Can the private cloud be used for real-time analytics?

The promise of the private cloud is tremendous. The idea of accessing resources, information, and software – on demand – through an in-house virtualized infrastructure is alluring: it offers the scalability, resource management and utilization, and time-to-market of the public cloud, but without giving up control, security, and payments to an outside service provider.

The quandary arises with the realization that the main benefits of the private cloud are usually limited to web applications rather than mission-critical applications the enterprise most depends on, such as analytics. Web applications are relatively simple, with each application running on a single server, and even multiple applications running on one server.

Financial services companies today must manage an increasing portfolio of applications that require higher throughput and lower latency - such as VaR calculations, P&L, Reconciliation, and more. Analytics processes are much more complex than simple web applications, and present unique challenges: 1) Analytics apps work on terabytes of data, from various sources and in multiple formats; 2) The same data often needs to be available to different analytics processes, leading to competition for resources. Also, the location of the data vs. the location of the process is often a factor, complicating system and resource manageme nt; 3) The need to gain meaningful information from the vast data streams in real time is becoming increasingly mandatory - and running complex algorithms on vast amounts of data is computation-intensive.

To handle the vast compute power needed to run complex analytical processes, many companies invest in extensive grid-based data centers. A computing grid enables dividing analytics process into batches, and running each batch on multiple resources in parallel - reducing processing time from days to hours and even minutes.

But compute grids are expensive, and therefore only available to organizations with considerable spend power. More importantly, the grid - based on traditional tier-based architecture with a data-store tier - does not provide all the performance necessary to handle the tera-data/real-time challenge. It does not make the data easily available for applications to consistently utilize in real time, nor does it provide the resource management required by complex applications. So while the grid reduces processing time of Excel-based reports, for example, it cannot handle ongoing and increasing data feeds from multiple sources, and convert them to multiple reports, for multiple users.

Another challenge computing grids cannot resolve is that enterprises can never know in advance exactly the amount of compute resources that will be needed, for how many processes, and for how long - nor is it financially practical to provision an infrastructure on an ongoing basis to handle maximum needs that arise only at peak times. If the system needs to scale even more, to handle more applications or ever-increasing data flows - the cost factor becomes even more critical.

Hence, while the private cloud solves the resource allocation problem better than the traditional grid, and less expensively, it does not provide the additional capabilities required by analytics applications, including: Managing more data, in less time; Running more applications, on fewer resources (increased utilization); Shrinking/growing the required infrastructure in accordance with actual needs (dynamic scaling); Enabling any single application to run on multiple resources (multi-tenancy); Simplifying provisioning, operations, and management; Maintaining reliability (redundancy), performance levels, and such business concerns as security, compliance, etc. All at no increased cost, or even actually reducing current cost levels (reduced TCO).

To summarize: Cloud computing can help reduce the cost of hardware, but current systems are not geared to handle complex applications. Compute grids enable processing complex calculations, but are expensive and cannot scale to actual needs, or handle the growing data challenge. So, most financial enterprises still face technology barriers to achieving the promised benefits of the private cloud, especially in regard to analytics applications.

In-Memory Application Platform = Cloud-Enabled Application Platform

Imagine that you could put all your data - terabytes of it - in-memory. What if you could manage your resources to have exactly the resources you need at any given time? What if you could run any application - no matter how complex - on multiple resources, so that your processing is sped up and your investment in IT infrastructure drops? What would that enable you to accomplish in your data center?

It would enable you to run any type of analysis, no matter how resource-intensive. It would enable you to run as many of these processes as needed, and to deliver real-time reports to as many people as need them within the organization. It would enable cost-effectively expanding your arsenal of business-critical processes at all levels of the enterprise.

GigaSpaces Technologies is a pioneer in developing in-memory infrastructure designed from the ground up to handle high-volume, high-performance, high-availability applications, where large data sets are a key application characteristic, and where data integrity cannot be compromised. Exactly the requirements for running financial services analytics applications. Because GigaSpaces has full integration with various cloud management mechanisms, the GigaSpaces eXtreme Application Platform (XAP) is uniquely geared to leverage a private cloud infrastructure for these memory- and data-consuming applications.

GigaSpaces provides an end-to-end solution for the entire set of requirements - functional, scalability, provisioning, and load-balancing - in a single platform, that: Runs real-time, event-driven analytics in-memory, close to the data; Uses standard programming models, ensuring interoperability, enabling re-use of existing assets, and shortening time-to-service; Has built-in multi-tenancy support. Multiple analytics processes can run on a shared infrastructure, with strict isolation, as if each were running on a dedicated resource.

The result is that you no longer need to just imagine running your calculations on your enterprise's private cloud. Now, financial services enterprises have a solution for real-time, complex analytics applications, with an absolute need for data integrity - all while saving on IT resources and investment, as compared to current solutions that offer far less performance.

In-Memory Application Platform + High-Capacity Hardware = Cloud-in-a-Box

Even if you recognize the benefits of running complex applications on the private cloud, you might be thinking, "That's all well and good, but I don't have a private cloud set up in my data center." You might be concerned with the cost and effort required to set it up, and worried about fitting all the parts together - the hosting, the hardware, and the platform on which your applications run.

To help enterprises overcome this obstacle, GigaSpaces has developed a joint solution with Cisco to provide a single, integrated system that unites computing, networking, storage access, and virtualization. The GigaSpaces platform can run directly on the Cisco Unified Computing System (UCS) hardware, with no intermediary hypervisor or operating system. What this provides is essentially a Private-Cloud-in-a-Box - just connect the "metal", and it's already a cloud. No set-up, configuration, or expensive investment of time or money is required.

In addition to the simplicity the GigaSpaces/Cisco solution brings to adopting and maintaining a private cloud, the integration also takes utilization to the next level. Cisco UCS offers a large number of cores and extensive memory capacity, designed to support parallel processing and offer the improved performance and increased efficiency of the next-gen data center. The multi-tenancy and memory capacity of the Cisco UCS, coupled with the complete GigaSpaces in-memory stack, enables processing all the terabytes of data and complex, mission-critical applications the enterprise might require.

The overall result: Massive reduction in data center footprint, and significant improvement in performance and scale of computational workloads.

For more information: www.gigaspaces.com/cloud