1. UNIMI: Provisioning Services Stream Process Architecture

2. Server and Desktop Provisioning
Create a virtual image of a server’s or Desktop’s workload: OS, applications and configuration
Virtualize
Store the virtual workload image on a network storage device
Store
Stream the workload image on-demand to virtual and bare-metal servers/Endpoints
Stream
Stream A
Virtualize
Store
…and Provisioning Server does just that. It changes the game for IT. It does this by streaming workloads on-demand rather than pushing full images onto each target system. There’s a lot of really cool and complex technology under the hood that I could spend an entire day talking about but really the model is quite simple, it’s virtualize, store, and stream.
You virtualize a system image by capturing it from a reference system.
That image is stored somewhere on the network
And then Provisioning Server streams it to any number of target devices.
We’re going to walk through a number of use cases, but in all of them, the model stays the same: Virtualize, store, and stream.

3. Provisioning Server For Desktops
Using Citrix Provisioning Server for Desktops OS-streaming functionality, two Virtual Disks are created – one configured with the Windows XP operating system and software and one with Windows Vista and software. The vDisks are stored as files on the network.

4. Provisioning Server For Desktops
LAN-connected “diskless” workstation. No software pre-loaded and the system is configured to boot from the network.

5. Provisioning Server For Desktops
The XP operating system and software vDisk is streamed to the desktop using Citrix Provisioning Server for Desktops OS-streaming functionality.
Workstation is powered on.

6. Provisioning Server For Desktops
Workstation image delivery and management are centralised; data security is assured because there is no persistent storage at the desktop.
PC processing power is harnessed for an uncompromised user experience.

7. Provisioning Server For Desktops
Workstation image delivery and management are centralised; data security is assured because there is no persistent storage at the desktop.
Provisioning the workstation with the Vista vDisk is quickly accomplished by rebooting.

8. PXE Booting a Provisioning Services Target
DHCP Server
IP Info
IP Info
IP Info
Target Device 1
Provisioning Services
Workloads
DHCP Discover
BOOT A
Target Device 2
TFTP
PXE
Network Storage B
Provisioning Services allows multiple end points, be it Virtual machines, or physical to boot off of a single image using the PXE Protocol. When a PXE Enabled Target boots it will send out a DHCP Discover packet, [CLICK]
Both the DHCP AND Citrix PXE Service will respond with offers, DHCP will send the IP information and the PXE Service will reply with the hostname/IP of the TFTP server and the bootstrap file name. [CLICK]
The target will then download the Bootstrap file from the TFTP Server [CLICK] and begin streaming the image.
The bootstrap can also be loaded by booting to a ISO file via the BootDevice manager.
Bootstrap
Bootstrap
Bootstrap
PXE Info
PXE Info
PXE Info
Target Device 3

9. Failover & Login Process
StreamServer
:vDisk3
PVS Target Device
Login Process
Streaming from
Request for vDisk3
This slide is going to show exactly what happens when a target device fails over.
The Target is up and streaming from
[CLICK]
becomes unresponsive.
Target refers to the bootstrap for other options.
The Target will try to contact the first server in the bootstrap. In this case
Target will go through the login process, authenticating with the MAC address and requesting vDisk assignment exactly as it would if it was booting for the first time.
In this example the Target is assigned vDisk3 streaming from Server
The Target connects to and requests vDisk3.
Failover is accomplished.
Bootstrap X

10. Load balancing with High Availability (cont)
PVS2
PVS1 X
If a PVS Server goes down
[CLICK]
all the targets connected to that server will failover.
Once the down PVS Server comes back up however, those targets will not automatically rebalance. They will stay connected to the server they are on until a rebalance command is sent. All newly booted Targets will stream to PVS1.

11. Where to store the cache file in a HA environment
Cache on local RAM
Cache on Server Disk
Cache on Local Disk
Another factor to consider when designing a High Availability solution is where to store the cache files.
There are three options to storing cache: Cache on Local Ram, Cache on Server, Disk, Cache on local Disk. We will discuss the pros and cons of each of them in regard to High Availability.
The main thing to keep in mind when choosing the cache location is that the cache needs to be available to both servers so in the event of a failover the Target Device can still access it’s cache file.

