Enterprise End User Summit 2014

Compared to traditional hosting, cloud computing provides attractive economic incentives while simultaneously improving ease of management. But it also provides an additional layer of complexity, and many existing security policies may fail to take cloud-specific concerns into account.

This presentation will cover some of the fundamental differences between cloud deployments and bare-metal hosting. The role of the hypervisor in providing fundamental security will be discussed, along with the potential worst case scenarios of various breaches and mitigation strategies (existing and future) for reducing the risk.

The final topic will be new technologies that permit cloud deployments to be more secure than traditional hosting, giving admins greater trust in the state of their systems.

Financial trading and modeling, telecommunications, and traditional high performance computing applications have requirements that apparently preclude virtualization. Yet in each case there are compelling reasons to virtualize these applications, from economics to better automation.

Mike Day will discuss emerging virtualization technologies for Linux that show promise running High-performance and deterministic applications.

Mike will lay the foundation with Linux Containers (LXC) and KVM, using Linux memory, cpu, and scheduling configurations. He will then cover new technologies including Cloud Operating Systems such as OSv and Partitioning hypervisors such as Jailhouse. These new technologies omit many traditional features to optimize for size, performance, or both.

Mike will discuss the outlook for emerging technologies and how they may effect the broader Linux ecosystem.

Since mainframes traditionally run mission critical workloads, serviceability has always been of high importance to the platform. Customers expect the root cause of a problem to be found with the first set of debugging data collected. Reproducing the error is not welcome and must be avoided. With this in mind, IBM provided the s390dbf kernel tracing infrastructure when Linux was ported to the s390 architecture at the beginning of 2000. This was about eight years before Linux ftrace and tracepoints were integrated into the upstream Linux kernel. In some areas, tracepoints offer advantages, for example, per-CPU trace buffers or an event streaming capability. Yet, tracepoints do not offer all features that are needed for s390. This presentation reviews the two tracing features and shows what would have to be added to tracepoints to achieve the same functionality as s390dbf.

Over the past year, cgroup has been going through major restructuring to address various issues, some of them fundamental, in the existing cgroup design and implementation.  The unified hierarchy will effectively serve as the v2 interface of cgroup.  This session presents the design of and rationales for the unified hierarchy and its progress.

Larger enterprise systems need to minimize recovery times and provide enough information for error analysis with kernel dumps. As core counts and memory sizes have grown, crashdump function has struggled to scale and provide the information needed for analysis. This talk will cover recent practical improvements to crashdump technology for better performance and function on larger systems. It will also cover potential future improvements.