SlideShare a Scribd company logo
Introduction to StorPool Storage
Boyan Ivanov, CEO
1960s and on
Mainframes
Client-Server with
Local Storage
1980s and on
Client-Server with
Local Storage
Shared-Disk
Storage Software
1990s and on
Dual-Controller
Shared-Disk Arrays
2000s and on
Shared-Disk
Storage Software
1990s and on
2010s and on
Software-Defined
Storage
SIMPLE & PREDICTABLE
1-3 RACKS
THEY
WORK…
Dynamic
provisioning
?
Self-service
?
Parallel service for
1,000s of workloads
?
Always-on
?
99.999% with
maint. included
?
No single point
of failure
?
Consistent low
latency
?
Linear
Scalability
?
BUT WHAT IF
YOU NEED
AT SCALE
10
racks
200
compute servers
1PB
saved data
5,000
VMs
200+
compute servers
4,000+
10+
racks
200+
1PB+
saved data
20 PB+
5,000+
100,000+
VMs
АТ LARGE SCALE = SCALE × 20
ROAD
TO
People are
simply unhappy
No automation,
ticket-based
IT gets in the
way of “the
business”
Slow to address
changes in
market dynamics
Squeezed profit
margins due to
org & tech debt
ROAD
TO
Mainframes Dual-Controller
Shared-Disk
Storage
Shared-Disk
Storage Software
Software-Defined
Storage
Client-Server with
Local Storage
Mainframes Dual-Controller
Shared-Disk
Storage
Shared-Disk
Storage Software
Software-Defined
Storage
Client-Server with
Local Storage
IS A RADICAL DEPARTURE
StorPool Storage
AGENDA
Introduction to
StorPool Storage
Boyan Ivanov, CEO
Tech Excellence
Boyan Krosnov, CTO
Helping IT Teams
Do Storage Right
Alex Ivanov, Product Lead
StorPool IN A NUTSHELL
Of Flash Storage
Deployments
100s
Countries of
Deployment
30+ Profitable
& Growing
RUNNING WORKLOADS FOR
CUSTOMERS
Dynamic provisioning
Self-service
Parallel service for 1000s of workloads
Always-on
99.999% with maintenance included
Consistent low latency
Linear scalability (no data silos)
No single point of failure
HOW DID WE MAKE THIS?
Dynamic provisioning
Self-service
Parallel service for 1000s of workloads
Always-on
99.999% with maintenance included
Consistent low latency
Linear scalability (no data silos)
No single point of failure
SILICON VALLEY
OF THE EASTERN
BLOCK
KNOW-HOW
FROM A DIFFERENT
DOMAIN
High Performance
Networking
IT Service Providers
COMMITMENT
& GRIT
StorPool Storage
Metrics & Analytics Alerts & Reporting VolumeCare CMP Integrations
FULLY MANAGED SOLUTION
Parallel multi-node
shared-nothing
architecture
Always-on:
Non-disruptive
everything
Made for large-scale
modern IT
infrastructure
StorPool Storage
Software only,
hardware-agnostic
Linearly scalable
Block first,
some file
IT TEAMS HAVE
LONG-TERM
RELATIONSHIPS WITH
Datacenter
providers
Hardware
vendors
Distributors
and resellers
PROVIDES STORAGE AS A SERVICE UNDER
BRING YOUR OWN DEVICE TERMS
Colocation
Cooling
Electricity
Hardware
People Software
ROI
TCO
Planned
maintenance
Expansion
costs
Controller
upgrades
Drive
replacement
End-of-life
refresh
Storage
Management
Overhead
Optimize on-premises
clouds’ TCO and ROI
through data storage
Software
Licensing
Infrastructure
Management
Overhead
Costly & Complex
Datacenter Design
Compute
Servers
StorPool
Storage
Dual-Controller
Shared-Disk
Arrays
5-YEAR TCO
AT SCALE
Considering Scope1
TCO Factors
52%
SAVED
StorPool
Storage
Dual-Controller
Shared-Disk
Arrays
5-YEAR TCO
AT SCALE
Considering Scope 2
TCO Factors
59%
SAVED
StorPool
Storage
Dual-Controller
Shared-Disk
Arrays
5-YEAR TCO
AT SCALE
Considering Scope 3
TCO Factors
60%
SAVED
WE NEED TO CHANGE THE WAY WE
THINK ABOUT DATA STORAGE
@StorPool
StorPool.com StorPool Storage
Tech Excellence
Boyan Krosnov, CTO
AGENDA
Introduction to
StorPool Storage
Boyan Ivanov, CEO
Tech Excellence
Boyan Krosnov, CTO
Helping IT Teams
Do Storage Right
Alex Ivanov, Product Lead
The Old Way:
Dual-Controller
Shared-Disk
Arrays
Base array
Disk shelf
Disk shelf
HOW DO SHARED-DISK ARRAYS
FIT INTO THE TYPICAL DATACENTER
ENVIRONMENT?
