Configuring the HPE Proliant DL380 Gen9 24-SFF CTO Server as a Vertica Node

Click here for a PDF version of the DL380 Hardware Guide.

Using the HPE DL380 Gen9 24-SFF Server as a Vertica Node

The Vertica Analytics Platform software runs on a shared-nothing MPP cluster of peer nodes. Each peer nodes is independent, and processing is massively parallel. A Vertica node is a hardware host configured to run an instance of Vertica.

This document provides recommendations for configuring an individual DL380 Gen9 24-SFF CTO Server as a Vertica node.

The recommendations presented in this document are intended to help you create a cluster with the highest possible Vertica software performance.

This document includes a Bill of Materials (BOM) as a reference and to provide more information about the DL380 Gen9 24-SFF CTO Server.

Recommended Software

This document assumes that your services, after you configure them, will be running the following minimum software versions:

  • Vertica 7.2 (or later) Enterprise Edition. This is the most recent release as of April 2016.
  • Red Hat Enterprise Linux 6.x.

If you are running Red Hat Enterprise Linux 7.x, watch for information in this document that is clearly marked as RHEL 7.1 specific.

Selecting a Server Model

The HPE DL380 Gen9 product family includes several server models. The best model for maximum Vertica software performance is the DL380 Gen9 24-SFF CTO Server (part number 767032-B21).

Selecting a Processor

For maximum price/performance advantage on your Vertica database,  the DL380 Gen9 24-SFF servers used for Vertica nodes should include two (2) Intel Xeon E5-2690v3 2.6 GHz/12-core DDR4-2133 135W processors.

This processor recommendation is based on the fastest 12-core processors available for the DL380 Gen9 24-SFF platform at the time of this writing. These processors allow Vertica to deliver the fastest possible response time across a wide spectrum of concurrent database workloads.

The processor’s faster clock speed directly affects the Vertica database response time. Additional cores enhance the cluster’s ability to simultaneously execute multiple MPP queries and data loads.

Selecting Memory

For maximum Vertica performance, DL380 Gen9 24-SFF servers used as Vertica nodes should include 256 GB of RAM. Configure this memory as follows:

  • 8 x 32 GB DDR4-2133 RDIMMs, 1DPC (32 GB per channel)
  • In the field, you can expand this configuration to 512 GB by adding
    8 x 32 GB DIMMs.

A two-processor DL380 Gen9 24-SFF server has 8 memory channels with 3 DIMM slots in each channel, for a total of 24 slots. DL380 Gen9 24-SFF memory configuration should comply with DIMM population rules and guidelines:

  • Do not leave any channel completely blank. Load all channels similarly.
  • Populate the maximum number of DIMMS per channel (DPC) to 2. Doing so allows you to use the highest supported DIMM speed of 2133 MHz. DPCs with 3 DIMMs run at 1866 MHz or lower.

Note Follow these guidelines to avoid a reduction in memory speed that could adversely affect the performance of your Vertica database.

The preceding recommended memory configuration is based on 32 GB DDR4 2133 MHz DIMMs and 256 GB of RAM. That configuration is intended to achieve the best memory performance while providing the option of future expansion. The following table provides several alternate memory configurations:

Sample Configuration Total Memory Considerations

8 x 16 GB DDR4-2133 RDIMMs

1 DPC (8 GB per channel)

128 GB

A low-memory option for systems with less concurrency and slower speed requirements.

16 x 16 GB DDR4-2133 RDIMMs

2 DPC (16 GB, 16 GB per channel)

256 GB

A slightly less expensive option for 256 GB of RAM that does not allow expansion in the field.

8 x 32 GB DDR4-2133 RDIMMs

1 DPC (32 GB per channel)

256 GB

The standard memory recommendation for Vertica.

16 x 32 GB DDR4-2133 RDIMMs

2 DPC (32 GB + 32 GB per channel)

512 GB

A high-memory option that may be beneficial to support some database workloads.

