Published on February 23, 2014
1. File Sharing and Failover Cluster This section defines the approach to file sharing and MS Failover Cluster including specific design decisions. 1.1 Summary Product development and <CUSTOMER> leadership have indicated that ongoing support costs are ultimately the most important factor in design decisions. Ease of implementation or recovery time should always be considered; however these items would have less of an impact on the design than ongoing support costs. The approach to file sharing design is to keep administrator support as simple as possible. 1.2 Design Decisions Decision Point Dell Decision Storage Configurations LUNs High speed - 6010XVS or Low speed - 6510E Low/6510: WITNESS: Witness/Quorum drive SQLDATA: SQL Data SQLLOGS: SQL Logs SOFTWARE: ISO, Software, and Citrix Administration USERDATA: User Data, Home Folders, Single Sign-on Store Justification This layout enables more effective capacity planning and protects production data from nonproduction growth. Windows User Profiles and Application Streaming Profiles require smaller, faster storage. High/6010XVS: APPSTREAM: Streaming Profiles PROFILES: User Profiles Launch Scripts Location (as needed) out of scope: SHAREA-???FS01LaunchScripts$ Hosted Application Data Location Streaming Application Profiles Location out of scope SHAREA-???-FS01AppStream$ Launch scripts are only required by some XenApp hosted applications. This might be deployed as part of an additional service; however this item is currently out of scope for the base offering. XenApp hosted applications are not included in this offering Streaming Application Profiles are stored on a central SAN share hosted by the cluster. These files are critical to proper functioning of streaming applications.
Decision Point Dell Decision Profile Directory Location SHAREA-???-FS01XDProfiles$ Home Directory Location Out of scope: SHAREA-???-FS01XAProfiles$ SHAREA-???-FS01users$ SSON Credential Store SHAREA-???-FS01CitrixSync$ License Server Justification For simplification of support, it is important for all users to have a common home drive path. This might be deployed as part of an additional service; however this item is currently out of scope for the base offering. <SERVER>-???-LIC01 1.3 Design Details There were 3 file storage options presented to delivery leadership. The 3 options are as follows: Horizontal Scaling – the original design included 1 file server per 500 seat pod. Each file server would be limited to 1 vCPU and 2TB of disk space. Hybrid Scaling – this option offered fewer file servers with larger disk sizes, either by adding more LUNs or by utilizing the MS iSCSI Initiator Centralized File Cluster – this option includes a single centralized file server utilizing MS iSCSI Initiator and MS Failover Cluster. Horizontal scaling results in 1 file server for each 500 seat pod. As an example, an 8,000 seat customer would start with 16 file servers. According to Dell Services operations groups, organizing profiles and user folders across 16 file servers will be an excessive administrative burden. Horizontal scaling also requires more Windows Server licenses and more resources to support the greater number of servers. Horizontal scaling would provide better server response times, especially as the number of file servers increases. Centralized file services offer a reduction in support costs and a reduction in the total number of supported systems; however this approach also creates a single point of failure. For this reason, MS Failover Cluster is included with this option. Hybrid scaling is a compromise between horizontal scaling and the centralized file cluster. With this option, there are fewer file servers that host more data per server. This would be accomplished by either adding more LUNs or by presenting larger storage LUNs directly to the MS iSCSI Initiator including with the guest operating system. A mount point solution could also be considered for this option. These facts were presented to leadership during a <CUSTOMER> design meeting. The decision methodology was to develop scoring categories like support costs, risk, performance, scaling, complexity, etc. These categories were scored and weighted according to leadership priorities. Long term administrative support costs were identified as the factor with the most weight. Horizontal scaling received a total score of 136, hybrid
