Planning your migration project for deploying a virtualized & consolidated environment

1)
Identify and appoint your organization’s business & IT key players and stakeholders who will affect the migration project. All must be engaged in robust project assessment and analysis and be consulted for inputs to carefully prepare and account for all their requirements for the migration planning.

2)
Make initial assessment study to evaluate application readiness and tools validation, determine database and source code dependencies, and identify potential concerns and risk involved.

3)
Evaluate the portability of your custom-made and ISV applications to the new z/Linux-on-system z environment. This is a critical step in your consolidation project.

4)
Specify clearly and agree on migration project success criteria.

5)
Prepare to implement a pilot proof of concept (PPOC) to demonstrate the technology and success criteria, identify problem areas and evaluate risks. Select an application best suited for PPOC and that can benefit greatly from z/Linux on System z platform.

6)
Subsequent to successful PPOC completion, work out a detailed migration roadmap and base implementation plan.

7)
Work out carefully thought off detailed migration planning checklists. You should include checklists for planning HW, SW, tools, network, training and applications.

8)
Determine the System z resources (HW, SW, network, storage, power, floor space, etc …) required to consolidate your distributed workloads. This is a critical step you must make before moving ahead with the consolidation of your production environment. IBM provides a consolidation study using a specialized tool (RACE tool - the IBM Rehosting Applications from Competitive Environments).

9)
Secure executive sponsorship and business commitments. Allocate required resource, provide for needed z/Linux-z/VM skill development and manage expectation with a negotiated service level agreement (SLA).

10)
Implement the migration according to plan with robust execution to produce a z/Linux-on-System z environment mature enough for testing and validation.

11)
Conduct vigorous and critical testing of the new environment and resolve all performance and operational issues and be certain that all expectations and success criteria are met before committing to move on and deploy to production. Get the required signoff and approval for migrating to production.

Why virtualizing and consolidating with z/VM-on-System z

Virtualization provides a technology to consolidate the workload of many distributed physical servers onto a much smaller number of physical servers and hardware devices. Such consolidation can be optimized to maximize resource utilization and simplify the IT infrastructure and management, and to reduce the costs and budget required to license the software, and manage, house, power, and cool the data center.

IBM’s System z® with z/VM® in combination with Linux offers most powerful, maximum capability and industry leading virtualization technology. Here is a summary of the key characteristics of the z/VM® virtualization that make it superior to other virtualization solutions available in the industry. It is crucial for you to be aware of, fully appreciate and evaluate these characteristics in order to determine which virtualization solution is more appropriate for your IT data center:

Unparalleled Sharing Capabilities

Over many decades, system z and IBM mainframes have unceasingly acquired unparalleled resource sharing capabilities. These share-all capacities of the z platform offered unique benefits to z/Linux guest servers running on the z hardware:

In z/VM® virtualization all resources on the System z can be shared between all z/Linux servers:

Virtualization is implemented by a hypervisor which is a software code that runs on the hardware of a single computer system. The hypervisor creates an environment in which many guest operating systems can run concurrently and each isolated in a discrete virtual machine. Each virtual machine seems to the operating system running inside as if it were itself a real computer system equipped with its components including CPU, memory, and I/O. Applications running inside the virtual machine are not aware that the physical hardware has been virtualized. When the guest operating systems make accesses to these virtual components, they are intercepted by the hypervisor and translated and mapped into accesses to real physical components. This powerful technology creates the capacity to efficiently share the various resources of the computer system between multiple guest operating systems. This sharing capacity results in greatly increased and efficient total utilization of the system resources.

You can prioritize your z/Linux servers in allocating the shared resources. z/VM can reallocate all hardware resources to all Linux servers and provide optimal capacities to the guest servers that need more resources. More mainframe hardware capacities can be added to z/VM to accommodate the growth in z/Linux workloads.

Superior capacity utilization and workload balancing can be achieved across the entire virtualized environment as results of this full resource sharing capability.

z/VM virtualization enables efficient sharing of all I/O paths by all z/Linux servers. The I/O requests, as in z/OS, are off loaded to the mainframe SAP processors that are dedicated to IO processing. This hardware assist for I/O reduces the CPU load on the IFL engines processing the z/Linux workloads and deliver greater performance. Virtualization on the mainframe delivers I/O performance and throughput that are unmatched in the industry.

z/VM employs OSA (Open Systems Adapters) to connect to the physical external network and provides virtual networking (LAN) for Linux servers using VSWITCH functions. The z/VM hypervisor uses the VSWITCH capability to map virtual network interfaces of the z/Linux virtual machines into real physical Ethernet adapters on the real LAN in order to connect directly to OSA. Extra OSA adapters can be added without having to bring down z/VM and the z/Linux servers. The hypervisor also supports failover mechanism between physical adapters installed on the mainframe. This simplifies the routing of the network traffic and greatly reduces the overhead and cost.

z/VM can also provide for Linux workloads very fast connection to DB2 data using in-memory Hipersocket connections to z/OS (internal TCP/IP network buses between LPARs).

z/VM also provides for shared mini disk storage. Read-only file system such as WAS executable binaries can be placed in a shared mini disk so that large number of z/Linux servers can share accessing it. This improves cost because one has only a single copy of the code to maintain. In z/VM you can also cache to memory appropriate, less frequently written, mini disks to reduce IO traffic.

The hypervisor in z/VM does direct hardware support virtualization. In contrast, VMware’s hypervisor must translate, trap and emulate privileged commands. This degrades VMware’s throughput and scalability.

