Selecting a Service Platform - The Practice of Cloud System Administration - page 50

Forexample,aserviceprovidermayofferfourzones:U.S.EastCoast,U.S.WestCoast,
Western Europe, and Eastern Europe. Each zone is built and managed to have limited de-
pendencies on the others. At a minimum, customers should locate a service in one zone
with plans for failover in another zone. A more sophisticated plan would be to have the
service run in each zone with load balancing between all locations, automatically shifting
traffic away from any zone that is down. We will cover this in more detail in Chapter 6 .
Such geographic diversity also permits customers to better manage the latency of their
service. Information takes time to travel, so it is generally faster to provide service to
someone from a nearby datacenter. For example, a service may be architected such that a
user's data is stored in one zone with a backup kept in one other zone. Users from New
York would have their data stored in the U.S. East Coast zone, with backup copies stored
intheWesternEuropezone.Duringanoutage,theuserisservedfromthebackupzone;the
service would not be as fast in such a case, but at least the data would be accessible.
IaaS providers have expanded beyond offering just simple machines and networks.
Some provide a variety of storage options, including relational (SQL) and non-relational
(NoSQL or key/value) databases, high-speed storage options, and cold storage (bulk data
storage that is inexpensive but has latency on the order of hours or days). More advanced
networkingoptionsincludevirtualprivatenetwork(VPN)-accessibleprivatenetworksand
loadbalancingservices.Manyprovidebothlocalloadbalancingandgloballoadbalancing,
as will be described in Chapter 4 . Some provide elastic scaling services, which automatic-
ally allocate and configure additional machines on demand as capacity is needed.
ProvidersthatofferbothIaaSandPaaSoftenblurthelinebetweenthetwobyproviding
high-level managed services that are available to both.
3.1.2 Platform as a Service
PaaS enables you to run your applications from a vendor-provided framework. These ser-
vices offer you a high level of value, as they manage all aspects of the infrastructure, even
much of the application stack. They offer very elastic scaling services, handling additional
loadwithout anyinputrequiredfromyou.Generally youarenotevenaware ofthespecific
resources dedicated to your application.
For example, in Google AppEngine, you upload application-level software and Google
takescareoftherest.Theframework(platform)automatically providesloadbalancingand
scaling. The more active your users are, the more machines Google allocates to your ap-
plication. Internally the system is managing bandwidth, CPU allocations, and even authen-
tication. Your application and hundreds of others might be sharing the same machine or
your application may require the resources of hundreds of dedicated machines. You do not
have to manage such decisions except to limit resource use to control costs.