ToR
SWITCH
SPINE
SWITCH
Slow adoption
of innovations
X
High-cost system
lifecycle
X
Multi-rack availability
requirements
X
Bottlenecks at
the controllers
X
Always-on
operating model
X
Datacenter-scale
Ethernet network
X
Building block
concept
X
BUT WHY IS THE OLD MODEL
NOT WORKING?
Hard to manage
at scale
X
Uses commodity
servers
Fits datacenter
designs
Fits customer
workflows
Addresses demanding
applications
Extracts near-100%
of HW potential
Has high-speed
data efficiency
Scales linearly Fades into the
background
THE IDEAL PRIMARY
STORAGE PLATFORM
RADICAL DEPARTURE
StorPool Storage
IS THAT
RADICAL DEPARTURE
StorPool Storage
(up to 4,032 nodes)
Network – 2x/4x 10-100 GbE
per storage node
Storage Node Storage Node Storage Node
Parallel NVMe/TCP Service
Parallel iSCSI Service
Highly Available NFS Service
Server Server
SSD
Server
SSD
SSD
SSD
SSD
SSD
SSD
SSD
SSD
iSCSI,
NFS
NVMe/TCP,
iSCSI, NFS
NVMe/TCP,
iSCSI, SP Block,
NFS
NVMe/TCP, iSCSI
SP Block, NFS
VMware Hyper-V
block
block
block
KVM Hypervisor
StorPool Client
block device driver
block
block
block
Linux Bare-Metal
StorPool Client
block device driver
StorPool Storage
Metrics & Analytics Alerts & Reporting VolumeCare CMP Integrations
Parallelism and
shared nothing
architecture
Exceptional latency
and throughput
Up to 64 sub-clusters
managed as a unified
large-scale system
Linearly scalable large
sub-clusters of capacity
& performance
Datacenter-scale
deployments
Data management,
incl. backup and DR
StorPool Storage
ARCHITECTURE AND
HOW IT WORKS
StorPool Framework
upstream drivers:
- block device
- vhost
hardware threads, tasks,
scheduling, loader
memory management huge
pages, shm, allocator
network protocols:
- Ethernet
- TCP/IP (ARP, ICMP, TCP) & BGP
downstream drivers:
- pci device
- nvme, linux block
- ixgbe, i40e, mlx4, mlx5, bnx2x,
bnxt, virtio-net, linux net
- rdma
- ioat
- hsleep, ksleep
Linux,
KVM
integrations
StorPool Protocol
on-disk
format
Block Client NVMe/TCP
iSCSI
Bridge mgmt
Server
Beacon
service discovery
service monitoring
quorum
PMem
NVMe
SATA/SAS
VMware
Hyper-V
Oracle
LVM
CLI
GUI
ARCHITECTURE
DEEP DIVE
LINEARLY SCALABLE CAPACITY
AND PERFORMANCE
453
823
773
1,686
75
106
71
106
4k qd 7x64 70/30[kIOPS]
16k qd 7x64 70/30[kIOPS]
2,000
1,500
1,000
500
0
3 servers,
12 NVMes
4 servers,
16 NVMes
5 servers,
20 NVMes
6 servers,
24 NVMes
7 servers,
28 NVMes
3 servers,
12 NVMes
4 servers,
16 NVMes
5 servers,
20 NVMes
6 servers,
24 NVMes
7 servers,
28 NVMes
4k qd 7x1 100% read[us]
4k qd 7x1 100% write[us]
100
75
50
25
