Solution configuration best practices | Dell Ready Bundle for SAP with Unity Storage | Dell Technologies Info Hub (2024)

Overview

This section describes the best practices that Dell EMC recommends for configuring the Ready Bundle for SAP Landscapes solution.

Server configuration

Dell EMC recommendsthe following bestpractices for configuring thePowerEdgeservers:

• Use the latest BIOS and firmware.
• Use the latest vendor-subscribed HBA drivers.
• Enable Intel hyperthreading.

Local area network

Dell EMC recommends the following best practices for configuringtheLAN:

• Use two 10 GbE S4048-ON network switches for end-to-end client traffic to prevent a single point of failure (SPOF).
• Use a minimum of two network cards to provide HA and preventa SPOF.In our testing, we used two Broadcom 10GbE cards to enable connectivity betweentheSAP host andtheexternal network.
• Use VLAN tagging to sustain switch failure and NIC teaming to provide multitier network HA.
• Use NIC teaming to provide HA at the NIC port levelandfor virtual and physical network connections.

Storage area network

Dell EMC recommends the following best practices for designingthe storage area network (SAN):

• Use redundant Connectrix 32 Gb/s-capable switches or directors (DS-6610B, DS-6620B, ED-DCX6) to prevent a SPOF.
• Include the optional Enterprise Bundle with Fabric Vision for the DS-6610B and DS-6620B switches (Enterprise Bundle is included with the ED-DCX6 directors).
• Use two FC HBA cards for HA.
• Enable HA by making crisscross cable connections between redundant FC HBA cards and redundant FC switches.
• Use name zoning of end devices (FC HBAs) and Unity ports while creating zones. Name zoning of end devices provides better flexibility because it does not tie them to specific physical ports on either the switch or the storage array.
• Ingeneral, for best availability and performance, Dell EMC recommends that you zone or mask each server HBA port to one or two Unity ports. Where possible, the Unity ports should be on different SPs.Where possible, avoid creating a bandwidth mismatch between a host and the SPsit accesses, such as a single 16 Gb/s FC host initiator accessing four 16 Gb/s FC SPs. In this type of configuration, the host mightsimultaneouslytransmitI/Orequests to all four storage ports and be overwhelmed by the response from the targets, leading tospreadingcongestion and a potential negative impact on overall SAN performance.
• Use the Brocadefabric performanceimpact (FPI)monitoringand Monitoring and Alerting Policy Suite (MAPS) to monitor the SAN for”slow drain”events. For more information, see theBrocade Monitoring and Alerting Policy Suite Configuration Guide.

Storage configuration

The Dell EMC Unity array is a virtually provisioned,flash-optimized storage systemthat isdesigned for ease of use. This section introduces specific configuration recommendationstoenableyou to obtaingood performance from your Unity array.

Themain designprinciplessupporting goodperformance are:

• Maximizeflash capacity
• Distribute workloads across available storage resources
• Simplify the configuration
• Design for resilience
• Maintainthelatest released Unity Operating Environment(OE)version

Flash drives

Dell EMC storage arrays use multiple types of drives with varying capacities and performance levels. The arrays try to match data with the optimal drive typebased on storage profiles. Typically, mission-critical data is stored on high-performance drives, while less-critical data is moved to high-capacity drives.

Unity all-flash technology supports high-capacity flash drives that provide very high IOPS per terabyte (IOPS/TB). By taking advantage of the increases in flash memory densities and economies of scalethatindustry-standard flash-drive technologyprovides, Unity all-flash technologyoffersa differentiated capability with respect to many all-flash alternatives.

Storageprovisioning

As a fully unified storage system, the Unity array offers both block and file provisioning in the same enclosure.The system provisionsdisksinto storage pools that can host both block and file dataandsupportsconnectivity for both block and file protocols.Dell EMC recommends the following best practices:

• Block connectivityUse iSCSI or FC to access LUNs, consistency groups, VMware datastores, and VMware VVols.
• File connectivityUse NASservers to hostfile systems that are accessed using Server Message Block (SMB) shares or Network File System (NFS) shares. NASservers can also host VMware NFS datastores.

Dynamic pools

UnityOE version 4.2 on Unity All Flash modelssupports dynamic pools. Dynamic storage pools apply RAID to groups of drive extents from drives within the pool and allow for greater flexibility in managing and expanding the pool. Dynamic pools must be all-flash pools and cannot be built with HDDs (SAS and NL-SAS).

Dynamic pools do not require hot spares. A dynamic pool automatically reservesspare space in the pool at a rate of 1 drive’s worth of spaceforevery 32 drives. If adrive failureoccurs, the data that was on the failed drive is rebuilt into the spare capacity on the other drives in the pool.

