StorPool presents at Cloud Field Day - the leading technology event focused on the impact of cloud technologies on enterprise IT. During the event, the high-performance block storage specialist will showcase how its storage technology allows cloud builders to easily outperform cloud titans like AWS, Microsoft Azure and GCP.
Performance is of major importance for modern applications and workloads. No matter if you run a private cloud or deliver public cloud services for customers, you need to ensure the excellent performance for the workloads running on the cloud. Often misunderstood, storage has a direct impact not only on the reliability of cloud services, but also on the performance of the entire cloud.
https://storpool.com/news/storpool-presents-at-cloud-field-day-9
2. 2
1. Introduction to StorPool Storage
2. Who uses StorPool and why
3. Understanding performance
4. Performance measurement and comparison of leading public clouds
5. How StorPool-powered clouds compete
6. Conclusions
Agenda
3. ● Chief of Product & Co-founder at StorPool
● 20+ years in ISPs, SDN, SDS
● IT Infrastructure with a focus on invention,
performance & efficiency
About me
https://www.linkedin.com/in/krosnov/
@bkrosnov
bk@storpool.com
3
Boyan Krosnov
4. About StorPool
4
● Leader in high-performance, scale-out, block-storage software for
demanding primary workloads
○ Runs on standard servers and turns them into high-end, all-flash
storage system
○ Focus on New-Age IT stacks – KVM, Kubernetes (still supports legacy
VMware/Hyper-V/XenServer, etc. stacks)
○ Comes as a Fully Managed Storage – mission critical 24/7 support,
constant monitoring, proactive issue resolution, SLA
● Founded in 2011 and scaled up to a global vendor with customers on 4
continents, profitable & growing
● HQ in Sofia, Bulgaria
● StorPool is built from the ground up to provide public/private cloud builders
with the fastest and most reliable software-defined storage solution on the
market!
5. 5
StorPool in a nutshell
Best-in-class Software-defined
Scale-out Block Storage Platform
SDS 2.0 – Feature Rich Shared
Storage System Faster than Local
SSD
Purpose Built Fully Distributed System –
With own On-disk Format, Protocol,
Quorum and more…
Replace SAN, AFA and Other Storage
Software
Robust and
High-quality
End to End
Data Integrity
Scale-out / Low
Latency /
High Availability
NVMe
Performance
Feature-rich Modern
SDS 2.0 System
Purpose Built for
the Cloud Era
6. ● SDS - runs on Linux
● No special hardware requirements
○ Just NICs and NVMes
○ Also supports SAS & SATA
drives, controllers
● Scale-out system, active/active
multi-controller, pooling of capacity,
pooling of performance
● Copy-on-write on-disk format
● End-to-end data integrity
● 3-way replication
Architecture
6
● Low latency <100 μs end-to-end
(measured in VM)
● High throughput - 1M IOPS per
storage node (10M+ IOPS in a
10-node system)
● Highly efficient storage protocol and
implementation, Kernel bypass
● API control. Common deployments
are with 10000+ volumes, snapshots
● Integrations with OpenStack,
CloudStack, Kubernetes, etc.
● Snapshots and multi-site capabilities
● Monitoring and fine-grained metrics,
analytics
8. storpool_server instance
1 CPU thread
2-4 GB RAM
NIC
storpool_server instance
1 CPU thread
2-4 GB RAM
storpool_server instance
1 CPU thread
2-4 GB RAM
storpool_block instance
1 CPU thread
NVMe SSD
NVMe SSD
NVMe SSD
NVMe SSD
NVMe SSD
NVMe SSD
KVM Virtual Machine
KVM Virtual Machine
• Highly scalable & efficient parallel architecture
• Scales up in each storage node & out with multiple nodes
25GbE
. . .
