viders competing to deliver very similar
product in a highly price-competitive
environment is termed perfect competition by economists. Perfectly competitive markets, such as those for milk,
gasoline, airline seats, and cellphone
service, are characterized by a number
of supplier behaviors aimed at avoiding the downsides of perfect competition, including:
˲˲Artificially differentiating the
product through advertising rather
than unique product characteristics
˲ ˲ Obscuring pricing through the use
of additional or hidden fees and complex pricing methodologies
˲ ˲Controlling information about
the product through obfuscation of its
specifications
˲ ˲ Compromising product quality in
an effort to increase profits by cutting
corners in the value delivery system
˲ ˲ Locking customers into long-term
commitments, without delivering obvious benefits.
These factors, when applied to the
cloud computing market, result in a
product that does not meet the enterprise requirements for deterministic
behavior and predictable pricing. The
resulting price war potentially threatens the long-term viability of the cloud
vendors. Let’s take a closer look at how
perfect competition affects the cloud
computing market.
Variable performance. We frequently
see advertisements for cloud computing breaking through the previous
price floor for a virtual server instance.
It makes one wonder how cloud providers can do this and stay in business.
The answer is that they over commit
their computing resources and cut
corners on infrastructure. The result
is variable and unpredictable performance of the virtual infrastructure. 5
Many cloud providers are vague on
the specifics of the underlying hardware and software stack they use to deliver a virtual server to the end customer, which allows for overcommitment.
Techniques for overcommitting hardware include (but are not limited to):
˲ ˲Specify memory allocation and
leave CPU allocation unspecified, allowing total hardware memory to dictate the number of customers the hardware can support;
˲ ˲ Quote shared resource maximums
instead of private allocations;
the cloud
computing market
is becoming more
crowded with
large providers
entering the playing
field, each one of
which trying to
differentiate itself
from the already
established players.
˲ ˲ Offer a range of performance for a
particular instance, such as a range of
GHz; and
˲ ˲ Overallocate resources on a physical server, or “thin provisioning.” Commercial virtualization management
software such as VMWare or Virtuozzo
offer the ability to overallocate resources on the underlying hardware, resulting in reduced performance during
peak loads.
Like overcommitment, limiting
access to infrastructure resources or
choosing lower-priced, lower-perfor-mance (and potentially older) infrastructure is used by vendors to make
providing cloud computing at rock-bottom prices viable. We entered the
cloud provider business after discovering we could not guarantee enterprise-grade performance to our customers
by reselling other vendors’ cloud services due to their corner-cutting. Here
is a list of some of the strategies the author has seen over the years:
˲ ˲ Traffic shaping. A new client asked
us to move their existing VPDC to our
infrastructure. Shortly after initiating
the transfer, the data rate dropped from
approximately 10Mbit/sec to 1Mbit,
where it remained for the duration of
the transfer. This behavior speaks pretty strongly of traffic shaping. Because
the client based their downtime window for the data center move on the assumption that the connecting network
was gigabit Ethernet, they missed their
estimate by over a week.
˲ ˲ Using older gigabit or fast Ethernet
networking. An ISP that was selling an
Amazon-like cloud computing product
connected their servers to the Internet
using fast Ethernet switches. If a customer wanted faster connectivity, there
was an up charge per port.
˲ ˲ Recycling failed disk drives. A client
leased several servers in a private cloud
from a very large ISP. He had previously
experienced several drive failures with
this ISP, so he decided to check up on
the hardware by running smartctl to
assess the health of the drives. What he
found was shocking to him: the ‘new’
servers he had just received had disk
drives in them that were over three
years old! When he challenged the ISP,
he was told their policy was to replace a
drive only when it fails.
˲ ˲ Deploying older CPU technology. We
were asked to manage a client’s ap-