FlexPod Datacenter with Citrix XenDesktop and NVIDIA GRID

FlexPod Datacenter with Citrix and NVIDIA enhances virtual desktop infrastructure to encompass real-time 3D graphics capabilities. This new solution, FlexPod Datacenter with Citrix XenDesktop and NVIDIA GRID brings together best-in-class components to overcome the challenges graphics-intensive visualization workloads.

 

Screen Shot 2015-03-03 at 1.23.37 PM

Posted in Uncategorized | Leave a comment

OpenStack vs AWS

Service OpenStack Project Name Amazon
Virtual Machines Compute Nova EC2
Virtual Block Devices Block Storage Cinder EBS
Object Storage Object Storage Swift S3
Virtual Networks Networking Neutron (formerly Quantum) VPC
App Orchestration Orchestration Heat CloudFormation
App Monitoring Measurement Ceilometer CloudWatch
Tenant Auth / Namespace Identity Keystone IAM
Boot Images Image Service Glance AMI
GUI Dashboard Horizon Management Console
Posted in Uncategorized | Leave a comment

Citrix XenDesktop RAM Cache – An awesome way to save your IOPS

We have been fighting storage IO battle since day one. There are many flash technologies on the storage vendor side. Now Citrix Provisioning Service has an awesome and easy to use feature called RAM Cache on the hypervisor RAM that can reduce  storage IO by 90%. RAM cache is great for PVS non-persistent desktops. For MCS or any persistent desktops, RAM cache is not ready yet.

I tested RAM Cache last year. The result was unbelievably good. I shared at NetApp Insight Vegas and Berlin. In January, I attended Dan Feller‘s session “Technical Update for XenApp and XenDesktop – 2015 Edition” at Citrix Summit. Now it is time to promote RAM Cache.

How to use RAM Cache? At PVS console, select vDisk properties and choose Cache type as “Cache in device RAM with overflow on hard disk”. This will use hypervisor RAM first and then use hard disk. Choose the RAM size based on OS type.

RAM sizing consideration

  • 256 MB for Win7 32bit
  • 512 MB for Win7 64bit
  • 2GB-4GB for XenApp VM

More detail on sizing RAM cache, check  Size Matters: PVS RAM Cache Overflow Sizing.  Also as a best practice, defragment the vDisk before deploying the image and after major changes.

ram cache conf

Now let me share the test setup and result:

Physical hardware

Cisco UCS Server CPU:  20 CPUs X 2.799 Ghz   Intel CPU E5-2680 v2@ 2.80GHz

Server RAM:   48 GB

Hypervisor: vSphere 5.5

 Storage:

 NetApp FAS3240 with 2 shelves HDD disks

 VM Configuration

Win7 VM:

VM CPU: 1vCPU

VM RAM: 2GB

VM OS: Windows 7

vDisk: 20GB

VM write cache: 6GB

RAM CACHE: 256MB

Load generation tool

LoginVSI 4.1

 Test procedure: 

I ran 4 tests to find out if RAM cache is useful and what is the optimized RAM Cache size compared to Citrix’s recommendation.

1. Baseline without RAM cache

2. With RAM cache size as 256MB

3. With RAM Cache size as 512MB

4. With RAM Cache size as 1GB

Test result: I use NetApp perfstat to capture the write cache NFS volume total IOPS.

ram cache testing

The blue line is the result without RAM Cache. The IO saving is tremendous.

What is the lesson learns?

Increase your write cache when you use RAM Cache

Storage is 4k based and RAM is 64MB. “If transitioning an environment from Cache on HDD to Cache in RAM with overflow to disk and the RAM buffer is not increased from the 64MB default setting, allocate twice as much space to the write cache as a rule of thumb.”

Different performance on different hypervisor 

Reduce storage IO. The data is from Dan Feller’s deck at Citrix Summit.

– By 99% with XenDesktop

– By 80% with XenApp on Hyper-V

– By 24% with XenApp on vSphere

Defragment your vDisk to save your write cache size

“Defragmenting the vDisk resulted in write cache savings of up to 30% or more during testing.”

 

 

Posted in storage, virtual desktop | Leave a comment

Hope I can meet you at Vegas!

