For tier-one asset managers looking for advanced, low-latency trading capabilities, however, cloud-based trading services have been a much tougher sell. Although the buy side is under increased pressure to streamline trading operations and reduce operational overhead, the benefits of traditional application service provider (ASP) systems have been offset by their functional and architectural limitations, including latency concerns. For example:
-- Limited functionality. ASP trading systems, particularly those offered by brokers, lack the kind of trading functionality that buy-side traders need in today's increasingly complex market environment. Portfolio trading in particular is something that ASP vendors have struggled to support over the years, due in part to legacy technology that was originally designed to support single-stock order execution.
-- No customization. ASP system providers generally offer a standard, non-customizable product. Although this utility-like model helps ASP vendors scale, it does not address the needs of clients with more complex trade workflows. Firms that want customized trading solutions and systems integration beyond standard order management system (OMS) connectivity need to look elsewhere.
-- Performance limitations. ASP systems generally rely on a shared infrastructure model. As a result, all client trades are executed in a single, centralized trading environment. Periods of high volatility can place a significant strain on the system as a whole, slowing trading for all users, regardless of their trading activity. Like an overburdened electrical grid during the summertime, even moderate users will be subject to brownouts caused by system stress.
In short, cloud-based trading solutions have never been appropriate for low-latency trading. For high-speed trading firms, the logical alternative is the locally deployed, enterprise-class trading system. Deployment in third-party hosting facilities certainly has been an attractive option, particularly for the proximity advantages, but referring to this model as a true cloud solution would be inaccurate. Servers must be requisitioned and racked in the data center, market data connections enabled and tested, access to brokers certified--essentially, the same process a client would go through to install the system locally.