August 07, 2006
GemStone EVP on Maximizing Grid, SOA Deployments
In this Q&A with GRIDtoday, GemStone Systems executive vice president of marketing and strategy Shankar Iyer discusses his company's Enterprise Data Fabric (EDF). Iyer explains where EDF fits into today's modern architectures (Grid, cluster, SOA, etc.), as well as in what markets the company sees the strongest demand and the most potential for the solution.
GRIDtoday: First, can you give me your definition of an Enterprise Data Fabric?
SHANKAR IYER: An Enterprise Data Fabric (EDF) is a distributed, operational data layer that serves as a foundational information infrastructure for the demanding needs of modern day architectures such as SOA, Grid/cluster computing and OLTP applications. Envision an EDF as an active data network which harnesses memory and disk across several hundreds or even thousands of distributed nodes to store, distribute and analyze data and deliver it to any application in a fast, consistent and reliable manner. An EDF holds the relevant "operational data subset" required for a specific workflow, transaction or computation close to the point of use, enabling highly responsive, information-driven applications across a distributed network.
Gt: How does an EDF complement distributed technologies like Grid computing and SOA?
IYER: SOA and Grid computing initiatives often present significant data management challenges. These are distributed deployments that require data from multiple sources to be provisioned onto the grid or to the services in a dynamic manner to support reliable and speedy business operations.
An EDF provides a distributed data layer required to maximize SOA and Grid deployments, significantly improving data virtualization, access and delivery to meet the demanding run-time metrics of these environments. Applications can scale transparently with an EDF and manage incremental additions of new users and new data sources, as well as increasing data volumes. An EDF offers guaranteed data access and reliability through its support for n-way replication and built-in persistence schemes for critical data elements. The flexible storage topologies of an EDF enable data to be stored and accessed in a fine- grained manner, improving application efficiency. Finally, an EDF can capture streaming "data in motion" and the temporal value of dynamic data, which is a requirement to support real-time decisions so common in analytic applications today.
Gt: How does an EDF utilize virtualization to manage data?
IYER: By acting as the uniform, universal data resource for distributed applications, the EDF virtualizes access to the underlying data sources through location transparency and multi-format support.
An EDF enables any enterprise node to transparently access data from any other node in a location-independent manner, without knowledge of actual data source or access protocol. Behind the scenes, data can be fetched from a peer node, from disk, or from a database.
An EDF can also deliver data to applications in the specific format they require, be it objects, XML, tables or events (data in motion). By encapsulating the necessary "operational data" for business processes, an EDF provides a layer of virtualized operational intelligence.
Gt: How does GemStone's EDF fit into the overall concept of a fabric?
IYER: GemStone's GemFire suite embodies all the characteristics of an enterprise data fabric. GemFire offers distributed data caching for storing objects, XML documents and relational tables while ensuring reliable data distribution and transport with messaging-like semantics. The event processing and SQL-based continuous querying functionality of GemFire enable intelligent data analysis. GemFire also provides open APIs for connectivity to backend data sources of any kind and allows data to be accessed via multiple APIs such as JCache, SQL, JDBC, C#/.NET, C++, XPath, SOAP and XML:DB.
Gt: Other vendors offer what they call "application fabrics." How does a data fabric differ from an application fabric? How are they similar?
IYER: The term "fabric" applies to both in that the fabric is pervasive and virtualizes applications or data regardless of the underlying hardware resources on which they are provisioned. However, they have very different pieces of functionality and fit at very different levels in an IT infrastructure. A data fabric is very much part of an application architecture and is something that's much more than a convenience. It offers new, additional pieces of functionality not found in existing data infrastructure today, without which it's impossible to deploy highly responsive real-time business processes. My understanding of an application fabric is that it makes it easy to provision applications on a Grid- like infrastructure by providing fault tolerance and operational management features across multiple OSs and platforms.
Gt: What kind of demand have you seen for your EDF, and how has this manifested itself in terms of customer wins, etc?
IYER: GemStone's mission is to identify large-scale business problems and drive demand across a wide range of enterprises for this category of infrastructure software that we and the industry call a "data fabric." We chose to pursue financial services enterprises and the federal sector, where the three-fold business drivers of "high speed operational efficiency," "deep customer intimacy" and "real-time compliance and risk" manifested themselves the most.
Our next focus was to build breadth and significant depth in our technology and product offering. We are very proud of our foray and success thus far. We have a significant number of the top 20 financial services firms as our customers and expect these customers to be deploying us across the enterprise as a true fabric. In that sense, we have differentiated ourselves from other vendors that have approached this problem piecemeal vs. a complete solution. We believe that solving the hardest problems on Wall Street, in London and Washington, D.C., has helped us create a compelling product suite with a significant level of differentiation -- both functionality-wise and attribute-wise. As a result, we are seeing tremendous traction in other industry segments, as well.
In addition, our corporate history of supporting mission-critical global installations 24x7 and our close working relationships with major partners such as IBM and Intel offers us an advantage that will help us drive significant adoption for this category of infrastructure software.
Gt: In what markets have you seen the greatest demand for EDF and similar technologies? Are there any specific markets you are looking to target in the future?
IYER: We believe that financial services, in particular capital markets will drive the greatest demand for the EDF in the short-term. The other sectors of the financial services industry such as asset management firms, retail institutions and insurance companies are starting to adopt the EDF in certain projects such as online services, real-time portfolio management and compliance.
Real-time operational intelligence is an area that we see gaining significant traction in our government business, and we expect this to start to play out in the commercial sectors, especially in sectors where there is inventory replenishment such as retail, consumer packaged goods and certain parts of manufacturing. We are also seeing strong demand in the online portals and media industry segment, as content delivery is a key challenge for companies in that market. Similarly, logistics and supply chains are other areas where increasingly decisions are being made at the "edge" of the enterprise. This decentralized decision making approach is driving the need for an intelligent data fabric to be present both at the edge and the core of the enterprise.
Gt: What do you see in the future for data fabrics and, more generally, data grids? How do you see demand for these solutions growing as Grid computing, SOA, etc., continue to evolve?
IYER: In the future, we see the data fabric emerging at the only data layer for all distributed applications, since by definition almost all applications are required to respond at the "right" time. This data layer will disintermediate existing sources of data such as databases. The data fabric is the manifestation of the notion that the "network is the database." Clearly, SOA and Grid computing are two drivers that require the data fabric. We view "data grids" as one manifestation of the data fabric when you deploy grids. In addition, we see that technology initiatives such as operational business intelligence, real- time analytics and decision support, straight-through processing, and online portals will all require their own specific manifestations and implementations of the data fabric.
Gt: Is there anything else you'd like to add?
IYER: It's apparent from previous history that technology adoption happens in waves. We are in the midst of a change in technology adoption, where the hardware evolution toward utility computing and virtualization coupled with the move to Grid computing and SOA has necessitated that old paradigms be challenged. The emergence of an EDF as an infrastructure is a result of this requirement to deliver a truly operationally intelligent, active, distributed data infrastructure that supports these highly information-based initiatives. We, along with other vendors and analysts that have helped create this space, welcome this development, as it puts true innovation to work to solve practical problems.