Hope you are as excited as we are for Insight 2014 Las Vegas! I will have 6 NetApp break sessions and a XenDesktop on Hyper-V lab. This will keep me busy.  Also you can find me at virtualization booth.  Hope I can meet you in Vegas!

MB-1-2025 – Citrix ShareFile Data Mobility for Enterprise
Date/Time: Tuesday, 10/28/14, from 13:30 to 14:30

MB-1-2025 – Citrix ShareFile Data Mobility for Enterprise
Date/Time: Wednesday, 10/29/14, from 14:45 to 15:45

VI-3-2132-TT – Citrix, VMware, and All-Flash FAS: A Reference Architecture Deep Dive
Date/Time: Monday, 10/27/14, from 13:00 to 15:15

VI-2-2024 – VDI Design, Architecture, and Best Practices for Citrix and VMware
Date/Time: Tuesday, 10/28/14, from 16:15 to 17:15

VI-3-2132-TT – Citrix, VMware, and All-Flash FAS: A Reference Architecture Deep Dive
Date/Time: Tuesday, 10/28/14, from 11:15 to 13:30

VI-2-2024 – VDI Design, Architecture, and Best Practices for Citrix and VMware
Date/Time: Wednesday, 10/29/14, from 16:15 to 17:15

VI-3-2132-TT – Citrix, VMware, and All-Flash FAS: A Reference Architecture Deep Dive
Date/Time: Thursday, 10/30/14, from 11:00 to 13:15

 

Screenshot 2014-10-24 10.55.10

Posted in Uncategorized | Leave a comment

Citrix XenDesktop Design Guide is published

Design guide for Citrix XenDesktop on NetApp Storage is published today. The Citrix XenDesktop 7.5 system, which now incorporates both traditional, hosted Windows 7 or Windows 8 virtual desktops, hosted applications, and hosted shared Windows Server 2008 R2 or Windows Server 2012 R2 server desktops (formerly delivered by Citrix XenApp) provides unparalleled scale and management simplicity while extending the Citrix HDX FlexCast models to mobile devices.

NetApp clustered Data ONTAP—with its key capabilities such as nondisruptive operations, unified storage, multiprotocol architecture, secure multi-tenancy, storage efficiency, read and write performance, and cost efficient data protection—is ideal for cost-effectively designing and deploying end-to-end storage solutions for desktop virtualization based on a single or mix of various XenDesktop FlexCast models.

This technical report provides key storage design and architecture best practices for deploying a mix of FlexCast technologies on a NetApp clustered Data ONTAP storage array. It also includes performance testing results and solution scalability guidelines that prove that NetApp clustered Data ONTAP can cost effectively scale to thousands of desktops without adding complexity.

Screenshot 2014-10-23 12.51.31

Posted in Uncategorized | Leave a comment

Big Hammer for virtual desktop

As a solution architect in a storage company, I am often asked by customers and sales engineers how many IOPS per desktop. My answer is always it depends. Virtual desktop IOPS is a myth. It varies by application, the stage of the workload, provisioning methods, and hypervisors. VDI IOPS is like different size nails, some are small like 2-3 IOPS per desktop, some are big like 20-30 IOPS. The easiest way is to have a big hammer to be able to hit nails in all size.  NetApp finally revealed all flash FAS ( AFF), the big hammer.

big-hammer

 

NetApp offers All-Flash  with advanced data management delivering high performance, advanced data management, and low cost per desktop. The NetApp All-Flash FAS solution shares the same unified storage architecture, Data ONTAP OS, management interface, rich data services and advanced feature sets as the Hybrid-FAS solution. This unique combination of All-flash media with Data ONTAP delivers the consistent ultra-low latency and high IOPS of all-flash storage, with the industry-leading clustered Data ONTAP OS. In addition, it offers proven enterprise availability, reliability and scalability; storage efficiency proven in thousands of VDI deployments; unified storage with multiprotocol access; advanced data services; and operational agility through tight application integrations.

We recommend to use AFF when the IOPS is > 20IOPS per desktop or ultra low latency  like < 1ms is required.

Check our new reference architecture on All flash FAS with VMWare View. AFF with XenDesktop is coming in a couple of months. Yes, I am doing the work.  🙂

Screen Shot 2014-06-24 at 11.04.46 PM

Posted in Uncategorized | Leave a comment