12. X Cache on Local RAM Shared Storage PVS1 PVS2 Cache vDisks
If caching on local RAM the amount of RAM allotted needs to be large enough to support the target device between reboots. If it is not it the Target Device will run out of cache space and eventually bluescreen.
In this slide the two Targets to the left are booted to PVS1 the two on the right are booted to PVS2
[CLICK]
If PVS1 experiences an outage the two targets on the left will failover to PVS2 and still have access to their cached information.
Pros:
Ram cache has very quick read/write times
Cons:
Ram is expensive and in some environments it is impractical to load the extra memory into the end point devices.
Cache
vDisks
SQL database

13. X Cache on Local Disk Shared Storage PVS1 PVS2 Cache vDisks
In this slide the two Targets to the left are booted to PVS1 the two on the right are booted to PVS2
[CLICK]
If PVS1 experiences an outage the two targets on the left will failover to PVS2 and still have access to their cached information
Hard Drive needs to be formatted with at least 1GB of free space. PVS will use the first Hard Drive that matches these requirements.
If multiple local hard drives are installed there is no way to configure PVS to cache on a certain drive.
Pros:
Most end point devices already have a local hard drive which would make local hard drive caching cost effective.
No extra shared storage space is needed for the cache file.
Less traffic on the network.
Less strain on the Provisioning Servers
Cons:
Some thin clients do not have local hard drives.
Cache
vDisks
SQL database

14. X Cache on Server Disk Shared Storage PVS1 PVS2 Cache vDisks
In this slide the two Targets to the left are booted to PVS1 the two on the right are booted to PVS2
[CLICK]
If PVS1 experiences an outage the two targets on the left will failover to PVS2 and still have access to their cached information
Pros:
Cache in one centralized location
Cons:
All reads and writes are sent through the provisioning server reducing performance.
Large Shared storage solution is required.
Larger strain on network since all of the cache reads and writes are going over the network.
Cache
vDisks
SQL database

15. Cache on Server Disk (incorrect configuration)
Shared Storage
PVS1
PVS2 X
In this slide the two Targets to the left are booted to PVS1 the two on the right are booted to PVS2
[CLICK]
If PVS1 experiences an outage the two targets on the left will failover to PVS2 but do not have access to the write cache. Eventually the Target Device will attempt to access critical information written to the cache file and will bluescreen.
Cache
vDisks
SQL database

16. Possible single point of failure
Shared Storage
PVS1
PVS2
DHCP
PXE
TFTP
High Available provides failover for the Streaming Service only. Single points of failure could still exist that will effect the employment.
Previously slides showed a high level very simple view of a provisioning server deployment, focusing on the target devices, Provisioning Servers and shared storage. Provisioning Server solution relies on other components as well. High Availability only provides failover for the streaming service. All other components need to be setup with failover separately.
In this section we are going to talk about these components and the options.
[CLICK]
These componets are
DHCP
PXE Server
TFTP
And the Network.
Switch
Cache
vDisks
SQL database

17. Possible single points of failure
TFTP
Network
Storage
Bootstrap Delivery
TFTP, Network, Storage and bootstrap delivery could be single points of failure if not configured properly.

18. TFTP Redundancy
DHCP Option 66 only allows for one TFTP server to be entered.
Using TFTP Clustered Servers or PXE Service instead of DHCP options will solve this issue.
This slide is going to talk about possible ways of setting up redundant TFTP.
If using DHCP and want fully available TFTP a load balancer such as (netscaler glsb) must be used to make the TFTP highly available.
This is not needed if instead of using DHCP, PXE is used since multiple PXE Services can run on the same network.

19. Network Redundancy Target Device PVS Server Target Device PVS Server
Switch
PVS Server
Target Device
This slide is going to talk about possible ways of setting up redundant networks:
PVS is only as redundant as the network it is running on.
NIC Teaming
Load Balancing Switches and routers with failover
Dual connections to backend storage with multiple paths.

