Creating a New Virtualization Management Paradigm With Search-Based Technology

Virtualization is transforming self-evident physical machines into multiple virtual machines (VMs), which can be cloned instantly at no perceived cost and moved seamlessly from one physical machine to another. While the power of virtualization is enticing, its management implications are daunting. A completely new management protocol is needed to match the dynamic nature of virtual environments and keep pace with their evolution as they move beyond the enterprise and into the cloud.

The most logical approach to the virtualization management challenge is to establish search engine technology as the foundation and main integration gateway of the management system architecture. In a traditional business environment, the architectural layer of a management system is usually occupied by a relational database management system (RDBMS). Queries for information are performed using Structured Query Language (SQL). In contrast, search technology supplements or replaces a RDBMS with multiple data stores consisting of Extensible Markup Language (XML) and other types of documents that are linked together to represent important data relationships. Instead of using SQL, queries for information are performed via simple natural language and keyword oriented searches that traverse structured, semi-structured and unstructured data.

Search technology can quickly provide full visibility into the evolution, behavior and state of VMs. Moreover, it makes it easier to manage user facing applications and virtual operating systems along with the underlying virtual hardware, hypervisor and physical hardware as a cohesive service. The universally recognized search user interface (UI) with its advanced visualization capabilities can deliver a real-time status of an entire virtual / physical infrastructure or any subset thereof. With the addition of a historical data store, it can track all evolution and movement of resources and generate structured insight from unstructured, rapidly changing data.

The virtualization management challenge

Historically, applications have been physically bound to the underlying hardware and operating system (OS). However, in the virtualized world, applications and services are abstracted into a loosely coupled arrangement that promotes service levels and efficiency via dynamic movement across the underlying infrastructure. This increased complexity is exacerbated with the addition of hypervisor and guest OS layers in the application stack. As a result, the amount of management data, the rate at which it changes and the distribution of this data up and down the stack and across physical machines is overwhelming. Fundamentally, the virtualization management challenge is rooted in an information management problem, one that traditional management solutions are not architected to solve.

Since the virtualization management challenge is radically different from anything IT has ever experienced before, it's important to understand the underlying issues in order to appreciate how search technology can be applied to solve them.

New user experience

While the power of queries is obvious, the real potential of search-based management for virtual environments lies in the user experience and engagement model, which is quite different from what traditional management tools provide. Consider the following:

  • Intuitive User Interface (UI): The intuitive interface that search provides is a great equalizer, and promises to make all administrators more effective - regardless of experience or tenure.
  • Interactive Experience: Search technology facilitates a rich interactive experience with the user who enters a question in the search bar, then evaluates the answers that are served up as search results. The question can be refined and asked again, or a totally different question can be asked. This cycle can be repeated until the desired result is obtained. Even when the questions are tough, search offers a simple way to ask complex questions. For the Admin, asking questions via natural language and regular expressions in a search bar trumps complex SQL queries. More questions can be asked, and more answers can be obtained - faster. At the same time, bad answers can be proactively filtered out via facets, keywords and search hint techniques. The interactive engagement model supports instant gratification by giving the Admin data relevant to the situation at hand so that he/she can act.
  • Time to Action: Data fuels analysis, and in the data gathering process Admins commonly open dozens of windows or management utilities in what is affectionately referred to as "console hell." Because analysis is a series of questions and lines of interrogation, the interactive nature of search is a natural fit - particularly when configuration, performance, state and event data is collected and indexed across all layers of the stack. In this case, a single question can be posed in a search bar by specifying multiple data type combinations, values and thresholds. And with the right data aggregation strategy, the search UI begs to become a single pane of glass that is used to drive analysis activities. This is not surprising given that the hardest and most time consuming part of administrative scenarios consists of evaluating the risks of actions, not actual action itself.
  • Dynamic vs. Static User Interface (UI): Traditional management tools force Admins into particular roles with constrained paths of exploration and analysis. Static UIs are presented in a standard tree-oriented format where the detail associated with hosts, VMs and so forth is limited and presented in a single, rigid format. It's a "one-way-fits-all" and "vender-knows-best" approach that doesn't scale to comprehend large, dynamic environments. In contrast, a search paradigm embraces freeform thinking and individual problem solving preferences. The dynamic UI returns rich information collections as familiar search result sets that are relevant to the current and previous lines of interrogation. More and more irrelevant data is filtered until the appropriate action is revealed, resulting in a very fluid and interactive Admin experience.
  • No boundaries: Most traditional management solutions are organized around data type with configuration management, performance management, monitoring and alerting (events), etc. The Admin is challenged to correlate information across these domains to support troubleshooting, provisioning, consolidation initiatives, etc. Conversely, a search management paradigm has the unique advantage of being cross discipline out of the box since search is inherently a data integration platform and knows no boundaries. It is easily leveraged to provide dynamic cross sections of unstructured data in ways that accelerate analysis and facilitate quicker management action.

Once in place, a search platform presents even more interesting opportunities. Dynamic data rendering leads to advanced visualization scenarios where Admins visually process large data collections or result sets that are graphically rendered. Visual cues, colors and patterns call out VMs, hosts, applications and services that are exhibiting interesting behaviors. All this can be managed within a constrained context that is defined by the Admin on the fly by leveraging filters and nested avenues of previous exploration.

When specialized search is combined with advanced visualization, the result is a highly interactive venue that promotes rapid analysis. But what is really exciting is the potential capabilities that can be built on top of search and how they can capture the virtualization genie and put him back in very manageable virtualized environments.