NYSE Euronext's new cloud services will open for business July 1, from its Mahwah, N.J., data center, letting companies buy pay-per-use computing with functionality and data honed to the needs of capital markets firms. The service's launch sparked speculation that more clouds will follow that are customized to the specific needs of different industries.
If the special-purpose cloud works in financial services, the thinking goes, would it work in healthcare, pharmaceuticals, public utilities, maybe even the airlines? Or will specialized clouds work for companies with particular technology needs, like security or high availability? "What we're seeing now is the first wave of special, fit-for-purpose clouds--community clouds--that are bringing together not just technology but the business intelligence, market access, and customer relationships, and truly delivering a full value set," says Howard Elias, president and COO of EMC, whose storage gear is used in the NYSE Euronext center.
Yet NYSE Euronext might be the exception, not the rule. Infrastructure as a service like Amazon's Elastic Compute Cloud did establish that companies want pay-per-use computing power, and it's true that general providers like Amazon can't deliver some of the more specialized--and higher-cost--services some industries need. But a close look at the special pivot point that NYSE occupies in financial services, and the new services it will provide, suggests that other industry-specific clouds will be tough to create. NYSE Euronext's plan is a bold move, but there are a number of reasons that approach may not work elsewhere. Instead, where specialized clouds do emerge, they're more likely to develop around technology or functions--like high security or data backup--than industries.
NYSE Euronext's cloud platform will provide online computing services for a number of capital markets functions. For example, by offering rapid provisioning of pay-per-use processing power combined with access to historical market data, it could be a platform for back-testing trading strategies, to see if they would have been profitable in past market cycles.
The platform might open a new market segment for NYSE. Already, it has a healthy colocation business, in which major trading houses rent space in NYSE's data center to be in close physical proximity to the exchange's trading engines, sparing the milliseconds it takes to go from a company's own data center to NYSE's. In automated trading, those split-second advantages actually matter.
NYSE hopes its new platform's pay-per-use approach will appeal to firms that don't trade enough to justify putting permanent servers in its data center, but still want low latency trading. "This opens up the addressable market," says Ken Barnes, senior VP of global platform services for NYSE Technologies, the business unit charged with marketing the exchange's technology services. "There is a wide array of parties that are willing to run a trading system on this type of infrastructure."
A colocated trader must be a member of the NYSE, and a seat costs about $1 million. Barnes believes some license holders will make their trading systems available and rent their identification number to customers, giving what may be a small trading desk a few minutes or hours of access to financial markets on a footing that's only slightly slower than high-performance trading shops. Multinational companies, for example, have a treasurer's office managing cash, short-term investments, currencies, and commodities that do trading that's often less time sensitive and not as frequent as that of Wall Street firms.
NYSE won't disclose what it will charge for its new on-demand services, only that the rates are higher than those of an infrastructure provider like Amazon, which charges 8.5 cents an hour for a Linux server. But the fees are much less than the price of admission to access the matching engine. "People who would normally say, 'I can't afford a colocation operation' can quickly set up a server in the OnDemand Compute Cloud," says Barnes.
The cloud infrastructure won't be quite as fast as dedicated servers. With colocation, the cables connecting each trader to the trading engines are the same length, to eliminate the possibility that one has even the slightest advantage over another. Likewise, NYSE cloud infrastructure will be equidistant. But using virtual machines on shared servers, while lowering costs, adds latency. NYSE will offer, for a higher price, the option of putting a trader on its own on-demand x86 servers, with no virtualization involved.
NYSE promises that its new services will be secure, employing VMware-based vShield security systems for the virtualization layer and a private network in lieu of the Internet. Beyond that, NYSE declines to go into detail on those technologies.
When NYSE announced its cloud June 1, it had only two small firms as beta customers, Millennium and Pico. It's possible other firms are testing the service under wraps; if so, NYSE didn't mention them.
However, NYSE doesn't need huge volume to support its offering. NYSE already had to build its data center to run its conventional exchange business, and it left a huge, unfinished section of the building for the growth of its colocation and other technology services. Anything its cloud services bring in "is gravy," not a required return on investment, Barnes says. He declines to set targets for that business, but NYSE Technologies has a goal of doubling overall revenue, to more than $1 billion in 2015.
So let's add up NYSE's advantages in this market, as a sort of checklist to see if other industries could have a similar player: It has a market opportunity that could let it expand its customer base, as well as provide services to existing customers. It's a highly regulated and security-conscious industry, and NYSE is broadly accepted as an honest broker among very competitive parties. It has data it can offer along with its technical services--data that would be difficult for customers to duplicate on their own. And, it could offer this without investing a lot of capital beyond what it needs to run an existing, profitable business.
You can come up with a few companies in a few industries that can match that mix of advantages, but not many. That's reason for skepticism when people start talking about a future filled with industry-focused clouds.
Specialized Technology Clouds
While NYSE's model will be hard to duplicate, we will see the market for online IT services branch into specialties, as vendors try to differentiate and earn higher margins. Vendors already are testing this model, with some early efforts focusing more on technology or business function specialization than vertical industry platforms.