25GbE
8
9. 9
1. Introduction to StorPool Storage
2. Who uses StorPool and why
3. Understanding performance
4. Performance measurement and comparison of leading public clouds
5. How StorPool-powered clouds compete
6. Conclusions
Agenda
10. Customer profile
Typical
● Hosting, MSP, Public Cloud, IaaS
● New IT, Modern IT - automation, APIs, integration/orchestration,
continuous improvements
● > 100TB Tier 1 Primary Storage (SSD, NVMe)
● 1 to 10s of locations
● KVM hypervisor and/or containers/Kubernetes
… and also
● Sizable private clouds - test/dev, production SaaS, Web/Mobile
Apps, Gaming
● Multiple stacks - VMWare, Hyper-V, XenServer
● Upgrade for legacy IT
10
11. The Project:
IT unification & consolidation: standardizing on a New-Age &
scalable IT stack company wide. Hardware from multiple
vendors - HPE, IBM, EMC, NetApp, Cisco, etc.
The goals:
Consolidate 10s of deployments from legacy platforms
(acquired MSPs) to one new, future-proof platform.
Benefit from modern IT practices - automation, scalability,
hardware independence
Streamline business operations and improve profitability
The solution:
New-Age IT IaaS platform - Software-Defined (SDDC design)
The New-Age stack: KVM+OpenNebula, StorPool, Mellanox,
Cumulus Linux
Case study #1 - Dustin
11
Company Profile:
Managed Service Provider
1.275 Billion Euro Net Sales
1800+ Employees
6 locations in Sweden, Norway,
Denmark
Grown through acquisitions
Had multiple IT stacks (VMware,
Hyper-V, Linux)
https://storpool.com/storpool_
case-study_dustin_2020
12. Case study #2 - Amito
12
Company Profile:
UK-based Cloud and Managed Service
Provider
Founded in 2000
Presence in 28 data centers worldwide
The Project:
"Тhe most difficult part of the infrastructure stack for high
availability virtualized servers is storage. Amito was using quite a
few different solutions, from big-name vendors to in house
developed HA SANs based on ZFS. None of these solutions
brought the expected results."
The goals:
Implement a high-performance storage across multiple data
centers to support Amito’s broad range of applications, with a
hugely different set of performance and availability requirements.
The solution:
High-performance data storage platform with CloudStack across
several clusters. All storage clusters have been faultless since they
implemented them and with zero downtime. Amito is currently
using approx. 35% less storage than on their previous solution
because of the way StorPool is built.
https://storpool.com/amito-c
ase-study-storpool
13. Case study #3 - Katapult by Krystal
13
Company Profile:
Katapult by Krystal
One of the largest independent UK web
hosting companies.
Founded in 2002 and steadily
grown over the last 18 years
Comprehensive portfolio of hosting,
cloud and VPS services
The Project:
Virtual Infrastructure as a Service platform built for extreme
performance and simplicity with several high-performance
NVMe storage clusters running on StorPool Storage.
The goals:
A storage solution which keeps the promise of a modern SDS;
high availability, extreme performance, a high level of flexibility
and guaranteed data integrity.
The solution:
Virtual Infrastructure as a Service platform, ensuring high
performance, robust API, unique space savings features and
the extremely high level of data protection offered from
StorPool’s triple data replication.
https://storpool.com/krystal-k
atapult-case-study
14. 14
1. Introduction to StorPool Storage
2. Who uses StorPool and why
3. Understanding performance
4. Performance measurement and comparison of leading public clouds
5. How StorPool-powered clouds compete
6. Conclusions
Agenda
15. Why performance
● Better application performance -- e.g. time to load a page, time to
rebuild, time to execute specific query
● Happier customers in cloud and multi-tenant environments
● ROI, TCO - Lower cost per delivered resource (per VM) through
higher density
● For public cloud - win customers over from your competitors
● For private cloud - do more with less; win applications /
workloads / teams over from public cloud
15
24. Storage benchmarks
Beware of caveats...