In contrast, in the VMware virtualization the resource sharing is very constrained:

The distributed physical servers are grouped at the physical level to form clusters. A group of Linux servers run on an assigned physical hardware cluster. In VMware the CPU and memory, network interface and internal storage routes assigned to a guest Linux server must be provided by the same physical server in the cluster. Moreover, the hardware resources in one cluster cannot be assigned to Linux servers defined to other clusters.

VMware cannot reallocate the non utilized hardware resources on one cluster to the Linux servers running on an over utilized distributed cluster. In order to obtain more hardware resources Linux server need to be moved to underutilized cluster. This Linux server movement adds a substantial performance and administration overhead.

The IO sharing in VMware is severely limited to only the IO paths that connect the physical server, supporting the guest Linux server, to the storage devices. All I/O requests must be dispatched and run on the CPU and are serviced by the CPU’s of the same supporting physical server which could create a performance bottle neck for IO intensive Linux workloads and limit the number of new Linux servers that can be provisioned within the cluster.

VMotion function in VMware can be exploited to achieve some workload balancing within the cluster. It allows Linux servers to be moved within the cluster from a constrained physical server to a physical server with free capacity. However, moving Linux server adds an overhead to an already resource-constrained cluster.

To install additional network adapters the physical server must be powered off and all Linux servers moved to other available physical servers. This would add overhead and might require extra spare hardware resources.

z/VM sharing capabilities including virtual networking, in-memory mini disk caching and execute-in-place file systems mechanism, enable collaborative environment for its co-existing z/Linux virtual machines to work and interact with more synergy and parallelism.This contributed greatly to z/VM’s capacity to virtualize and support such large number of virtual machines on a single system z IFL engine. Applications can be parallelized to run on multiple z/Linux virtual machines to boost throughput and maximize availability. You can use z/VM to build not only isolated z/Linux environments but also collaborative machines that might be required to virtualize the entire infrastructure of the data center.

Superior Consolidation Platform

z/VM in system z offers the most efficient, utmost mature and ultimate industry-grade server consolidation platform:

z/VM provides much more efficient virtualization technology with enormously greater hypervisor scalability and superior virtual server (CPU/IFL, memory & IO) scalability. One can consolidate hundreds or even a thousand of discrete distributed servers into a single mainframe. This greatly advantages the data center:

allows very large scale consolidation (high density of z/Linux servers in a mega-consolidation)

delivers massive savings on software licensing and IT support and administration staff

extensive power and cooling, networking and floor space savings

exceptionally cost effective and achievable disaster recovery

could generate a huge cut in overall TCO and a great ROI compared to virtualization on distributed and commodity hardware.

z/VM offers highly developed built-in systems management tools that can be applied to greatly reduce data center’s maintenance and support cost. z/VM also provides for accounting and chargeback models.

Superior system z Capacity on Demand (CoD) capabilities to dynamically and quickly add and allocate extra physical capacities (CPU/IFL & memory) to existing z/VM infrastructure. In contrast, expanding physical capacities in VMware is a disruptive process because it requires adding other physical servers to the present IT environment. Provisioning new servers in VMware distributed IT environment is also inefficient and time consuming due to lack of a true CoD.

System z virtualization offers Virtual Machine Resource Manager to manage resources for applications and workloads running on z/Linux servers contained in a z/VM LPAR.

Presents efficient platform to easily, rapidly and cost effectively provision new guest machines for applications’ development, build or test.

System z LPAR, the most robust and highly advanced partitioning capabilities, allow users to deploy multiple images of z/VM on a single mainframe. Each z/VM copy can own large numbers of z/Linux servers. In contrast, standard x86 servers in large part can host only a single image of VMware which largely constrains the attainable virtual machine density.

Wide range of software products covering middleware, applications and tools are becoming available and certified to run on z/Linux. This has opened great opportunities for migrating workloads from distributed servers and consolidating on z/Linux-on-z/VM System z platform.

If the data lives on the mainframe, then there is a very compelling reason to have the servers running on the mainframe as well and connect to the data through super fast
in-memory links.

The Reliability of the Hypervisor Layer

z/VM on system z provides a first-rate unmatched reliability for the virtualized infrastructure:

Virtualization technology has been created and more greatly evolved in the mainframe than in other platforms.

z/VM has evolved and advanced over more than 4 decades to become the most mature virtualization technology in the industry. During all this time z/VM has been used by IBM to test future versions of IBM’s operating systems.

Mainframe’s architecture has been designed early on for self-virtualization and therefore greatly advantaged z/VM.

In contrast, VMware runs on x86’s architecture that has not been designed for virtualization and it will make it more difficult to virtualize. VMware ESX’s track record does not measure up to z/VM’s.

z/VM’s hypervisor, that controls the resource allocation for many guest operating systems, enables the virtualization of not only processor and memory resources but also of I/O and networks.

z/VM greatly leverages the superior System z’s capabilities and highest hardware reliability and its advanced IO capacity for the advantage of its guest operating systems and applications. System z’ hardware reliability is quantum leaps beyond the reliability level you usually obtain in the x86 sphere.

Leveraging a co-residency with z/OS®

Linux-on- z/VM co-residency with the industry most powerful z/OS operating system allows z/Linux workloads to leverage greatly from z/OS world:

Share some of z/OS’s quality-of-service (QoS) and first-rate performance renowned strengths.

Access and connect to z/OS’s DB2 and workloads through high-speed Hypersockets Internal TCP/IP links. This provides enormous performance benefits, super security and network simplification to the co-resident z/Linux severs.

Gain continuous application availability by leveraging DB2 data-sharing technology deployed in z/OS environment running in the same box or in a system z parallel sysplex cluster. z/Linux servers can connect to data-sharing enabled BD2 for 24/7 availability and superior DB2 capacity and scalability.