The primary cause of extended server outages is administrator error. A troubling aspect of this situation is that currently, most server administration is performed manually and is getting continually more complex, leading to more room for error. The term "server management" has been bandied around for years but is largely misused.
"Server management" refers to server monitoring and does not
include the active administrative tasks (configuration changes, proactive
maintenance) that most administrators continually execute against their stable
of servers. Currently, systems administrators use a "toolbox" of utilities for
maintenance and administration containing everything from low-end commercial
off-the-shelf tools (COTS) to homegrown tools to high-end frameworks. These
processes are typically highly manual, poorly (if at all) documented, and not
shared across the enterprise (a condition we project will remain the norm
through 2003). This limits scalability of administrator-to-server ratios and can
cause them to degrade (with a commensurate increase in associated costs).
Although server monitoring has reached a level of reasonable maturity,
automated server management is just now emerging as a necessity. This is driven
by new application architectures employing on larger numbers of smaller servers
(horizontal scaling) to provide distributed power and high availability. Common
"server farm" configurations (both consolidations and new infrastructure
build-out), as well as emerging "blade" servers, pose a problem to
administrators-- keeping these servers in sync without drastically increasing
staffing levels. Although hardware vendors currently provide administrative
technology for their servers, these tools are often single-server-centric and
single-vendor-centric, lacking the broad automation capability required.
Through 2002, numerous server management tools will hit the market, largely
from emerging vendors. By YE03, Tier 1 infrastructure and application management
(IAM) players (Computer
Software) will enter this market, primarily via acquisition.
Demand for cross-platform server management will initially be strongest in
the service provider segment, though service providers are slowing overall
capital spending (through 2002), and by 2003/04, enterprises will drive to adopt
automated server management as a discipline. Initial technologies are focused
around administration of patches and software configurations, complementing the
server administration tools the hardware vendors currently provide, but with a
focus across platforms and acting on multiple machines simultaneously.
By 2005, automated server administration will be required for organizations
that want to treat server and disk capacity as a utility (utility computing,
grid computing), with these server management technologies tracking what
resources are available, allocating them appropriately, tracking usage, and
Hardware and operating system vendors have identified server management as a
competitive differentiator and are investing. Current investments are around
stronger instrumentation (e.g., better data on how the system is running and how
an application is using those resources) and exposing more capability for
administrators (e.g., new parameters to tune). The added instrumentation will
mature first and be the strongest of the infrastructure provider's efforts
While infrastructure vendors' initial work targeting server administration is
underway, these efforts will remain immature through 2005 and be best for a
single-vendor environment through 2007. This will still leave the cross-platform
issue unsolved and will be insufficient for the application layer.