0
125
NVMe SSD
KVM Virtual Machine
KVM Virtual Machine
NVMe SSD
NVMe SSD
NVMe SSD
NVMe SSD
NVMe SSD
NIC
server instance
1CPU thread
2-4 GB RAM
server instance
1CPU thread
2-4 GB RAM
server instance
1CPU thread
2-4 G RAM
block instance
1CPU thread
1 GB RAM
25GbE
25GbE
Kernel
bypass
Kernel
bypass
DATA
PROTECTION
Storage Node
SSD SSD
SSD SSD
Client
Volume/LUN
Storage Node
SSD SSD
SSD SSD
Storage Node
SSD SSD
SSD SSD
Storage Node
SSD SSD
SSD SSD
Storage Node
SSD SSD
SSD SSD
Distributed
storage and
processing
Synchronous
replication across
storage nodes
PERFORMANCE CONSISTENCY,
FOR REAL!
Average 95th 99th 99.9th
Read latency 133µs 224µs 358µs 498µs
Write latency 66µs 81µs 99µs 293µs
NEW APP ADDED:
Transactional
workload with
4KB read/write
SYSTEM LOAD:
50,000 IOPS
per TB Usable
4KB 70/30
read/write
500,000
IOPS total
MULTI-RACK SETUP
StorPool Storage can be deployed “horizontally” in multiple racks
StorPool Cluster
HARDWARE REFRESH CYCLE
HARDWARE REFRESH CYCLE
Without service
interruption
Replace servers
one-by-one
No maintenance
windows
No complex data
migrations
Integrations:
Kubernetes CSI, OpenStack Cinder,
CloudStack, OpenNebula, OnApp
StorPool VolumeCare
(enables snapshot-based
backup and DR services)
StorPool API
Kubernetes
OpenStack
CloudStack
OpenNebula
OnApp
Bespoke/custom
integrations
INTEGRATIONS & AUTOMATIONS
SCOPE OF MANAGED SERVICES
Continuous Service Improvement
DESIGN DEPLOY FINE-TUNE
MONITOR
MAINTAIN
WHAT ENABLES OUR MANAGED
SERVICES AT SCALE
Мetrics, analytics Еvents and alarms,
multi-channel alerting
Аutomation
tooling
Unification of
everything possible Cross-immunization
Hyperscaler-like approach
to infrastructure
LESS HASSLE &
FEWER PEOPLE
INVOLVED
Higher overall cloud
efficiency
Higher performance
Higher density
Fewer servers
=>
=>
=>
&
60% more
applications per rack
Fewer compute
servers needed for
applications
60% less electricity
consumption
They're paying less $ even though electricity prices went up!
USA electricity prices are 2x higher.
Parallel, always-on, no silos, automated.
No toil, no hassle.
@StorPool
StorPool.com StorPool Storage
Helping IT Teams
Do Storage Right
Alex Ivanov, Product Lead
AGENDA
Introduction to
StorPool Storage
Boyan Ivanov, CEO
Tech Excellence
Boyan Krosnov, CTO
Helping IT Teams
Do Storage Right
Alex Ivanov, Product Lead
“The hyperscalers have long proven
that commodity hardware is good
enough to run industrial-grade
workloads. The trick is in applying
the right software layer to get the
most out of the available hardware.”