Selecting and Configuring Storage

Configure the storage hardware as follows for maximum performance of the DL380 Gen9 24-SFF server used as a Vertica node: 

  • 1x HPE DL380 Gen9 24-SFF CTO Chassis with 24 Hot Plug SmartDrive SFF (2.5-inch) Drive Bays
  • 1x HPE DL380 Gen9 2-SFF Kit with 2 Hot Plug SmartDrive SFF (2.5 inch) Drives Bays (on the back of the server)
  • 1x HPE Smart Array P440ar/2GB FBWC 12Gb 2-ports Int FIO SAS Controller (integrated on system board)
  • 2x 300 GB 12 G SAS 10 KB 2.5 inch SC ENT drives (configured as RAID1 for the OS and the Vertica Catalog location)
  • 24x 1.2 TB 12 G SAS 10 KB 2.5 inch SC ENT drives (configured as one RAID 1+0 device for the Vertica Data location, for approximately 13 TB total formatted storage capacity per Vertica node)

You can configure a Vertica node with less storage capacity:

  • Substitute 24x 1.2 TB 12 G SAS 10 KB 2.5 inch SC ENT drives with 24x HPE 600 GB 12 G SAS 10 K 2.5 inch SC ENT drives.
  • Configure the drives as one RAID 1+0 device for the Vertica data location, for approximately 6 TB of total data storage capacity per Vertica node.

Alternatively, you can configure the 23rd and 24th 1.2 TB (or 600 GB) data drives (for 22 active drives in total) as hot spares. However, such configuration is unnecessary with a RAID 1+0 configuration.

Vertica can operate on any storage type. For example, Vertica can run on internal storage, a SAN array, a NAS storage unit, or a DAS enclosure. In each case, the storage appears to the host as a file system and is capable of providing sufficient I/O bandwidth. Internal storage in a RAID configuration offers the best price/performance/availability characteristics at the lowest TCO.

A Vertica installation requires at least two storage locations―one for the operating system and catalog, and the other for data. Place these data locations on a dedicated, contiguous-storage volume.

Vertica is a multithreaded application. The Vertica data location I/O profile is best characterized as large block random I/O.

Drive Bay Population

The 26 drive bays on the DL380 Gen9 24-SFF servers are attached to the Smart Array P440ar Controller over 4 internal SAS port connectors (through a 12 G SAS Expander Card) as follows:

  • Cages 1, 2, 3—Drive bays 1‒8, 9‒16,17‒24 (8 drives each)
  • Cage 4—Drive bays 25 and 26 (2 drives)

For example, an ideal implementation of the recommended 26-drive configuration is:

  • 300 GB drives placed in bays 25 and 26, with the 2SFF drive expander in the rear of the server.
  • 2 TB (or 600 GB) drives placed in all bays on the front of the server. This approach spreads the Vertica data RAID10 I/O evenly across the SAS groups.

Note For best performance, make sure to fully populate all the drive bays. 

Protecting Data on Bulk Storage

The HPE Smart Array P440ar/2GB FBWC 12 Gb 2-ports Int FIO SAS Controller offers the optional HPE Secure Encryption capability that protects data at rest on any bulk storage attached to the controller. (Additional software, hardware, and licenses may be required.) For more information, see HPE Secure Encryption product details.

Data RAID Configuration

The 24 data drives should be configured as one RAID 1+0 device as follows:

  • The recommended strip size for the data RAID 1+0 is 512 KB, which is the default setting for the P440ar controller.
  • The recommended Controller Cache (Accelerator) Ratio is 10/90, which is the default setting for the P440ar controller.
  • The logical drive should be partitioned with a single primary partition spanning the entire drive.

Place the Vertica data location on a dedicated physical storage volume.  Do not co-locate the Vertica data location with the Vertica catalog location. Vertica Packard Enterprise recommends that the Vertica catalog location on an Vertica node on a DL380 Gen9 24-SFF server be the operating system drive.