● Performance numbers from hardware configurations totally
unlike what you’d use in production
● Synthetic tests with high iodepth - 10 nodes, 10 workloads *
iodepth 256 each. (because why not)
● Testing with ramdisk backend
● Synthetic workloads don't approximate real world
24
25. ● Design benchmarks which reflect your use-case and application
● Measure what matters. Examples:
○ developer productivity - simple SQL database for up to X users, so no
need to pay for complexity of clusters; runs CI/tests in half the time
○ Efficiency - $ per user, $ per features
● If you can't measure what matters directly, find good proxies. Example:
○ "I can't run my entire stack as a benchmark, but I know it consists of
a load balancer and a transaction-heavy database, so I'll use a load
balancer and a DB benchmark"
Benchmarks
25
26. 26
1. Introduction to StorPool Storage
2. Who uses StorPool and why
3. Understanding performance
4. Performance measurement and comparison of leading public clouds
5. How StorPool-powered clouds compete
6. Conclusions
Agenda
27. * - ramdisk used to reduce usable RAM to 16 GB
VMs and block storage
Provider Service name
$/GiB/mont
h
IOPS limit
Size of
volume
[GiB]
Monthly
cost
Katapult Shared disk NVMe $0.15 unlimited 1024 $154
StorPool Private Cloud 3 NVMe unlimited 500 n/a
AWS EBS gp2 $0.10 3,072 1024 $102
Google Cloud SSD Persistent Disk $0.17 15,000 1024 $174
Google Cloud Balanced Persistent Disk $0.10 6,144 1024 $102
Microsoft Azure Premium SSD $0.12 3,500 1024 $123
Digital Ocean Block Storage $0.10 10,000 1024 $102
Provider Instance name region
monthly cost (with
12 month
commitment)
vCPUs RAM
Katapult ROCK-24 London $120 8 24GB*
StorPool Lab one-8vcpu-16gb Sofia n/a 8 16GB
AWS Compute optimized: c5.2xlarge us-east-2 $245 8 16GB
Google Cloud General purpose: n2-8vcpu-16gb us-central1 $197 8 32GB*
Microsoft Azure
Compute optimized: Standard_F8s_v2 - 8
vcpus, 16 GiB memory
East US 2 $235 8 16GB
Digital Ocean CPU Optimized Droplet: 16GB sfo2 $160 8 16GB
27
28. ● Previous version of our tools and methodology:
○ https://storpool.com/storage-performance-and-resilience-
testing
● We'll be releasing updated tools and method with the write-up
in the next month
○ coremark, fio, rsync, pgbench, sysbench
● Until then drop us an email at info@storpool.com
Benchmarks
28
29. ● Storage heavy, a little CPU
○ FIO, rsync
● Storage & CPU
○ pgbench, sysbench
● CPU, RAM*
○ coremark
● Network*
* - future additions to our suite
Tools used
29
41. 41
1. Introduction to StorPool Storage
2. Who uses StorPool and why
3. Understanding performance
4. Performance measurement and comparison of leading public clouds
5. How StorPool-powered clouds compete
6. Conclusions
Agenda
42. ● Storage performance determines application performance to a
large degree.
● VMs with seemingly identical parameters like CPU and memory
can have orders of magnitude difference in application
performance.
● Both IOPS and latency are important for application performance.
● Applications perform measurably better on StorPool-powered
clouds - up to 2.5 times better than the second-best public cloud
offering and in some case more than 10 times better than the
other leading clouds
Storage performance = Application performance
42
43. 1. Top performance, Top availability, Top support
2. Great solution economics permit competitive positioning of cloud
service
3. StorPool manages storage layer so highly skilled team can focus on
improving the services
How StorPool-powered clouds compete
43
44. 44
1. Introduction to StorPool Storage
2. Who uses StorPool and why
3. Understanding performance
4. Performance measurement and comparison of leading public clouds
5. How StorPool-powered clouds compete
6. Conclusions
Agenda
45. 1. Good application performance is critical for every cloud
2. When measuring performance, measure what matters to you
3. 2x-3x higher application performance (per $ !) than hyperscalers is
within reach for small to mid-sized public and private clouds. Half
cost for the same workload!
4. On your next project work with partners who understand
performance. You can gain a lot!
5. StorPool delivers a unique combination of performance, reliability,
scalability, ease of use and proactive mission-critical support. The
best storage for demanding public and private clouds.
Conclusions
45