StorPool Storage
Metrics & Analytics Alerts & Reporting VolumeCare CMP Integrations
Dynamic provisioning
Self-service
Parallel service for 1000s of workloads
Always-on
99.999% with maintenance included
Consistent low latency
Linear scalability (no data silos)
No single point of failure
Issues with performance
consistency
Infrastructure
sprawl
ENTER THE STORAGE GAME
Complexity Endless cycles
High-cost
workarounds
HARD & EXPENSIVE SYSTEM LIFECYCLE MANAGEMENT
License tied to
specific hardware
Systems rarely add
performance during
lifecycle
Modernize, and you
need a new license
Time-consuming HW
modernization
Fixed overhead
Perfect for prophets
Pay-as-you-go?
Typically a fixed
monthly cost
Hiked MSRPs
Enable fictional
60% - 90% discounts
Support and drives
Tagged as % of MSRP
ZOOMING INTO THE
STORAGE GAME
VENDORS NEED ROI FOR EACH
SYSTEM THEY DEVELOP
WHY?
THE IMPORTANCE OF BLOCK
STORAGE PERFORMANCE
Slower block
storage
performance
Slower
application
response
times
Reduced
user
productivity
Delayed
time-to-
market
Reduced
revenues
and margins
PRODUCTIVITY PLUMMETS WITH
SLOW APP RESPONSE TIMES
80%+
for experts
75%+
for average
95%+
for novices
Transaction
rate
4,000
3,500
3,000
2,500
2,000
1,500
1,000
500
0
0.25 0.5 0,75 1.0 1.25 1.5
Expert
Average
Novice
Application response time (seconds)
SLOW APP RESPONSE TIMES COST
MONEY, ESPECIALLY AT SCALE
$150k
per month
for 50 users
$300k
per month
for 100 users
$9 million
per month
for 3,000 users
1 2 3
Potential monthly savings from rapid application response
for varying numbers of simultaneous users
10
9
8
7
6
5
4
3
2
1
0
0
300
Users
200
Users
100
Users
50
Users
X
$100,000
per
month
Application response time (seconds)
IMPACT OF FASTER APP RESPONSE
TIMES AT SCALE?
Faster
response
times
More
productivity
More
savings
1 2 3
Potential monthly savings from rapid application response
for varying numbers of simultaneous users
10
9
8
7
6
5
4
3
2
1
0
0
300
Users
200
Users
100
Users
50
Users
X
$100,000
per
month
Application response time (seconds)
CULPRITS FOR BLOCK STORAGE
PERFORMANCE DEGRADATION
Usable capacity
utilization ≥ 50%
Storage drive failures,
large drive rebuilds
Data migration
between data silos
Inefficient storage
software stack
Cleanup processes
Synchronous
replication
Write penalties due
to write algorithms
Network/system
interconnect mismatch
TRADITIONAL APP RESPONSE TIME
WORKAROUNDS
Faster, bigger direct-attach storage
SSDs, NVMe SSDs, Persistent Memory, etc.
Faster, better server-storage interconnect
Higher bandwidth, NVMe-oF
Application splitting & database sharding
On-prem or in the cloud
More, faster, newer storage systems
Come with faster drives and interconnect
At large scale? Not so much...
Easy to do at small scale
Storage, servers, software, network, racks, etc.
Leads to infrastructure sprawl
Leads to orphaned storage, a.k.a. siloed storage
App instances require separate I/O
Splitting, patching, troubleshooting,
updating, load balancing...
Waste costs across the board
TRADITIONAL APP RESPONSE TIME
WORKAROUNDS LEAD TO
EXCESSIVE 2X TO 3X MORE
EXPENSIVE INFRASTRUCTURE TCO!