For more information, read Before You Install Vertica in the product documentation, particularly the discussion of Vertica storage locations.

Note  Vertica does not support storage configured with the Linux Logical Volume Manager in the I/O path. This limitation applies to all Vertica storage locations including the catalog which is typically placed on the OS drive.

Linux I/O Subsystem Tuning

To support the maximum performance DL380 Gen9 24-SFF node configuration, Vertica Packard Enterprise recommends the following Linux I/O configuration settings for the Vertica data location volumes:

  • The recommended Linux file system is ext4.
  • The recommended Linux I/O Scheduler is deadline.
  • The recommended Linux Readahead setting is 8192 512-byte sectors (4 MB).

The current configuration recommendations differ from the previously issued guidance due to the changes in the Vertica I/O profile implemented in Vertica 7.x.

System administrators should durably configure the deadline scheduler and the read-ahead settings for the Vertica data volume so that these settings persist across server restarts.

Caution Failing to use the recommended Linux I/O subsystem settings will adversely affect performance of the Vertica software.

Data RAID Configuration Example

The following configuration and tuning instructions pertain to the Vertica data storage location.

Note The following steps are provided as an example, and may not be correct for your machine.

Verify the drive numbers and population for your machine before running these commands.

  1. In Red Hat Enterprise Linux 7.x, to load the modules that HPSSACLI requires, execute the following commands:

    modprobe sg
     modprobe hpsa hpsa_allow_any=1
     first
  1. View the current storage configuration:

    # hpssacli
    >controller slot=1 show
  1. Assume that the RAID1 OS drive is configured as a logical drive 1 and the 24 data drives are either “loose” or pre-configured into a temporary logical drive. To create a new logical drive with recommended parameters, any non-OS logical drives must be destroyed, and content will be lost. To destroy a non-OS logical drive:

    >controller slot=1 ld 2 delete forced
  1. Create a new RAID10 data drive with 512 K strip size:

    >controller slot=1 create type=ld raid=1+0 ss=512 drives=allunassigned
  1. Partition and format the RAID10 data drive:

    # parted -s /dev/sdb mklabel gpt mkpart primary ext4 0% 100%
     # mkfs.ext4 /dev/sdb1
  1. Create a /data mount point, add a line to the /etc/fstab file, and mount the Vertica data volume:

    # mkdir /data
     [add line to /etc/fstab]: /dev/sdb1  /data  ext4  defaults,noatime  0 0
     # mount /data
  1. So that the Linux I/O scheduler Linux Read-Ahead, and hugepage defragmentation settings persist across system restarts, add the following five lines to /etc/rc.local. Apply these steps to every drive in your system.

    Note The following commands assume that sdb is the data drive, and sda is the OS/catalog drive.

    echo deadline > /sys/block/sdb/queue/scheduler
     blockdev --setra 8192 /dev/sdb
     echo never > /sys/kernel/mm/redhat_transparent_hugepage/enabled
     echo never > /sys/kernel/mm/redhat_transparent_hugepage/defrag
     echo no > /sys/kernel/mm/redhat_transparent_hugepage/khugepaged/defrag
     echo deadline > /sys/block/sda/queue/scheduler
     blockdev –setra 2048 /dev/sda
  1. After you have configured the storage, run vioperf to understand the baseline I/O performance.

The minimum required I/O is 20 MB/s read and write per physical processor core on each node. This value is based on running in full duplex and reading and writing at this rate simultaneously, concurrently on all nodes of the cluster.

Vertica Packard Enterprise recommends that I/O is 40 MB/s per physical core on each node. For example, the I/O rate for a server node with 2 hyperthreaded 6-core CPUs is a required minimum of 240 MB/s. Vertica Packard Enterprise recommends 480 MB/s. A properly configured DL380 Gen9 significantly outperforms this.