SunGard is considering offering data recovery as an online service. SunGard believes Amazon EC2-style infrastructure as a service will be widely available by the end of 2012, with businesses of all kinds making use of it. As companies get comfortable using online computing, either for failover or to deal with spikes in demand for a particular application, SunGard sees a big opportunity in delivering data replication and system recovery as an online service from shared computing resources--what it's calling a replication cloud. Such services could provide a recovery mechanism for critical systems, whether they're running in enterprise data centers or on infrastructure as a service.
If a financial application fails, for example, its owner wants it back up and running in a matter of minutes. That's possible with identical virtual machines waiting to be spun up in a secondary data center, a service SunGard already offers, but those VMs also need immediate access to the failed application's stateful data--to know whether a transaction has been completed or must be rerun, for example. A recovered system could get swift access to stateful data if there were a continuous mirror image fed into a remote data center that also holds the backup virtual machines, says Indu Kodukula, CTO and executive VP of products at SunGard. Replicated data is also needed for so-called cloud bursting--when infrastructure as a service takes over for in-house servers that run out of capacity. Those online servers will need stateful data to spring into action. Kodukula says customers of such a service might even pay based on the speed of recovery.
Kodukula wouldn't say when SunGard might make such a recovery service available. That kind of service wouldn't be created just for one industry, but companies with high-availability needs, like financial services or telecom firms, would likely gravitate to it.
Likewise, Harris Corp.'s recently announced Cyber Integration Center in northern Virginia is designed as a highly secure data center to offer infrastructure as a service not for any one industry. That said, the center's location and design, along with Harris' history as a government and military contractor, mean those sectors will likely be prime market segments.
Harris' new data center is just far enough from Washington, D.C., to be outside the immediate blast zone of a nuclear strike. It's hypersecure operations result from software systems that are inspected periodically for their digital "fingerprint"--looking at the configuration and workload and comparing it with the original fingerprint in a system registry. It's a strategy of not just securing the perimeter against attacks, but of "deep and pervasive monitoring of running systems," says Rich Plane, head of development and delivery for the center (and Harris' former CTO).
Harris hopes that approach will convince otherwise skittish companies and government agencies to try a cloud platform. The center's security and reliability features could make it attractive to healthcare providers, as well. Harris guarantees its center will provide HIPAA-compliant data handling and meet regulatory requirements such as Payment Card Industry standards.
Google App Engine might be yet another kind of specialized cloud. While App Engine, like Amazon EC2, can be used for generic infrastructure as a service, it also can be seen as something of a specialized cloud for running apps that rely on data from Google's search or location services, which the company reveals through APIs.
Google's new prediction API, for example, combines features of location awareness with buyer preferences. An application using the API could let a person give the names of favorite restaurants, then collect data on the restaurants online and use the data as a baseline. When that person enters the name of a new restaurant in the app, it goes through the API to use the baseline data to find characteristics at the restaurant that match the user's preferences. It's a way of leveraging Google's search infrastructure that might be embraced by the tourism, restaurant, or retail industries.
Another example is the Google "prospective search" API. An application using it would ask a user about queries the person is interested in answering, build an index of them, then tell the prospective search API to examine incoming documents and check to see if they have information relevant to the query. It's like search in reverse, and could be used by industries where customer service is vital, such as automatically looking for dissatisfied customers in user comments on the Web. Or, it might be of use to a crisis management team that needs to know all information coming into a company relevant to a particular problem.
Vendors are in the early stages of developing specialized clouds. One of the top requests businesses have for App Engine is that Google bring it out of preview mode and support it as a full- fledged product. That will happen "later this year," says Greg D'Alesandre, senior product manager for Google App Engine. The company has dropped plans for a business-oriented version, called App Engine for Business, and instead is concentrating on having the core product meet business needs.
SITA, an IT services firm specializing in the airline industry, earlier this year announced it would develop a cloud platform for the airline industry, promising everything from infrastructure and platform services to software and desktops as a service, tailored to the industry's needs. But what we haven't seen, and aren't likely to see thrive, is a cooperative industry cloud platform--something like pharma companies or banks building or sharing a data center resources in a partnership to provide shared, online computing resources.
Why not? It's true that a strong regulatory climate gives a group of industry players a well defined arena in which they must operate, and that in turn sets potential terms for pooled services to emerge. But someone has to pay for the initial investment, supervise the cloud's development, and accept the risks of working in an emerging technology. NYSE, with its clear business case for reaching new customers, was in an unusually strong position to make such an investment.
Asked what other companies may offer a cloud platform for capital markets, Barnes ticks off the advantages NYSE Technologies had heading into the effort: "You need to run hard-core IT trading operations. You need data centers. You need interesting applications, interesting content, and a captive audience." Drop the "trading" piece from hard-core IT operations, and Barnes has offered a short list of what an entity might need for an industry cloud. That last requirement, the captive audience, could be hard to duplicate in pharmaceuticals, healthcare, the airline industry, utilities or most any other regulated industry.
NYSE's Barnes says a number of IT services companies could build a cloud tailored to a particular industry, but he contends, "They don't understand the unique attributes of our industry." It's true that business knowledge has to be intimately meshed with the technical knowledge, and that an architect with a deep understanding of the industry needs to design the vertical industry cloud.
Don't count out large IT service providers, if a critical mass of clients urge them to provide clouds tailored to an industry or technical need. How much critical mass that would require is the big question. More specialized industry clouds are likely to emerge, but they must command some key pivot point in their industry to thrive. In most industries, few such points exist.