scaling scored 158, and centralized file clustering scored 171. <CUSTOMER> leadership designated centralized file cluster as the preferred design. Virtualization of Failover Clusters - VMWare does not support MS Failover Clusters on virtual machines using iSCSI; however Microsoft does support this configuration for Premier customers if the storage is presented directly to the MS iSCSI Initiator. The complexity of the support policies, and considerations for DIY whitepapers prompted leadership to prefer a physical cluster until these support policies are changed. Failover Cluster Type - the witness (quorum) disk is a disk in the cluster storage that is designated to hold a copy of the cluster configuration database. (A witness disk is part of some, not all, quorum configurations) For this two-node cluster, the quorum configuration will be Node and Disk Majority, the default for a cluster with an even number of nodes. Node and Disk Majority means that the nodes and the witness disk each contain copies of the cluster configuration, and the cluster has quorum as long as a majority (two out of three) of these copies are available. “No majority node, quorum only” mode is not recommended because the quorum disk is a single point of failure. File System Type - For native disk support included in failover clustering, use basic disks, not dynamic disks. NTFS format is required for the quorum/witness disk. NTFS format is recommended for all other shared disks. GPT is best practice for Equallogic volumes and is required for volumes larger than 2TB. Volumes should not exceed 17TB without explicit design of the volume to include cluster size adjustments. LUN Organization – For ease of management, more effective capacity planning, and to prevent non-critical data from filling up more critical file volumes, Dell will provision separate LUNS for each of these activities: WITNESS: Witness/Quorum drive (Low/6510) SQLDATA: SQL Data (Low/6510) SQLLOGS: SQL Logs (Low/6510) SOFTWARE: ISO, Software, and Citrix Administration (Low/6510) APPSTREAM: Application Data and Streaming Profiles (High/6010XVS) PROFILES: User Profiles (High/6010XVS) USERDATA: User Data / Home Folders, Single Sign-on SSPR (Low/6510) LUN Sizes - for 500 seats are as follows: WITNESS: 1 GB SQLDATA: 44 GB including: XenApp Farm: 4 GB XenApp Logging: 4 GB XenDesktop Site: 20 GB XenDesktop Logging: 4 GB PVS Server Farm: 2 GB VCenter: 10 GB SQLLOGS: 10 GB SOFTWARE: 50 GB
APPSTREAM: 10 GB PROFILES 5 GB USERDATA: 2.5 TB XenApp Farm Data Store Database - approximately 100MB or every 250 servers and 50 published applications in the farm. The current <CUSTOMER> offering includes only application streaming to the virtual desktop and applications installed into the virtual desktop image. There will be only 2 servers in each data store and very few print drivers. For initial deployment and to allow for growth, 4 GB will be allocated for XenApp database. XenDesktop 5 Site Database - Citrix published number of 20,000 desktops requires 150 GB database. 5000 desktops would require approximately 40 GB. 500 users would require approximately 10GB. For initial deployment and to allow for growth, 20 GB will be allocated for XenDesktop database. vCenter Database - vCenter database will scale at a rate of 2-3 GB per 2000 seats but should not exceed 10 GB. For <CUSTOMER>, we will estimate 10GB for initial deployment. SQL Logs - Full Recovery model will be used. Total disk capacity is driven primarily by XenDesktop. In a 24 hour period, one desktop will consume 1.45 MB, 20,000 desktops will consume approximately 29 GB. Across all databases, <CUSTOMER> estimates approximately 2 GB per 500 seats. For initial allocation and to allow for growth, 20GB will be designated for SQL Logs. User Data – 5 GB per user or approximately 2.5 TB for 500 seats. Windows User Profiles - 5 MB per user or approximately 2.5 GB per 500 seat pod with local logons to virtual desktops only. Separate profiles are used for XenApp hosted applications, so this requirement would be doubled if XenApp is included for the deployment. For current <CUSTOMER> design, Windows user profiles will be stored on a smaller, faster volume and user data would be separated onto a larger, slower volume. For initial deployment and to allow for growth, 5 GB will be allocated for XenApp database. Application Streaming – Dell Services provided an average of 50 MB per application or approximately 1.5 GB for the first 30 applications. For initial deployment and to allow for growth, 10 GB will be allocated for Application Streaming Profiles. Provisioning Server (PVS) – Minimum size for a PVS SQL Database is 250 MB. For initial deployment and to allow for growth, 2 GB will be allocated for the PVS database. Launch Scripts – These files are critical to proper functioning of hosted applications and should be stored on fast storage. These are typically a small amount of data that can be collocated on the APPSTREAM volume. This type of configuration would be deployed as part of an additional XenApp service and is currently out of scope for the base offering. Network Applications – PeopleSoft Tools, for example, might be configured to be executed directly from a network share. Centralized application files stored on SMB shares are currently out of scope for <CUSTOMER>.
