SPECstorage Solution 2020_ai_image Result: Qumulo, Inc. – Azure Native Qumulo – Public Cloud Reference

Copyright © 2016-2024 Standard Performance Evaluation Corporation

Qumulo, Inc. SPECstorage Solution 2020_ai_image = 700 AI_Jobs
Azure Native Qumulo – Public Cloud Reference Overall Response Time = 0.82 msec

Performance

Business
Metric
(AI_Jobs)
Average
Latency
(msec)
AI_Jobs
Ops/Sec
AI_Jobs
MB/Sec
50 0.936 21752 4890
100 0.831 43505 9778
150 0.820 65258 14671
200 0.807 87011 19558
250 0.779 108764 24449
300 0.793 130517 29339
350 0.783 152270 34229
400 0.797 174022 39117
450 0.814 195775 44011
500 0.814 217528 48902
550 0.830 239281 53790
600 0.865 261034 58682
650 0.877 282787 63575
700 0.891 304540 68462
Performance Graph

Product and Test Information

Azure Native Qumulo – Public Cloud Reference
Tested by Qumulo, Inc.
Hardware Available May 2024
Software Available May 2024
Date Tested May 28th, 2024
License Number 6738
Licensee Locations Seattle, WA USA

None

Solution Under Test Bill of Materials

Item No Qty Type Vendor Model/Name Description
1 33 Microsoft Azure Virtual Machine Microsoft Standard_E32d_v5 The Standard_E32d_v5 Azure Virtual Machine runs on the 3rd Generation Intel® Xeon® Platinum 8370C (Ice Lake) processor in a hyper-threaded configuration. Each VM features 32 vCPU cores, 256 GiB of memory, 8 NICs, and supports up to 16000 Mbps of network egress bandwidth. Network ingress is not metered, as described https://learn.microsoft.com/en-us/azure/virtual-network/virtual-machine-network-throughput (“Ingress isn’t metered or limited directly”). For testing, 32 VMs were used as Spec Load Generators, and 1 VM was designated as the Prime Client. Each VM utilized a single NIC with Advanced Networking enabled.

Configuration Diagrams

  1. ANQ-reference-architecture

Component Software

Item No Component Type Name and Version Description
1 Azure Native Qumulo filesystem File System 7.1.0 The Azure Native Qumulo (v2) File System is a cloud service offering provisioned directly from the Azure portal. This architecture is entirely different from previous releases of Qumulo Core Filesystems. Customers specify the initial capacity and a customer subnet delegated to Qumulo.Storage/FileSystems, as explained here: https://learn.microsoft.com/en-us/azure/virtual-network/subnet-delegation-overview. Qumulo uses vNet injection to add only IP addresses to the customer network. Customers can request performance increases on-demand through service requests. From a service consumption perspective, the IP addresses are the only architectural component customers need to be aware of, similar to how a virtual network service endpoint functions.
2 Ubuntu Operating System 22.04 The Ubuntu 22.04 operating system is deployed on all Standard_E32d_v5 VMs.

Hardware Configuration and Tuning – Virtual

Component Name
Parameter Name Value Description
Accelerated Networking Enabled Microsoft Azure Accelerated Networking enables single root I/O virtualization (SR-IOV)

Hardware Configuration and Tuning Notes

None

Software Configuration and Tuning – Virtual

Component Name
Parameter Name Value Description
sysctl 1 custom sysctl settings used

Software Configuration and Tuning Notes

/etc/sysctl: net.core.wmem_max = 16777216 net.core.wmem_default = 16777216
net.core.rmem_max = 16777216 net.core.rmem_default = 16777216 net.ipv4.tcp_rmem
= 1048576 8388608 16777216 net.ipv4.tcp_wmem = 1048576 8388608 16777216
net.core.optmem_max = 2048000 net.core.somaxconn = 65535 net.ipv4.tcp_mem =
4096 89600 4194304 net.ipv4.tcp_window_scaling = 1 net.ipv4.tcp_timestamps = 0
net.ipv4.tcp_sack = 1 net.ipv4.tcp_no_metrics_save = 1 net.ipv4.route.flush = 1
net.ipv4.tcp_low_latency = 1 net.ipv4.ip_local_port_range = 1024 65535
net.ipv4.tcp_slow_start_after_idle = 0 net.core.netdev_max_backlog = 300000
vm.dirty_expire_centisecs = 100 vm.dirty_writeback_centisecs = 100
vm.dirty_ratio = 20 vm.dirty_background_ratio = 5 net.ipv4.tcp_sack = 0
net.ipv4.tcp_dsack = 0 net.ipv4.tcp_fack = 0 fs.file-max = 2097152

Service SLA Notes

Azure Service SLA:
https://www.microsoft.com/licensing/docs/view/Service-Level-Agreements-SLA-for-Online-Services