Your I/O performance should be close to:

  • 2,200 MB/s for read and write when using 15K RPM drives.
  • 2,200 MB/s write and 1,500 MB/s read when using 10 K RPM drives.
  • 800+800MB/s for rewrite
  • 7k+ for seeks

If you do not see those results, review the preceding steps to make sure you configured your data storage location correctly.

For more information, see vioperf in the product documentation.

Selecting a Network Adapter

To support the maximum performance MPP cluster operations, DL380 Gen9 24-SFF servers used as Vertica nodes should include at least two 10 Gigabit Ethernet (RJ-45) ports:

  • HPE Ethernet 10 Gb 2-port 561FLR-T Adapter in the FlexibleLOM slot

Alternatively, if you need SFP+ networking,  DL380 Gen9 24-SFF servers used as Vertica nodes should include at least two (2) 10 Gigabit Ethernet (SFP+) ports.

  • HPE Ethernet 10 Gb 2-port 546FLR-SFP+Adapter in the FlexLOM slot

A Vertica cluster is formed with DL380 Gen9 24-SFF servers, associated network switches, and Vertica software.

When used as a Vertica node, each DL380 Gen9 24-SFF server should be connected to two separate Ethernet networks:

  • The private network, such as a cluster interconnect, is used exclusively for internal cluster communications. This network must be the same subnet, dedicated switch, or VLAN, 10 Gb Ethernet. Vertica does TCP P2P communications and UDP broadcasts on this network. IP addresses for the private network interfaces must be assigned statically. No external traffic should be allowed over the private cluster network.
  • The publicnetwork is used for database client (i.e., application) connectivity, should be 10 Gb Ethernet. Vertica has no rigid requirements for public network configuration. However, Vertica Packard Enterprise recommends that you assign static IP addresses for the public network interfaces.

The private network interconnect should have Ethernet redundancy. Otherwise, the interconnect (specifically the switch) would be a single point of a cluster-wide failure.

Cluster operations are not affected even in the event of a complete failure of a public network. Thus, public network redundancy is not technically required. However, if a failure occurs, the application connectivity to the database is affected. Therefore, consider public network redundancy for continuous availability of the entire environment.

To achieve redundancy on both the private and public networks:

  1. Take the two ports from the Ethernet card on the server and run one to each of the two top of rack switches (which are bonded together in an IRF).
  2. Bond the links together using LACP.
  3. Divide the links into public and private networks, using VLANs.

Configuring the Network

The following figure illustrates a typical network setup that achieves high throughput and high availability. (This figure is for demonstration purposes only.)

This figure shows that the bonding of the adapters allows for one adapter to fail without the connection failing. This bonding provides high availability of the network ports. Bonding the adaptors also doubles the throughput.

Furthermore, this configuration allows for high availability with respect to the switch. If a switch fails, the cluster does not go down. However, it may reduce the network throughput by 50%.

DL380.Network.png

Tuning the TCP/IP Stack

Depending on your workload, number of connections, and client connect rates, you need to tune the Linux TCP/IP stack to provide adequate network performance and throughput.

The following script represents the recommended network (TCP/IP) tuning parameters for a Vertica cluster. Other network characteristics may affect how much these parameters optimize your throughput.

Add the following parameters to the /etc/sysctl.conf file. The changes take effect after the next reboot.

##### /etc/sysctl.conf
 # Increase number of incoming connections 
 net.core.somaxconn = 1024 
 #Sets the send socket buffer maximum size in bytes.
 net.core.wmem_max = 16777216
 #Sets the receive socket buffer maximum size in bytes.
 net.core.rmem_max = 16777216
 #Sets the receive socket buffer default size in bytes.
 net.core.wmem_default = 262144
 #Sets the receive socket buffer maximum size in bytes.
 net.core.rmem_default = 262144
 #Sets the maximum number of packets allowed to queue when a particular interface receives packets faster than the kernel can process them.
 # increase the length of the processor input queue 
 net.core.netdev_max_backlog = 100000
 net.ipv4.tcp_mem = 16777216 16777216 16777216 
 net.ipv4.tcp_wmem = 8192 262144 8388608 
 net.ipv4.tcp_rmem = 8192 262144 8388608
 net.ipv4.udp_mem = 16777216 16777216 16777216 
 net.ipv4.udp_rmem_min = 16384 
 net.ipv4.udp_wmem_min = 16384

