Users of a PaaS environment should not be directly concerned about how to scale their instance. "Elastic capacity" is one of the most appealing aspects of a PaaS solution, and a key driver for cloud computing in general.It is important to remember that organizationsdeveloping a PaaS solution internally have slightly different concerns than providers of a public PaaS solution.
For example, a public PaaS provider is providing a general SLA with customers that is measured in response time and availability. Most applications hosted in a public PaaS are there because a company is either unable or unwilling to devote their own data center capacity to hosting it. As such, it is probably a safe assumption that these applications will have a smaller potential set of users and perhaps lower capacity requirements. This assumption allows a public PaaS provider to more optimally use their own resources since many of the applications are mostly idle (of course they must be prepared for the chance that one app may suddenly require bursts of capacity).
In a private PaaS environment, the capacity demands of the applications may not vary as widely due to the fact that an organization's business applications (particularly internally-focused ones) are used by many common users with consistent use patterns. Therefore a private PaaS may not require such autonomous "elastic" capacity abilities that a public PaaS must provide. Scaling, therefore, is more likely an on-demand affair.
The key to effective on-demand scaling is accurate utilization metrics. The PaaS operations team must be able to quickly monitor the platform usage in terms of memory and cpu consumption to ensure the hardware is meeting the needs of the application. But consider that there may be several ways for the private PaaS provider to scale a Tenant's application:
Increasing EJB/JDBC pool sizes
Adding managed servers
Adding HTTP servers
Adding additional cache servers
Remember that unlike a public PaaS provider who has perhaps very little insight into the Tenant's applications, a private PaaS provider has far more tools at their disposal to scale up without having to resort to booting another virtual or physical server.
It should also be noted that scaling down should also be a part of the PaaS solution. In either public or private PaaS environments the key reason for scale down is resource reclamation. In a private PaaS environment this is critical for effective server utilization and cost reduction. For a public PaaS, scaling down is most likely the complementary behavior after a temporary scale up event (e.g. an e-commerce site just after the holidays). For a private PaaS, scaling down is most likely the result of analyzing usage statistics about an application that indicate its business value does not justify fully dedicated resources.