RAID specification

The Unity array applies RAID protection to the drives in a storage pool. Selecting the appropriate RAID type is an important decision for a successful implementation. Consider elements such as performance, capacity, and availability.

Dell EMC storage systems support RAID 1, RAID 5, and RAID 6. Each RAID type provides different performance, capacity, and protection levels, as follows:

• RAID 1—Provides data protection by mirroring data onto another physical disk, producing better performance with minimal or no performance impact in the event of a disk failure.
• RAID 5—Stripes data across disks in large stripe sizes. RAID 5 stores parity information across all disks so that data can be reconstructed, which protects against a single disk failure. With its high write penalty, RAID 5 is most appropriate in environments with mostly read I/O.
• RAID 6—Stripes data across disks in large stripe sizes. RAID 6 stores two sets of parity information across all disks so that data can be reconstructed. RAID 6 can accommodate the simultaneous failure of two disks without data loss.

Customers select the RAID type and disk type combination based on their requirements. In a Dell EMC FAST™ for Virtual Pools (FAST VP) environment, Dell EMC recommends RAID 5 for flash, RAID 1 for FC, and RAID 6 for SATA.Table 7shows how RAID levels compare in random read, random write, sequential read, and sequential write scenarios.

Table 7. Comparison of RAID levels

Unityglobal memory

Global memory is a crucial component in the Ready Bundle for SAP Landscapes solution architecture. All read and write operations are transferred to or from global memory. The processing of transfers between the host processor andSPs occursat much greater speeds than transfers involving physical drives. Unity OE uses complex, intelligent, statistical pre-fetch algorithms, which adjust to workload on the array by constantly monitoring, evaluating, and optimizing cache decisions.

Unity All Flash systems share a number of similarities in form factor and connectivity but scale in processing and memory capabilities from 48GB per SP to 256GB per SP.

Front-end connectivity

The Unity array provides multiple options for front-end connectivity through on-board ports directly on the disk processor enclosure (DPE) and through optional I/O modules. In general, Dell EMC recommends that you:

• Connect and configure all required front-end ports symmetrically across the two SPs to facilitate HA and continued connectivity in case of SP failure.
• For best performance, useall front-end ports that are installed in the system so that workload is spread across as many resources as possible.

For example, when configuring the four-port FC I/O module, zone different hosts to different ports so that all eight ports across the two SPs are used. Do not simply zone all hosts to the first port of each I/O module.

Fibre Channel

For FC, configure Dell EMC Unity CNA ports and I/O module ports with 8 Gb or 16 Gb small form-factor pluggable (SFP) transceivers. All FC ports can negotiate to lower speeds. Dell EMC recommends 16 Gb FC for the best performance.

Dell EMC recommends that you use single-initiator zoning when creating zone sets. For HA purposes, zone a single host initiator to at least one port from SP A and one port from SP B. For load balancing on a single SP, zone the host initiator to two portsfrom SP A and two ports from SP B. When zoning additional host initiators, where possible, zone them to different SP ports to spread the load across all available SP ports.

On hosts connected with FC, use multipathing software such as the Dell EMC PowerPathTMsolution. The PowerPath software coordinates with the Unity system to provide path redundancy and load balancing.

Setting I/O paths

FC zoning partitions the FC fabric into multiple subsets. A zone includes FC initiators and targets and enables communication between them. Consider the following when setting the I/O paths:

• Single initiator/multiple target zoning creates each FC zone with a single initiator (HBA port) and multiple targets (Unity front-end ports). Therefore, each HBA port requires a separate FC zone with Unity front-end ports. Create independent zones for each HBA installed in the host.
• WWN zoning contains only the host HBA port and the Unity front-end primary ports. In most cases, you do not have to include the Unity front-end reserve ports because they are not used for volume mapping. Dell EMC recommends creating zones by using a single initiator host port and multiple Unity ports.
• Multipathing specifies how many of the Dell EMC storage front-end ports that the system allows the volume to be mapped through. Dell EMC recommends that you set this option toOS Default.
• When using a mixed fabric (that is, a fabric containing two or more switches running different release levels of Fabric OS), use the switch with the latest Fabric OS level to perform zoning tasks.
• Perform zoning using the core switch rather than an edge switch.

Volume provisioning and mapping

Thinprovisioning enables you to increase capacity utilization bypresentingmore storage to a host than is physically consumed and by allocating storage only as needed from a shared virtual pool.

