Dell DVS Reference Architecture for Windows Server 2012 (update)

This architecture is a comprehensive update to a previous RA I wrote a few months back on RDS VDI using Server 2012. Below are the pertinent updated content sections of the document or you can pull from its published location on dell.com (August 2012): LINK

Summary

The Dell Windows Server 2012 RDS solution provides a robust and scalable VDI platform for pooled, personal and Session host deployments. Using VDI-optimized hardware in a configuration that has been validated and proven by Dell DVS Engineering, you can deploy Microsoft based VDI that is both cost effective and high performing. Our layered architecture provides flexibility to maximize your infrastructure investment with the capability to expand and contract where necessary.

New content:

• Single server 50 user/ POC offering
• Improved densities for both RDSH and RDVH session sources
• RDS options and Hyper-V architecture detail
• HA offering
• Updated test results and analysis

Local Tier 1 – Solution Layers

Only a single high performance Force10 S55 48-port switch is required to get started in the Network layer. This switch will host all solution traffic consisting of 1Gb iSCSI and LAN sources. Additional switches can be added and stacked as required to provide High Availability for the Network layer.

The Compute layer consists of the server resources responsible for hosting the user sessions, whether shared via RDSH (formerly Terminal Services) or pooled via RDVH (see section 4.5.1 for a detailed explanation of each role). The RDVH role requires Hyper-V as well as hardware assisted virtualization so must be installed into the parent partition of the Hyper-V instance. The RDSH role is enabled within dedicated VMs on the same or dedicated hosts in the Compute layer.

Management components are dedicated to their own layer so as to not negatively impact the user sessions running in the Compute layer. This physical separation of resources provides clean, linear, and predictable scaling without the need to reconfigure or move resources within the solution as you grow.The Management layer will host all the RDS VMs necessary to support the infrastructure as well as a file server to host SMB shares for user Profile Disks or data.

The Storage layer is made up by the capacity dense and performance capable Equallogic 4100E iSCSI array. 12TB is provided in base form that can scale as high as 36TB to suit your capacity requirements. A second 4100E can be added to group the two arrays to provided greater capacity or performance.

 

Local Tier 1 – 50 User/ Pilot

For a very small deployment or pilot effort to familiarize yourself with the solution architecture, we offer a 50 user/ pilot solution. This architecture is non-distributed with all VDI and Management functions on a single host. If additional scaling is desired, you can grow into a larger distributed architecture seamlessly with no loss on your initial investment.

Local Tier 1 – Combined

As a logical entry point to the distributed RDS solution stack, a combined architecture is offered to host both the RD Virtualization Host (RDVH) and RD Session Host (RDSH) roles within the same physical Compute host while separating the Management layer. This will enable users requiring either shared RDP or pooled VDI sessions to be hosted on the same physical server. The value of this solution is a minimum infrastructure investment with maximum VDI flexibility easily tailored to shared and pooled user types. Horizontal scaling is achieved simply by adding additional Compute hosts.

Local Tier 1 – Base

In the base distributed architecture the RDVH or RDSH roles are assigned to a dedicated Compute host. This architecture can support either a single RDVH or RDSH Compute host or one of each. This solution provides maximum Compute host user density for each broker model and allows clean linear upward scaling. You’ll notice that the hardware spec is slightly different for the two Compute host types, giving additional RAM to the virtualization host. This of course can be adjusted to suit your specific needs.

Fully Expanded

The fully expanded architecture provides linear upward scale for both the RDVH and RDSH roles optimized for 600 pooled VDI sessions or over 1000 shared. See Appendix for test results. This solution supports up to 4 Compute hosts of any combination running either RDVH or RDSH roles to meet the needs of the enterprise.

High Availability

High availability (HA) is currently offered to protect all layers of the solution architecture. An additional ToR switch is added to the Network layer and stacked to provide redundancy, additional Compute and Mgmt hosts are added to their respective layers, and Hyper-V clustering is introduced in the Management layer.

Solution Density Summary

Design Scale	Management Hosts	Compute Hosts	RDSH Sessions	OR		RDVH Sessions	HA
50 User / Pilot	0	1	100		50		-
Combined	1	1	130 + 75				+ 1 Compute +1 Mgmt
Base	1	1	260		145		+ 1 Compute +1 Mgmt
Expanded	1	2	520		290		+ 1 Compute +1 Mgmt
Fully Expanded	1	4	1040		600		+ 1 Compute +1 Mgmt

RDS Options

Server 2012 RDS provides a number of VDI options to meet your needs, all within a single, simple, wizard-driven environment that is easy to set up and manage.

