Policy-based application management and virtualization: The (missing) key components of data center automation

by Ahmar Abbas

November 11, 2003 —

The ever-increasing complexity of the enterprise data center is challenging IT managers to discover creative ways to manage it, while also keeping costs down. At the same time, many executives worry that the current inflexibility of the IT infrastructure hinders corporate business strategy. Product development, product launches, channel development, and other critical business activities are delayed or even abandoned due to the difficulty and cost of modifying today's information technology infrastructure.

In response to these challenges, major IT vendors have outlined initiatives that help enterprises automate the data center and increase flexibility by allowing resource sharing. Examples include Sun's N1, Computer Associates' on-demand computing initiative, IBM's e-business on demand, HP's adaptive enterprise, and Veritas' building blocks for utility computing.

These vendor initiatives may bear unique monikers, but the underlying strategies are remarkably similar. In each instance, an evolutionary approach is offered that starts with simple data center automation and progresses to an environment in which all resources and applications are virtualized and contain high degrees of self-management capabilities.

To accelerate their ambitious strategies, the major vendors have complemented and expanded their technology and product portfolios through acquisitions. The tempo of such acquisitions by large systems vendors has increased in recent months, with deals totaling nearly $1 billion over the last 12 months. Recent acquisition activities include Sun's purchase of Center Run, which complements the company's earlier acquisitions of Terraspring and Pirus, IBM's purchase of ThinkDynamics, and Veritas' purchase of Jareva.

The recent wave of acquisitions has focused on automating the management, provisioning, and sharing of various data center hardware resources. These are undoubtedly important pursuits, but glaringly absent from all of the current offerings is the ability to dynamically allocate resources to applications based on pre-defined policies and real-time business needs - in other words, the on-demand capabilities of these on-demand strategies. On the evolutionary path of Sun's N1, on demand arrives in phase 3, referred to as policy-based automation; similarly, IBM considers this level 5, dubbed autonomic data center management; and at Computer Associates it is phase 4, otherwise known as business-driven infrastructure management.

As each of these major systems vendors gets closer to phasing in true on-demand capabilities, we are likely to see a new wave of acquisitions centered around policy-based application management and application virtualization. These critical elements already exist in a few startups but are most prominent in Ejasent, a 4-year old technology company based in Mountain View, CA.

Ejasent has developed a broad suite of products that provide just-in-time application availability through its policy-based application management. Central to this capability is the company's application virtualization technology, which provides an abstraction layer between applications and the underlying infrastructure. This allows applications to be deployed to pools rather than to specific servers, to be invoked or revoked in seconds, to be migrated from a maxed out 2-way server to a 4-way server while maintaining session state and user connections, all of which result in significantly higher utilization of computing resources and optimized balance between availability and resource allocation.

What distinguishes Ejasent's technology in particular is its support for complex, multi-tiered enterprise applications (e.g., databases or complex ERP applications). Ejasent has filed for more than a dozen patents for its virtualization, connection migration, and policy management technology within the company's rather unique UpScale product.

Ejasent also has a complimentary metering and chargeback product, MicroMeasure, that it claims is the only product on the market that monitors usage of the difficult to measure logical resources (e.g., application usage, transactions, i/o, throughput, etc.) in addition to the typically measured physical computing resources (e.g., CPU, memory, and storage).

Ejasent has developed a strategy to go to market solely through partnerships with large system vendors, which gives potential acquirers the ability to closely evaluate the company's technology and products. Given that Ejasent's technology seems to strategically fit into many of the large system vendors' near-term utility computing roadmaps, it wouldn't be at all surprising if Ejasent were acquired. Ejasent's management team, however, believes the company can successfully continue to build a standalone business.

Some other likely acquisition targets include: VMware, a Palo Alto, CA-based virtualization provider; Opsware, a Sunnyvale, CA-based data center automation software provider co-founded by Marc Andreessen; and Egenera, a Marlborough, MA-based company that provides a virtualization solution coupled with an increasingly popular blade server architecture. Given the expected size of the utility computing market and the current high level of focus by major platform and software players, it is likely that only a few (if any) of these companies will remain independent over the next couple of years.

With enterprise customers starting to look to comprehensive on-demand utility computing solutions to solve the data center complexity challenge, the major vendors will need to make good on the solutions they have outlined. Accordingly, there is a certain level of urgency among the vendors to bolster their product and technology portfolios - both to satisfy customer requirements and to gain a substantial competitive advantage. Interesting times lay ahead.

Grid Technology Partners