HPE 3PAR Performance Best Practices Guide

Getting the best performance and resilience from your HPE 3PAR storage array doesn’t have to be difficult. You must lay the proper foundation, follow best practices when building on that foundation, monitor your progress, and correct course as necessary.

This blog will show you ways to implement your 3PAR storage such that you have the best possible performance and resiliency in your environment.

Fibre Channel Ports – Laying the Foundation

One of the most important building blocks of the 3PAR foundation is balance, particularly in the host ports for the storage array. For maximum balance, performance, and resilience, you should have a documented and followed process for allocating Fibre Channel host ports.

A typical process is to have the odd-numbered ports in one fabric (Fabric A) and the even numbered ports in the other fabric (Fabric B). If you are using more than the typical two fabrics, the ports of the same pair of nodes with the same ID should be connected to the same fabric, for example 0:2:3 and 1:2:3 in Fabric A and 0:2:4 and 1:2:4 in Fabric B.

Other best practices for Fibre Channel ports are dependent on whether you are using Brocade or Cisco switches.

Brocade Switches:

The easiest way to ensure things are properly zoned on Brocade switches is by enabling Smart SAN and using the SSMC or 3PAR CLI to make the configuration.

Smart SAN reduces complexity and difficulty by completely automating the zoning task. Smart SAN also utilizes target driven peer zoning (TDPZ), allowing a single target to be zoned with multiple initiators automatically when provisioning SAN hosts to 3PAR ports. TDPZ provides the benefits of initiator-based zoning (single initiator/single target) with fewer zones and optimal use of switch resources.

Figure 1 shows a traditional single initiator/single target zoning scheme for a 3PAR host port being zoned to HBA ports for three SAN-attached hosts:

Figure 1 – Single Initiator/Single Target Configuration

Three zones are necessary to zone the three hosts to the 3PAR host port. When using peer zoning as in Figure 2, the zoning configuration is much simpler because only one zone is needed. When scaled out over a large infrastructure, this results in a dramatically reduced active zoneset size and much easier administration.

Figure 2 – Peer Zoning Configuration

Cisco Switches:

Cisco switches use a similar feature to peer zoning called Smart Zoning with close to feature parity to Brocade Peer Zoning. Cisco Multilayer Director Switch (MDS) isn’t as tightly integrated with 3PAR OS as Brocade, however, and can’t take advantage of automation provided by TDPZ, requiring the administrator to manually create the zone.

3PAR Fibre Channel Port Configuration Best Practices

Some general best practices for host port configuration are:

• Use Smart Zoning or TDPZ. If those aren’t available, use single initiator/single target zoning.
• If you are planning to use the port persistence feature of HPE 3PAR, be sure to zone accordingly
• For best performance and resilience, zone each host HBA to at least two physical switches
• Hosts should be zoned to HBA node pairs

Auditing to make sure that your zoning configurations are within best practices can be a time-consuming and manual task. IntelliMagic Vision for SAN provides automated reports that point out places where your zoning doesn’t fall within best practices. This allows you to proactively address these issues before you find out the hard way when a key component fails, and an outage occurs.

Figure 3 shows one of these reports, showing that a host masking view doesn’t have the necessary supporting zoning in place. This report shows the connectivity from the host through the 3PAR host adapters down to the LUNs the 3PAR is serving through the host.

Figure 3 – Orphaned Ports

The diagram shows a host directly connected to the HPE 3PAR. If it were connected via SAN switches they would appear in the diagram between the host and the HPE 3PAR adapters. The red dots on the host indicate that there are HBA ports configured in the masking view that are orphaned because they lack connectivity to the HPE 3PAR array.

Block Storage

HPE have recently changed their stance for how block storage should be configured. In the past, RAID 5 was favored over other RAID types. In more recent versions of 3PAR OS, however, the suggestion is to use RAID 6 for maximum availability.

There are some exceptions, however. For Fast Class Common Provisioning Groups, (CPGs), while RAID 6 should be the default, if there is a very high write ratio, you should consider using a RAID 10 .

HPE Recommendations for High Availability

One of HPE’s key recommendations for high availability is to be on the recommended HPE 3PAR OS for your system so that you have the most recent bug fixes and performance improvements.

When standing up new systems or adding new workloads to a 3PAR array, it’s important to plan for device failure. You need to be sure that if a node fails, the other node in the node pair has sufficient resources to be able to assume the workload of the failed node.

The balance charts provided by IntelliMagic Vision provide an easy way to look at host port utilization over time to make sure that workloads haven’t shifted over time to put the array in a position where an overloaded host port can cause performance issues during failover.

The balance chart shown in Figure 4 makes it easy to see the average utilization (green dot), 10^th/90^th percentile (green box), and minimum and maximum utilization for each port (yellow box) over any time period.

Figure 4 – Balance Chart

In this case, port 0:4:3, which has a peak utilization of 86%, might have performance issues at peak utilization if it’s partner port, 1:4:3 (which has a peak utilization of 31% utilization) should fail.

Virtual Volumes

Since 3PAR deduplication works at a granularity of 16 KiB, consider using host allocation units of multiples of 16 KiB for volumes residing in CPGs that make use of deduplication. Similarly, since deduplication is done at the CPG level, consider putting virtual volumes with high duplicate affinity in the same CPG. Also bear in mind that data that has been previously deduplicated, compressed, or encrypted are poor candidates for deduplication.

Adaptive Optimization (AO)

When using 3PAR’s Adaptive Optimization feature, you should always size the solution assuming the nearline (NL) tier will contribute 0% of the IOPS required for the solution. Similarly, don’t create an AO configuration that has only SSD and NL tiers, as the algorithm will push as much of the performant data as possible to the SSD tier, resulting in less than desired outcome.

Finally, if SSDs are used in AO configurations, no thin-provisioned volumes should be directly associated with SSD CPGs. This assures that the SSD capacity is only consumed by the data that requires SSD performance.

Remote Copy

When zoning Remote Copy Fibre Channel ports, each zone should only contain one primary RC/FC port and one secondary RC/FC port. For ease of administration, all virtual volumes for a particular application should be in the same Remote Copy group. If you’re using synchronous mode, don’t replicate FC or SSD virtual volumes to NL target volumes because the performance of the NL disks might impact the performance of the primary volume due to much slower write acknowledgement.

Priority Optimization (QoS)

When implementing maximum IOPS/Bandwidth limitations, use System Reporter or a solution such as IntelliMagic Vision to quantify the performance needs of the solution so that rules can be set accordingly.

Figure 5 shows how easy it can be to analyze IOPS and Throughput over any given time period to make sure you fully understand the workload for a group of virtual volumes.

Figure 5: Throughput and IOPS

Ensure Your HPE 3PAR is Utilizing Best Practices

Making sure that you start with and keep a strong foundation doesn’t have to be a hard and time-consuming task. The proper tools can make that job much easier and free you up to spend more time innovating.

While there is reporting software that can be acquired from HPE or by external sources that can help gather and interpret performance and capacity metrics, IntelliMagic Vision provides the best combination of ease of setup, intelligent adjustable thresholds, ease of use, and sophisticated reporting capabilities.

IntelliMagic Vision provides AI-driven analytics for not only HPE 3PAR systems, but also multiple other storage platforms and Brocade or Cisco SAN switches, giving you the end-to-end visibility necessary in today’s fast-moving datacenter. Click here to learn more about IntelliMagic Vision for SAN or click here to start a free trial.