• Sessions, hosted by the RDSH role (formerly Terminal Services), provide easy access to a densely shared session environment. Each RDP-based session shares the total available server resources with all other sessions logged in concurrently on the server. This is the most cost effective option and a great place to start with Server 2012 RDS. An RDS CAL is required for each user or device accessing this type environment.
• Pooled VMs are the non-persistent user desktop VMs traditionally associated with VDI. Each user VM is assigned a dedicated slice of the host server’s resources to guarantee the performance of each desktop. The desktop VM is dedicated to a single user while in use then returned to the pool at logoff or reboot and reset to a pristine gold image state for the next user. Applications can be built into gold images or published via RemoteApps. A VDA license is required for each non-PC device accessing this type of environment.
• Personal VMs are persistent 1-to-1 desktop VMs assigned to a specific entitled user. All changes made by Personal VM users will persist through logoffs and reboots making this a truly personalized computing experience. A VDA license is required for each non-PC device accessing this type of environment.

Compute Server Infrastructure

The Compute host configuration varies slightly as to whether it will be hosting RDSH or RDVH roles, or both. The RDVH role must be enabled in the Hyper-V parent partition thus providing one RDVH role per Compute host if pooled or personal VMs are required. The RDSH role should be enabled in up to 4 VMs on a single Compute host to support up to 260 session-based users.

The requirements for RDSH VMs are outlined below. All application and non-OS related files should be installed in the 5GB data disk:

Role	vCPU	Startup RAM (GB)	Dynamic Memory			NIC	OS + Data vDisk (GB)	Tier 2 Volume (GB)
			Min|Max	Buffer	Weight
RD Session Host	8	16	512MB | 20GB	20%	Med	1	40 + 20	-

 

Management Server Infrastructure

The Management host configuration consists of VMs running in Hyper-V child partitions with the pertinent RDS roles enabled. No RDS roles need to be enabled in the root partition for Management hosts.

Management role requirements for the base solution are summarized below. Data disks should be used for role-specific application files/ data, logs, IIS web files, etc and should exist in the Management volume on the 4100E array. Please note that the Tier2 volume presented to the file server is designated as a pass-through disk (PTD).

Role	vCPU	Startup RAM (GB)	Dynamic Memory			NIC	OS + Data vDisk (GB)	Tier 2 Volume (GB)
			Min|Max	Buffer	Weight
RDCB + License Server	1	4	512MB|8GB	20%	Med	1	40 + 10	-
RDWA + RDG	1	4	512MB|8GB	20%	Med	1	40 + 10	-
File Server	1	4	512MB|8GB	20%	Med	1	40 + 10	2048 (PTD)
TOTALS	3	12				3	120 + 30	2048

 

Solution High Availability

Each layer in the solution architecture can be individually protected to prevent an extended service outage. The Network layer only requires an additional switch configured in a stack with the first.

Protecting the Compute layer for RDSH and RDVH is provided by adding an additional host to a collection, thus effectively increasing the hosting capacity of a given collection. Session requests will be fulfilled by all hosts in the collection and as a result, each will have reserve capacity to insure against a host failure. Care needs to be taken to ensure that user provisioning does not exceed the overflow capacity provided by the additional node. A simple fail-safe measure would be to ensure that the appropriate number of users entitled to connect to the environment be tightly controlled via Active Directory. In a failure scenario users working on a failed host would simply reconnect to a fresh session on a surviving Compute host.

To implement HA for the Management layer, we will also add an additional host but will add a few more layers of redundancy. The following will protect each of the critical infrastructure components in the solution:

• The Management hosts will be configured in a Hyper-V cluster (Node and Disk Majority).
• The storage volume that hosts the Management VMs will be upgraded to a Cluster Shared Volume (CSV).
• SQL Server will be added to the environment to support RD Connection Broker HA.
• Optionally SQL mirroring can be configured to further protect SQL.
• The RD Connection Broker will be configured for HA.

Volumes	Host	Size (GB)	RAID	Storage Array	Purpose	File System	CSV
Management	1	500	50	Tier 2	RDS VMs, File Server	NTFS	Yes
Management	2	500	50	Tier 2	RDS VMs, File Server	NTFS	Yes
SQL Data	2	100	50	Tier 2	SQL Data Disk	NTFS	Yes
SQL Logs	2	100	50	Tier 2	SQL Logs Disk	NTFS	Yes
SQL TempDB Data	2	5	50	Tier 2	SQL TempDB Data Disk	NTFS	Yes
SQL TempDB Logs	2	5	50	Tier 2	SQL TempDB Logs Disk	NTFS	Yes
SQL Witness	1	1	50	Tier 2	SQL Witness Disk	NTFS	Yes
Quorum 1	-	500MB	50	Tier 2	Hyper-V Cluster Quorum	NTFS	Yes
User Data	-	2048	50	Tier 2	File Server	NTFS	No
User Profiles	-	20	50	Tier 2	User profiles	NTFS	No
Templates/ ISO	-	200	50	Tier 2	ISO/ gold image storage (optional)	NTFS	Yes

 

Architecture Flow