Storage and Filesystems

Item No Description Data Protection Stable Storage Qty
1 Azure Native Qumulo persists data using Azure Blob Storage, which is a Microsoft-managed service providing cloud storage that is highly available, secure, durable, scalable, and redundant.Architecturaly, Read cache is serviced from an in-memory L1 cache and a NVMe L2 cache. The global read-cache is increased on-demand based on customer service requests, elastically. Write transactions leverage high performance Azure Managed disks, which act as a protected write-back cache for incoming writes, continuously flushing them to Azure Blob Storage. Provided by Microsoft Azure Blob Storage Azure Blob Storage 64
Number of Filesystems 1
Total Capacity 100 TB
Filesystem Type Azure Native Qumulo

Filesystem Creation Notes

Azure Native Qumulo is deployed directly from the Azure Portal, the Azure CLI,
or using an Azure REST endpoint. Customers specify the capacity of storage and
what subnet they want to deploy vNet-injected IP addresses in. Requested
aggregate performance capability of 100GB/s and 500,000 IOPS by submitting a
service ticket to the service in Azure.

Storage and Filesystem Notes

Azure Native Qumulo is a cloud storage service that has a different
architecture than traditional storage architectures, including a traditional
Qumulo Core cluster. Data is persisted to Azure Blob, with only caching
occuring at the compute layer. This disaggregates the storage persistence and
file system performance from data storage at rest, allowing each layer to scale
independently and elastically. The Azure Native Qumulo filesystem is built on
top of the Azure Managed Blob (LRS) object tier. The architecture acts as an
accelerator that executes parallelized reads which are prefetched from object,
and served directly from the filesystem to the clients (over NFSv3 in this
case). Read cache is serviced from an in-memory L1 cache and a NVMe L2 cache.
The global read-cache is increased on-demand, elastically. Write transactions
leverage high performance Azure Managed disks, which act as a protected
write-back cache for incoming writes, continuously flushing them to Azure Blob
Storage.

Transport Configuration – Virtual

Item No Transport Type Number of Ports Used Notes
1 NFS version 3 97 The mount options used were only tcp and vers=3 (without nconnect). IP addresses are distributed to clients using a round-robin DNS record. Each client in this benchmark used two of the vNet-injected IP addresses per NFS client, resulting in 2 NFS mounts per load generator, each to distinct IP addresses.
2 Azure vNet injected vNIC 64 Azure Native Qumulo uses vNet injection to place these NICs onto a custoer owned subnet. They provide access to the filsystem..

Transport Configuration Notes

None

Switches – Virtual

Item No Switch Name Switch Type Total Port Count Used Port Count Notes
1 Azure 100Gbps Ethernet with Enhanced Networking 97 97 In Azure, there is not awareness of what the actual physical switch port sizes are. Since a value is required we set it to the number of NICs used.

Processing Elements – Virtual

Item No Qty Type Location Description Processing Function
1 1 vStorage Azure Region EastAsia Azure Native Qumulo Cloud Storage Service Qumulo File System, Network communication, Storage Functions

Processing Element Notes

None

Memory – Virtual

Description Size in GiB Number of Instances Nonvolatile Total GiB
Azure Standard_E32d_v5 VM memory 256 33 V 8448
Grand Total Memory Gibibytes 8448

Memory Notes

None

Stable Storage

Azure Blob Storage is a cloud storage solution provided by Microsoft Azure
designed to store large amounts of unstructured data. This data can include
text, binary data, documents, media files, backups, and logs. Blob Storage is
optimized for storing and retrieving large blobs of data efficiently, providing
scalable, durable, and highly available storage for various types of
applications and workloads. Azure Blob Storage uses HMAC to ensure the
integrity of data. When data is uploaded, a hash is computed and stored. When
data is retrieved, the hash is recomputed and compared to ensure data
integrity. The Azure Blob Storage class used is the Locally Redundant Storage
(LRS) which keeps multiple copies of data within a single region.

Solution Under Test Configuration Notes

The solution under test was a standard Azure Native Qumulo cluster deployed
natively in Azure. The cluster is elastic both in performance and capacity and
can be requested to handle 100’s of GB/s of throughput, millions of IOPS, and
100’s of PB’s of capacity.

Other Solution Notes

Requested aggregate performance capability of 100GB/s and 500,000 IOPS for the
Azure Native Qumulo performance. This can be changed with a service ticket in
Azure. The Azure Native Qumulo (v2) File System is a cloud service offering
that is provisioned directly from the Azure portal. This architecture is
completely different than prior releases of Qumulo Core Filesystems. Customers
specify the capacity they wish to start with and a customer subnet delegated to
Qumulo.Storage/FileSystems.

Dataflow

None

Other Notes

None

Other Report Notes

None


Generated on Tue Jun 4 01:14:12 2024 by SpecReport
Copyright © 2016-2024 Standard Performance Evaluation Corporation

Scroll to Top