you.
Javier: Thanks Larry.