HPE ROM-Based Setup Utility (BIOS) Settings

Vertica Packard Enterprise recommends that you configure the BIOS for maximum performance:

System Configuration > BIOS/Platform Configuration (RBSU) > Power Management > HPE Power Profile > [Maximum Performance]

Additionally, you must modify the max_cstate value.

For Red Hat Enterprise Linux versions prior to 7.0, modify the kernel entry in the /etc/grub.conf file by appending the following to the kernel command:

intel_idle.max_cstate=0 processor.max_cstate=0

For Red Hat Enterprise Linux 7.x, /etc/grub.conf has become /boot/efi/EFI/[centos or redhat]grub/conf. You need to append the following arguments to the following kernel command:

intel_idle.max_cstate=0 processor.max_cstate=1 intel_pstate=disable 

HPE ProLiant DL380 Gen9 24-SFF Bill of Materials

The following table contains a sample Bill of Materials (BOM). Note the following about this BOM:   

  • Part numbers are listed for reference only.
  • Networking equipment and racks are not included.
  • All hardware requests should come through the CSE team for review.
Quantity Part Number Description Notes

1

767032-B21

HPE DL380 Gen9 24-SFF CTO Server

1

719044-L21

HPE DL380 Gen9 E5 2690v3 FIO Kit

For 12 cores

1

719044-B21

HPE DL380 Gen9 E5 2690v3 Kit

For 12 cores

8

726722-B21

HPE 32 GB 4Rx4 PC42133P-L Kit

For 256 GB of memory (expandable to 512 GB)

1

727250-B21

HPE 12 GB SAS Expander Card

1

724864-B21

HPE DL380 Gen9 2SFF Kit

1

SG506A

HPE C13-C14 2.5 ft Sgl Data Special

Requirement for deployments using HPE Intelligent PDUs

1

SG508A

HPE C13-C14 4.5 ft Sgl Data Special

Requirement for deployments using HPE Intelligent PDUs

1

768896-B21

HPE DL380 Gen9 Rear Serial Cable Kit

1

749974-B21

HPE Smart Array  P440ar/2GB FBWC 12Gb 2-ports Int FIO SAS Controller

2

785067-B21

HPE 300 GB 12 GB SAS 10KB 2.5 in SC ENT HDD

For operating system and database catalog

24

781518-B21

HPE 1.2 TB 12 GB SAS 10KB 2.5 in SC ENT HDD

For approximately 13 TB formatted storage capacity

1

700699-B21

HPE Ethernet 10Gb 2P 561FLR-T Adapter

For RJ-45 networking

2

720479-B21

HPE 800W FS Play Ht Plg Pwr Supply Kit

1

733660-B21

HPE 2U SFF Easy Rail Kit

Alternate Part Appendix

Processor

Quantity Part Number Description Notes

1

719046-L21

HPE DL380 Gen9 E5 2670v3 FIO Kit

12 cores at 2.3 GHz

1

719046-B21

HPE DL380 Gen9 E5 2670v3 Kit

Memory

Quantity Part Number Description Notes

16

726722-B21

HPE 32 GB 4Rx4 PC4 2133P-L Kit

For 512 GB of memory

Disk Drives

Quantity Part Number Description Notes

24

759212-B21

HPE 600 GB 12 GB SAS 10 KB 2.5 in SC ENT HDD

For approximately 6 TB formatted storage capacity

Networking

Quantity Part Number Description Notes

1

779799-B21

HPE Ethernet 10GbE 546FLR-SFP+ Adapter

For SFP+ networking

For More Information