SQL Temp Database - A "temp" database should be placed on a partition with at least 1 GB of free disk space and set to grow automatically. Citrix recommends 4GB if the farm is large and includes multiple print drivers. This database can stay on the C: drive of the SQL server as this would be on a faster storage volume. The system drive should be configured with sufficient free space to accommodate the temp database. License Server – The license server resource utilization is minimal and Citrix License server is cluster aware. The license files are not a significant amount of data; however these files should be considered production data. These files can be collocated with SQL databases on the SQLDATA LUN. This approach has been field tested across several Dell Services hosted customers. Single Sign-on Credential Store – This component is out of scope for the base <CUSTOMER> offering; however it may be sold as a separate service. The credential store for SSPR only deployments can be stored on the USERDATA volume. Larger deployments require a design process and may require a dedicated volume. EqualLogic Host Integration Tools – provides host-based solutions that expand Equalogic capabilities at no extra cost. The kit improves overall storage access performance and provides the following additional features: EqualLogic Remote Setup Wizard - enables you to quickly attach servers to SAN storage. The Remote Setup Wizard can initialize an array and set up or expand a group from a Windows server, automatically configuring access controls on both the group and the server. VDS Provider - enables you to use Microsoft Virtual Disk Service (VDS) and Microsoft Storage Manager for SANs to create and manage volumes in a PS Series group. Auto-Snapshot Manager - enables you to implement Microsoft Volume Shadow Copy Service (VSS) backups. 1.4 Additional Resources NTFS Chkdsk Best Practices and Performance Windows XP Professional Resource Kit - Disk Management Dell Whitepaper: Sizing and Best Practices for Deploying Microsoft Exchange Server 2010 on VMware vSphere and Dell EqualLogic Storage Windows and GPT FAQ MS Technet Forum replies from Microsoft Disk Partition Alignment Best Practices for SQL Server Deploying Windows Server 2008 with PS Series SANs Deploying MS SQL2008 DSS using Dell EQL 10GbE iSCSI Storage Setting up the License Server on a Microsoft Cluster
XenDesktop 5 Database Sizing and Mirroring
File Design: Â· Procurement Â· Inventory Â· Distribution File Design Name of the file: Purchase requisition Number of field : 11 Primary key : ...
1. 1. File Sharing and Failover Cluster This section defines the approach to file sharing and MS Failover Cluster including specific design decisions.1.1 ...
Moving File Shares to IBM FileNet P8 SummaryDavid ChampeauIBM FileNet Certified Professional 2. ... Design; Devices & Hardware ... GuideIT Delivery Design ...
Mounting Windows SMB File Shares Using Cifs. by boxfileku. on Oct 16, 2015. Report Category: Documents
... (Additional file 1) ... Conceptual design of the guideline implementability tool ... ” for guidelines would be a good delivery system.
View Brian Murphy ’s professional ... - Complex and secure MPLS Design combined with Delivery Zone ... Leverage AD DFS for file services shares. Team ...
Table C-6 Application and Service Tools for Troubleshooting. ... they can use different hardware architecture and design ... On the File menu ...
View 368 Delivery & Design posts, presentations, experts, and more. Get the professional knowledge you need on LinkedIn. LinkedIn Home What is LinkedIn?