20. Storage Redundancy Type of Storage Pros Cons CIFS on Windows Server
Easy to setup
Free with OS
Hard drives are cheap
Performance due to SMB traffic
Lack of scalability
HA requires two plus servers
CIFS on NAS Device
Device typically less expensive than SAN device
Built in redundancy for disks (depends on vendor/model)
Various vendors/models
Less expensive solutions equal single point of failure
Scalability can be an issue depending on vendor/model purchased
This slide is going to talk about possible ways of setting up redundant storage:
Clustered File Systems

21. Storage Redundancy (continued)
iSCSI to NAS Device
Moderate to setup
Device typically less expensive than SAN device
Built in redundancy for disks (depends on vendor/model)
Various vendors/models offer resiliency benefits
Less expensive solutions equal single point of failure
Scalability can be an issue depending on vendor/model purchased
Requires cluster file system when sharing LUN across multiple servers
iSCSI to SAN Device
Highly scalable
Built in resiliency
Build in features like storage replication and high performance disk arrays
Moderate to complex to setup
Typically more expensive than NAS devices
This slide is going to talk about bootstrap Delivery redundancy.
DHCP Clustering
Multiple PXE Servers

22. Storage Redundancy (continued)
Fiber to SAN Device
Highly scalable
Built in resiliency
Build in features like storage replication and high performance disk arrays
Complex to setup
Typically more expensive than NAS devices
Requires cluster file system when sharing LUN across multiple servers
This slide is going to talk about possible ways of setting up redundant storage:
Clustered File Systems

23. Bootstrap Delivery Redundancy
Solution
Pros
Cons
DHCP Clustering
Less traffic on the network
Moderate to setup
Multiple servers needed
More expensive
Multiple PXE Servers
Easy to setup
Less hardware needed.
More broadcast traffic on the network
This slide is going to talk about bootstrap Delivery redundancy.
DHCP Clustering
Multiple PXE Servers

25. Soluzione per UNIMI - Centralized HA per LAB
Target Devices
Provisioning Servers
Centralized HA stores the vDisks on shared storage:
Easier to manage, all vDisks are in one location only requiring one vDisk to be updated to make a change.
Requires shared storage.
More network traffic then distributed.
Target Devices
vDisks
Shared Storage

26. Soluzione per UNIMI completa
LAB 2
2 provisioning servers & 50 Target Devices
LAB 1 – 50 Target Devices
Target Device 1
LAB 3 – 2 provisioning servers & 10 Target Devices + 10 Remote Desktop Linux
Provisioning Services
Workloads A
Target Device 2
Network Storage B
Provisioning Services allows multiple target devices physical and virtual, to boot off of one streamed image saving greatly on management and storage costs.
Pxe Options can also be used. HA
Target Device 50

27. Proposta commerciale: LICENSING
- Lab1: 50 licenze XenDesktop Platinum per user
- Lab2: 40 licenze XenDesktop Platinum per user
- Lab3: 10 licenze XenDesktop Platinum per user

28. Proposta commerciale: PROVISIONING SERVERS (1)
Lab1: 50 users (licenze citrix + 1y di sw assurance) ,00 € (subscription ~ 60$ x lic)
2 Server reali con XenSever in HA ,00 €
2 Server virtuali con Provisioning Sever 5.6SP1
- Lab2: 40 users (licenze citrix + 1y di sw assurance) ,00 € (subscription ~ 60$ x lic)
- Lab3: 10 users (licenze citrix + 1y di sw assurance) ,00 € (subscription ~ 60$ x lic)
2 Server virtuali con XenDesktop 5

29. Proposta commerciale: PROVISIONING SERVERS (2)
6 x Server Reali:
DELL Server rack 1U PowerEdge R410
Processore: Processori 2x Intel Xeon E5506
Memoria: 24GB DDR3 1333MHz
Sistemi operativi - Opzioni di virtualizzazione: Citrix®  XenServer® 
HDD: 2x 146GB, SAS 3Gbps 15kRPM (hot plug)
Controller RAID: PERC H200A Controller di rete: 6x Broadcom NetXtreme II 1GbE
Alimentazione: Ridondante 500 W

30. Proposta commerciale: NETWORK STORAGE
NetApp FAS2040-R5
24 x 450 GB (15K)
8 (4+4) porte 1 Gb
4 (2+2) porte 8 Gb
Base Sw inclusi
Licenza CIFS-NFS
Garanzia 5y NBD
50.000,00€

31. Base pack comprende un set di funzionalità quali (Incluso) :
Snapshot™,
FlexVol®,
Thin Provisioning,
RAID-DP®,
FilerView®,
FlexShare® ,
iSCSI, FCP,
SyncMirror
NearStore®,
Deduplication,
System Manager,
Operations Manager,
HTTP, host utilities