Horizon Agent for Windows 1.5.1 was released this week. This is a complete rewrite of the agent and from a ThinApp point of view a big release.
Guest Blogpost by Chas Setchell…
Centralized management of VDI vs. physical computers saves on time and resources however depending on which kind of VDI deployment you have you could be realizing even more savings. Ideally the most efficient deployment type is Non-persistent / Stateless VDI, this is when you have one or many golden master VM’s that are then shared by a VDI pool. When your VM’s are Non-persistent / Stateless the user gets a fresh VM at login and at logout the VM gets recycled.
In a Non-persistent / Stateless VMware Horizon View deployment with Composer the admin creates a base image and performs a VM snapshot that is used as the starting point for the VM’s. When patch Tuesday rolls around and the admin will boot the golden master apply the patch and take a snapshot, the admin then needs to perform a recompose that will update all existing VM’s, and depending on the urgency of the update you might need to have users log off. This process will can take 1 or more minutes per VM.
What if you could take an active VMware Horizon View session and inject new applications, update existing applications including system and application patches in real time with no reboot, zero install, zero configuration, and no streaming.
Using CloudVolumes you get exactly that, let’s take a look at how this works in a VMware Horizon View environment. Let me introduce you to Appstacks. Appstacks are volumes that contain a single Application or Multiple Applications or anything you can install in a VM when it is in capture mode that enables the creation of a Appstack including dependencies, services, roles and even settings. A single AppStack can be shared across thousands of VM’s saving on storage and management.
How does it work? The CloudVolumes agent is installed in your Golden Master, when a user is login the Appstack’s get attached to the VM and any applications located in that Appstack are immediately made available to that user. When a user is already logged in to a VM when the administrator assigns a new Appstack CloudVolumes will attach the Appstack to the running VM and the agent will detect and inject the Application into the VM and will be ready use within seconds.
Using CloudVolumes Writable Volumes also enables some other exciting features for Non-persistent / Stateless VMware Horizon View environments. The CloudVolumes Writable Volumes will capture the changes made in the session providing Persona / User Profile Persistence across the Non-persistent / Stateless VDI pool.
This also enables the ability for User Installed Applications (UIA) or Admin Installed Applications (AIA).
A good example could be a User Installed Plug-in as the user changes VM’s the plugin will stay in the CloudVolumes Writable Volume. When a user logs out all of his changes are retained and again instantly applied to any pooled VM he logs back into.
There are many benefits to using CloudVolumes in a VMware Horizon View environment:
• Minimize the need for Department based Golden Master Images.
• Deploy Applications and Updates in real-time.
• Reduce Disk space by using only a single copy of the Applications and Updates across VMs.
• Writable Volume enables User Installed Applications for Non-persistent / Stateless.
• Change Golden Master without Reinstall of Core Applications.
• Change Golden Master without Impacting Users Writable Volume.
• CloudVolumes works with all existing tools and Application Virtualization including profile mgmt.
• Appstacks support even difficult applications with DCOM, Drivers and Services and even custom apps.
PernixData first beta test
From the time we heard about PernixData and watched PernixData CTO Satyam Vaghani’s videos form Storage Field Day http://pernixdata.com/SFD3 we were excited to get our hands on the PernixData Flash Virtualization Platform (FVP) product and give it a test.
What is PernixData and why should you care?
PernixData is a inline storage IO acceleration software that leverages flash storage on your VMware ESX host.
OK so you might be saying to yourself ok I have flash on my VMware ESX host why would I need a piece of software like PerixData FVP.
For Statefull / Persistent VM’s you typically want them running on some type of SAN that provides raid and enables you to use all the enterprise features from the VMware vSphere such as live migration. You also want to reduce the risk of data loss and having data on a single flash device could put you in that situation.
So back to why PernixData FVP, if you want to leverage your existing SAN that uses SAS or SATA disk and get higher performance without the risk of data loss that is where PernixData FVP comes in. PernixData FVP will leverage local flash across multiple VMware ESX servers increase performance and to prevent any data loss that might happen with local flash in a single system.
The fun stuff testing PernixData FVP
We tested PernixData FVP in our lab in the Cloud hosted at Baremetalcloud.com
The requirement is a VMware ESX compatible system with a flash drive.
If your system has only a single flash drive you will need to either boot and install VMware ESX from SAN, ISCSI, USB key or add a second drive. PernixData FVP needs the entire flash disk not just a vmfs volume.
- Build and install ESX on a compatible system “Do not install ESX on your Flash Drive”
- Enable SSH on ESX host and the PernixData FVP VIB
- Install the PernixData FVP server side component on a Windows 2k8 we used an existing SQL 2k12 database server.
- Register the PernixData FVP plugin in vCenter.
- Once you click getting started you will be taken though a guided setup.
- First you create a flash cluster and add the ESX hosts with the PernixData FVP VIB and Flash.
- Then you will add the associated VMware datastores that you want to accelerate.
- VM’s are now accelerated by PernixData FVP, yes it is that easy.
Quick VDI Boot Storm with 15 VMWare View VM’s with PernixData FVP
We wanted to get a little taste of how PernixData FVP would work with VDI specifically VMware View.
Since we were limited by what was available to us in the cloud we made use of what we could get.
Our testing is more functional than performance.
Our ESX host has 16gb of memory with 1 quad core CPU so not the most ideal VDI host.
1 256gb SSD flash device, 1 146gb SAS drive for OS, 1 iscsi LUN for Datastore.
We used VMware View to provision 15 VM’s once all VM’s were ready to use we did a simple boot storm by restarting all of the 15 running VM’s.
Let’s take a look at what the PernixData FVP performance tab shows
Frist up is VM IOPS this is the total number of IOPS generated by the VM’s in this case a peak of about 1800 in orange hitting the local flash, in purple the ISCSI Datastore about 200 IOPS resulting in a savings of 1600 IOPS on the backend.
VM Latency you can see that the flash is providing almost zero latency but the backend SAN Datastore has a latency peak of 200.
VM Throughput we noticed a peak of around 58000KBps to the flash and only 6000 to the SAN Datastore.
Hit Rate & Eviction Rate this is how much is hitting flash vs. going to SAN and you can see PernixData FVP is Caching a high percentage of the activity during this boot storm.
As you can see by the performance charts PernixData FVP can improve performance of VM’s by reducing latency and reduce the load on your backend storage.
We look forward to the general release of PernixData FVP and are excited to see what the PernixData team comes up with next.
written by Chas Setchell – chassdesk(at)gmail.com