Partition-based heap memory management in an application server

Abstract

Heap is an important shared resource in Java virtual machine. A problem with memory management in one component can affect the whole system and even result in crashing the virtual machine. In this paper, we propose an partition-based approach to manage heap in an application server. In our approach, the shared heap is divided into logical partitions, in which instances of application components and server components are allocated separately.

Supporting Reconfigurable Fault Tolerance on Application Servers

“…Fault isolation prevents other components from accessing the failed or recovering components. Similar techniques in operating systems [12] and in virtual machines [22] have already proven their effectiveness. Sandboxes provide fault confinement and isolation at middleware layer to ensure that: a) if an error in a component arrives at its sandbox's boundary and there aren't any components strongly depend on the component (explained in Section 4.1), the error cannot run across the boundary; and b) once an error is caught and the failed component is being recovered, incoming requests for the component would be blocked for fear of new failures.…”

Adaptive defenses for commodity software through virtual application partitioning

“…Sensitive memory objects are identified and used to build the appropriate code segments that can access them. A similar approach is also employed by Liu et al [24] to protect the Java virtual machine (JVM) against illicit memory operations that may corrupt it, causing it to crash. ABYSS [53] and AEGIS [46] both describe a secure chip architecture that enables some regions of an application to execute in a secure mode (encrypted), while the remaining ones executed in insecure mode (unencrypted), hence partitioning in this way their execution.…”