To obtain optimal performance,Dell EMC recommends the following best practices for creating thin LUNs:

• Place data and log files on separate LUNs.
• Leave room for data growth and avoid exceeding 80 percent capacity of the LUNs for database files.
• Carve LUNs for data files on SSD drives.
• Use the latest storage array firmware.
• Configure end-to-end multipathing for redundancy and throughput optimization.
• Ensure that I/O is distributedtoall front-end ports.

After creating the LUNs,create a server cluster for HA and map the volumes to the server cluster.Related initiators (HBAs) are grouped by host, while related devices are grouped by storage pool.

NAS servers

NAS servers host file systems on the Unity storage system. NASservers use virtual interfaces to enable host connectivity to SMB, NFS, and multiprotocol file systems, as well as VMware NFS datastores andVVols. Depending on what is enabled on the NASserver,you can createSMB file systems and NFS file systems separately or in a multiprotocol configuration. File systems and virtual interfaces are isolated to a single NASserver, allowing for multitenancy over multiple NASservers. NASservers are hosted onan SPand automatically fail over if the SP becomes faulted. Any associated file systems also fail over.

Unity filesystems

File systems are file-based storage resources that are hosted on NASservers. They are accessed through shares, which can be provisioned for SMB or NFS access, providing access for Windows or UNIX hosts. Depending on what protocol is selected for your file system, only shares of that type can be provisioned. For example,foran SMB file system, you can provision SMB shares, but not NFS shares. To provision both SMB and NFS shares from a file system, use a multiprotocol file system.

File systems can be extended and shrunk in size, and allocated space is automatically taken and reclaimed based on file system usage patterns. Finally, file systems can be configured with quotas to better regulate file system space usage. Tree quotas and user quotas are supported on file systems and can be used in combination.

Global file sytems

The distributed architectureof SAP NetWeaver systemsrequires that global file systems are physically shared from the ABAP SAP Central Services (ASCS) instance to all SAP NetWeaver Application Servers (AS) belonging to the same SAP systemusing the NFS protocol. Examples include:

• Global directories such as /sapmnt/SID and its subdirectories
• The global transport directory /usr/sap/trans

Twomethodsareused to physically share these global file systems:server-based NFSandstorage-based NFS.Unity file systems can provide storage-based NFS to satisfy the SAP global file system requirements.

Unity compression

To help reduce the total cost of ownership(TCO)and increase the efficiency of a Unity storage system,Dell EMC addedcompressionin Unity OE version 4.1, with initial support for”thin block” storage resources. Unity OE version 4.2 and laterversionsinclude support for”thin clones”and”thinfile” storage resources.

With compression, a smaller amount of space is required to store a dataset for compression-enabled storage resources. This space savings reduces the amount of physical storage that is required to store a dataset, which can yield cost savings. Space savings are also realized on the storage resources.

Snapshots and thin clones

Snapshots and thin clones inherit the compression setting of the source storage resource, which helps increase the space savings that they can provide.

Competing workloads

When adding workloads to the Unity system, consider the reported CPU utilization rates as well as the capacity and IOPS sizing. Brief spikes of high utilization are normal and expected on any Unity system.

In highly consolidated environments, databases and applications compete for storage resources. The Host I/O Limits quality of service (QoS) feature controls“noisy neighbors”and protectstheperformance of theproduction system.You can apply host I/O limits on LUNs, VMware vStorage,VMware NFS datastores, and their associated snapshotsto limit the number of IOPS serviced by the Unity system.

Use host I/O limits to limit incoming host activity on the basis of IOPS, bandwidth, or both. You can enforce limits on individual resources, or you can share a limit among a set of resources.

Host I/O limits can beeffectiveif an overload on the storage resources affects the performance of critical applications such as SAP production installations. Consider configuring host I/O limits on LUNs, datastores of non-SAP applications, or SAP nonproduction systems to limit the total IOPs or bandwidth, which might otherwise consume a large portion of the system’s resourcesand affect the performance of SAP production systems.

Data Domain configuration

Dell EMC recommends the following best practices forconfiguring yourDataDomain system:

• Use 10GbE instead of 1 GbE for the cabling.
• UseDDBoost for the backupandrestoreoperations.
• Try to keep the Data Domain system less than 85 percent full. Filesystemcleaning and other operationsarefaster and more efficient when the system has enough diskspace availableto performthem.
• Schedulefile system cleaningto run during times when active backups are not running, to avoid resource contention and extended cleaning times.
• If replicating toaData Domainsystemat another site, do notschedule the replication to overlapwith your active backup window.Both processes require substantial resources and less time if they run separately rather than concurrently.
• Never use encryption, multiplexing, precompression,or client-side deduplication from the client backup application. These featuresgreatly reduce the compression factor that is obtained on the Data Domain system.Perform theseactivitieson the Data Domain system only. Some backup applications enable these features by default, soensurethat none of them are enabled for your application.
• Avoid selecting the highest available values for configuration options (number of streams, throttling, replication,and so on). Usingmore moderate configurations often offersthe best overall performance.
• Configure multiple dedicated network connections between the Oracle RMAN server and the Data Domain system to maximize throughput and provide redundancy.
• Note that many factors determine capacity requirements, including the size of the database being protected, rate of change, type of backups being performed, and retention requirementsfor the backups. A knowledgeable Data Domain system engineer can perform a detailed analysis using tools and models that have been developed over thousands of real-world deployments.
• Usethe Oracle backup method ‘RMAN backup’toobtaina faster backup speed. The RMAN channel set followsthe “1 CPU with one channel” rule for the system.
• Restrict both NFS and CIFS sharesthatOracle RMANaccesses directlyto the servers that are running Oracle RMAN operations. Backup security to a Data Domain system is usually maintained by controlling access to the share points.

Running SAP on VMware

Dell EMC recommends that you adhere to the following VMware sizing rules and considerations when you run your SAP system on VMware.

• Enablehyper-threading.
• Disable power saving mode.
• Try to size the VM to fit within the NUMA node (nonwide).
• InstallVMwareTools andperform theconfigurationusingthe VMXNET3 network adapter.
• Spread the virtual disks across all four of the available virtual SCSI controllers to a VM to maximize performance.
• Spread the database files across multiple LUNs and LUN queues to maximize I/O performance.
• Separate database log and data files into different LUNs to minimize I/O contention.
• Use the Virtual Machine File System (VMFS) where possible to increase operational management efficiency.
• Use eager zeroed thick format for all virtual disks.
• Note that themaximum number of vCPUs in VMware is 128 for a single VM.
• Use vSphere HA for all SAP instances.
• InstallCentral Services instances (ASCS/SCS) in a standaloneVMand ensure thatthe VMisprotected with VMware FT,which supports up tofourvCPUs.
• Ensure that you use the latest SAP host agent.
• If you want site HA, configure two or more ESXi hosts in the vSphere cluster.
• Ensure that vSphere HA and DRS are active and that VM monitoring is activated.

Note:If an ESXi host fails, vSphere HA powers on the SAP VMs according to the VM restart priority. ASCS has the highest priority followed by the database, followed in turn by the application servers.

• Use affinity rules to ensurethatOracle databaseVMsare running on processor coresthat arefully licensed for Oracle in VMware environments. For more information, seeUnderstanding Oracle Certification, Support and Licensing for VMware Environmentson the VMware website.
• Ensure that the server resources exist to fail over to a single site. Configure admission control of vSphere HA for 50 percent CPU and 50 percent memory.
• Enable VM Component Protection (VMCP) to take advantage of new capabilities around All Paths Down (APD) and Permanent Device Loss (PDL).

To enable VMCP, check the Protect against Storage Connectivity Loss option. Then, in the Failure conditions and VM response list box under the vSphere HA Cluster setting, select Power off and restart VMs for:

• Response for Datastore with PDL
• Response for Datastore with APD

For more information, seeSAP on VMware Best Practices.

Figure 4shows our storage design layout for VMware VMs.

Figure 4. Storage design layout for VMwareVMs

SAP system architecture

Because SAP ERP is one of the most important transactional systems in a typical enterprise IT environment, the system architecturemustconsider both performance and availability.Dell EMC highly recommendsa distributed system architecture, where each of the main components resides on its own VM, as follows:

• ABAP central services (ASCS) instance―ASCS comprises a message server and an enqueuer server that are both SPOFs. Separating ASCS from an application server instance in a central system architecture can minimize the impact of other work processes. A lower chance of failure also provides for the highest level of protection by using VMware FT. The SAP shared file systems /sapmnt/<SID> and /usr/sap/<SID> can be stored on this instance and shared to all other SAP instances within the same system.
• Database instance―A dedicated database instance has full command of its VM resources and is isolated from any other possible threat to the stability of the database. Because the network traffic between the database instance and the application server instances is usually high and RAM state change within the VM is frequent, avoid using VMware FT to protect the database instance. Use operating system/database-specific clustering tools or vSphere HA to provide a higher level of protection, but do not use both together.
• Additional application server (AAS) instances―AAS is a scale-out architecture that performs most of the computational tasks when transactions and background jobs are executed. You can add AAS at any time for additional performance and availability. Login groups (T-Code: SMLG) usually manage access to provide flexibility and increase availability. If one application server instance fails, connected users lose connection and reconnect to other AAS instances, and transaction-in-flight is rolled back. Standard vSphere HA is sufficient to provide a quick restart from ESXi or operating system failures.