Some info on hosting services:
(From a Jupiter Communications Report 4/99 titled Hosting:
Build Management Expertise Internally; Outsource
Infrastructure Needs)
Question to site executives: Which criteria are the most important
(on a scale of 1 to 5) in choosing an external hosting provider?
Reliability 5
Scalability 4.4
Security 4.3
Customer care/ service 4.3
Reduction in number of hops 3.6
Service offerings 3.4
Price 3.3
Some other snips from the report:
Decreasing Congestion and Reducing Risk
Other major drivers of multiple geographical distribution of front-end
servers and back-end databases are the desire to reduce the risks of poor
site performance due to congestion at a single server cluster and the
elimination of single points of failure. Several Los Angeles-based
companies told Jupiter that they needed multiple, geographically dispersed
data centers because of earthquake concerns. Other site executives
expressed serious concerns about depending on the performance of a
single hosting company or network. These executives believe that the best
means of ensuring reliability is to spread the risk among multiple hosting
and network providers. An executive of one technology company summed
up the sentiment succinctly. ÒWe have been filing our S-1 follow-ups [for
public stock offerings], so I have been forced to really look at risk
assessments,Ó he said. ÒThe things we could get away with as a start-up
simply are not acceptable anymore. We canÕt go down, so we canÕt depend
on just one location or [hosting] vendor.Ó
----------
Evaluate vendor peering agreements
A chief determinant of the speed
of responses to page requests is the quality of the networks over which the
information must travel. These networks are established as a result of
peering agreements between a hosting provider and a phone company.
Data exchange between networks occurs at network access points (NAPs),
which may be either public or private. Congestion, especially at public
access points, can adversely affect the quality of vendor service. A hosting
provider can offer better service by maintaining high-quality peering
arrangementsÑpreferably privateÑat important exchange points.
Developers should sign a nondisclosure document with potential hosting
providers that allows them to technically evaluate the quality of their
peering agreements.
------------
Establish disaster-recovery/redundancy plans and eliminate single
points of failure
Earthquake. Flood. Fire. Such disasters are of concern
not only to insurance companies but also to mission-critical content
providers. Qualified hosting vendors generally offer nightly tape backups
to provide data recovery in the event of a mishap, and, for additional
service fees, offer it in closer intervals. Jupiter expects that a proliferation
of front-end Web servers and back-end databases in several geographical
areas will create redundancies and eliminate single points of failure within
the Web hosting infrastructure, though sites should still maintain disaster-
recovery contingency plans.
-------------
Establish SLAs
A service level agreement (SLA) defines minimum
service quality and indicates the penalties that follow when the hosting
provider fails to meet that level of service. The SLA has become a
common method of auditing and managing third-party vendors offering
ongoing technology services, such as hosting, customer service, and
general application maintenance. New media companies often skip SLAs
because they have yet to gather the data necessary to define appropriate
levels of service or penalties and are often new to the development
process. Though Jupiter recommends that an SLA should accompany each
service contract, nearly one-third of companies Jupiter surveyed have not
established SLAs with their hosting providers.
Defining an SLA will force a company to plan a contingency strategy for
dealing with downtime and compel vendors of services to explicitly state
the steps that they will take when a failure occurs. After addressing a Web
site failure, a site manager must quickly communicate the cause of the
problem and establish a timetable for correcting it. Additionally, the
technical team should be responsible for contacting the marketing team
and drafting an information document to assist customer service support.
An SLA is not a guarantee against loss of revenue, but organizations
should be able to assign penalties that address their lost sales and the cost
of hosting. The actual financial impact on a site is probably significantly
higher than that, because a site is likely to lose consumers to a
competitorÕs site when they are unable to complete a transaction. As stated
earlier, Jupiter Consumer Survey data shows that 47 percent of consumers
indicated that they have left a preferred site and chosen a competitorÕs site
as an alternative at least once due to a site failure.
The best time for site managers to introduce an SLA into negotiations is
after the principals have agreed on initial pricing terms. A vendor may be
tempted to build extra expense into a contract to cover potential liabilities.
However, SLAs are not insurance, and sites should not be required to pay
extra for them. SLAs are meant to penalize a vendor in the case that it fails
to provide a contracted service. Some vendors may resist agreeing to a
contract that includes an SLA. This is a sign of an unreliable vendor.
If a vendor refuses to negotiate reasonable terms for an SLA,
a site should find another vendor. |
