Press "Enter" to skip to content

Objective 2.5 – Build Performance Requirements into the Logical Design

Knowledge

  • Understand what logical performance services are provided by VMware solutions.

Memory

    • Transparent Page Sharing (TPS) – most guest virtuals will be using the same Operating System and drivers etc, this means there will be allot of identical data in memory. What TPS does is detect what virtuals have the same data in memory and then only keep one copy of it linking all other virtuals to that copy, this can save a tremendous amount of RAM.
    • Ballooning – a service which uses VMware Tools installed on the guest OS, if ESXi requires memory and a machine can spare it, a Driver in VMware Tools will balloon and take memory away from the guest OS causing it to page, then tells the Host it can now use the memory its taken.
    • Compression – ESXi will compress memory pages in physical RAM, while there is a small performance hit its not as noticeable as swapping to disk due to the speed of Physical RAM.
    • Caching – Allows the use of SSD drives to act as a Cache to allow for swapping to not take as big a hit as it would swapping to spinning disks.
    • Swapping – Swapping out RAM to Disk this is the last resort and if this is happening then its time for more Physical RAM.

Disk

    • Storage IO Control (SIOC) – This is a cluster wide contol of disk resources, its main goal is to prevent any single VM from  congesting a datastore(s). This enforces shares like are available with CPU and Memory onto Datastores as well.

Network

    • Network IO Control (NetIOC) – Enables basically QoS for virtual machines by prioritising traffic or partitioning bandwidth.
  • Identify and differentiate infrastructure qualities (Availability, Manageability, Performance, Recoverability, Security)
    • Availability – ability to access resources when needed, this is achieved using redundancy usually defined in the SLA as up time.
    • Manageability – is it Easy to deploy, Easy to administer, Easy to upgrade.
    • Performance – Measured usually by Throughput, latency or transaction time.
    • Recoverability – the ability to return the system to a working state after a failure.
    • Security – Correct people have access, firewalls need to be opened etc
  • List the key performance indicators for resource utilization.

KPI’s for performance would be Processor, Memory, Disk and Network

Skills and Abilities

  • Analyze current performance, identify and address gaps when building the logical design.

Current-State Analysis to collect the inventory of both Physical and Virtual, Storage, Networks and Applications.

Tools to be used would be:

  • VMware Capacity Planner
  • If available existing monitoring systems like SCOM, vCOPs
  • Guest OS tools like windows Perfmon or linux TOP
  • Using a conceptual design, create a logical design that meets performance requirements.

Lets use a example scenario here:

Client A wants a new testing environment, in this testing environment there will be 3 different areas needing resources R&D, HR and Apps Support. R&D needs be be guaranteed 50% of the resources and HR and Apps Support share the other 50% evenly.

Conceptually it would look like one big box and 2 smaller boxes(Conceptual Design is what the Client has put forward as they view it)

Logically we would picture it like this:

sample logical

  • Identify performance-related functional requirements based on given non-functional requirements and service dependencies.

Lets say for example a non-functional requirement was a monetary limit on what can be used on storage, this would limit the amount of disks and type of disks (SSD, SAS 15k 10k) which could limit the IOPS amount available.

  • Define capacity management practices and create a capacity plan.

From THIS Document:

A key goal for Capacity management is to provide value to the business; this can be done to great effect using VMware and consolidation.  Using consolidation it is possible to reduce the number of physical servers and to utilize those reduced physical servers at a more optimal rate.  We have completed a number of consultancy projects for our clients to assist in the consolidation of their infrastructure and certainly the key to this is in the planning stage.  In order to undertake this exercise a modeling tool is used.  Within this context there are two types of modeling tool.

 For a consolidation exercise we would look to use a tool such as VMware Capacity Planner that will analyze the current utilization of your server estate and determine your optimum consolidation requirements

 Once the consolidation exercise is complete the requirements of your modelling tool may change; obviously having our own modeling tool we are biased as to our choice, but whichever tool you choose it is important that it contains the following functionality:

    • The ability not only to model the resource consumption and the relative response time as well
    • If you are consolidating servers of differing OSs and platforms these need to be supported
    • To eliminate any issues with clock drift, etc, it is recommended that the data be captured via VCenter.

Going through any virtualization/consolidation exercise should generate some additional process metrics that can be used to demonstrate the effectiveness of the process and the utilization of VMware.  Within Business Capacity Management we are looking to provide metrics to demonstrate our value to the business and whether we are meeting their requirements.  The sort of metrics we could look at producing are:

    • % reduction in physical estate
    • % reduction in power consumption (useful as documented evidence in obtaining any environmental standard e.g. ISO14001)
    • % cost reduction in maintenance contract
    • Incorporate scalability requirements into the logical design.
    • Determine performance component of SLAs and service level management processes.
  • Incorporate scalability requirements into the logical design.

When doing the Current-State Analysis you should have a base line on how much capacity is required by the use of tools such as VMware Capacity Planner, you can then use tools such as vCOPS previously Capacity IQ to do a what if scenario to show how many more hosts may be required to scale out 10% etc

  • Determine performance component of SLAs and service level management processes.

From the previous documentation I linked

As we know, the response time of a service/application is a key performance metric and it is used to determine the end user experience of using a service that conventional metrics like CPU utilization, memory consumption, etc won’t necessarily show.

Data relating to the response time of a service can be difficult to obtain, but is critical to determining how a service is performing.  This sort of data will also become more important when we look at consolidating servers into a VMware environment as the resource limits that are applied when viewing the hardware as a whole will have a greater impact.

Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Anti SPAM BOT Question * Time limit is exhausted. Please reload CAPTCHA.