MORE, FASTER, NEWER STORAGE
ARRAYS HELP, TO A POINT
Faster and bigger
SSDs
More arrays enable
more app splitting &
DB sharding
Better overall
performance @
premium cost
Some scale-out
options with active-
active pairs
BUT, WHAT HAPPENS AT SCALE,
ESPECIALLY AT LARGE SCALE?
WHEN YOU HAVE TENS OR
HUNDREDS OF ARRAYS?
USABLE AND PERFORMANCE CAPACITY CHAOS
Continuous workload
monitoring
Constant rebalancing
between data silos
Performance
degradation due to
capacity consumption
Labor-intensive data
migrations for system
lifecycle management
Why? Vendors need ROI for each system they develop.
IN SHORT, ENTER THE STORAGE GAME
HARD & EXPENSIVE SYSTEM LIFECYCLE MANAGEMENT
License tied to
specific hardware
Systems rarely add
performance during
lifecycle
Modernize, and you
need a new license
Time-consuming HW
modernization
Fixed overhead
Perfect for prophets
Pay-as-you-go?
Typically a fixed
monthly cost
Hiked MSRPs
Enable fictional
60%-90% discounts
Support and drives
Tagged as % of MSRP
ENDING THE STORAGE GAME
Variable cost that
follows revenue
Easy for business
Pay-per-use
Pricing aligns with
end-user offerings
No pricing fugazis
With usage-based volume
discounts
24/7 support
Included with the StorPool
license
EASY, NON-DISRUPTIVE SYSTEM LIFECYCLE MANAGEMENT
Why?
Systems built with cheap
read-intensive
storage drives
Adding drives or servers
scales capacity and
performance
Modernize as vendors
introduce new server
generations
Add new hardware
generations
non-disruptively
WOULDN'T BE POSSIBLE WITHOUT...
Consistent
performance
even under
massively
parallel I/O
Easy
workload
consolidation
with less
infrastructure
RUNNING WORKLOADS FOR
CUSTOMERS
“The StorPool product is perfect from cost,
flexibility (grow-as-you-go), and performance
point of view. Second - and for us equally
important - is the way the StorPool team
helps us to build the most advanced hosting
platform and their fast and committed
support in case of an issue."
Hink Wiersema
Product Manager
and Architect
ATOS VIRTUAL ORACLE COMPUTING HOTEL
LAUNCHED IN 2013
Hinks' goals for what
should be offered to
end customers
Flexibility
Decrease
Oracle
licensing
Low Total
Cost of
Ownership
Running
workloads
for
Banks
Insurance
companies
Hospitals
In the
Netherlands
HARD & EXPENSIVE SYSTEM LIFECYCLE MANAGEMENT
License tied to
specific hardware
Systems rarely add
performance during
lifecycle
Modernize, and you
need a new license
Time-consuming HW
modernization
Fixed overhead
Perfect for prophets
Pay-as-you-go?
Typically a fixed
monthly cost
Hiked MSRPs
Enable fictional
60%-90% discounts
Support and drives
Tagged as % of MSRP
HINK‘S GOALS
Flexibility Low Total Cost
of Ownership
Decrease
Oracle licensing
DIFFICULT SYSTEM LIFECYCLE MANAGEMENT
License tied to
specific hardware
Systems rarely add
performance during
lifecycle
Modernize, and you
need a new license
Modernize to
next-gen HW?
Migration
Fixed Overhead
Perfect for prophets
Pay-as-you-go?
Typically a fixed
monthly cost
Hiked MSRPs
Enable fictional
60%-90% discounts
Support and Drives
Tagged as % of MSRP
HINK‘S GOALS
Flexibility Low Total Cost
of Ownership
Decrease
Oracle licensing
CRASHED STRAIGHT INTO THE STORAGE GAME
WORKING WITH HIS TEAM,
HE GOT TO A POINT WHERE HE FACED
5 SIGNIFICANT CHALLENGES
Maintain service
quality and
optimize prices
Align his storage
pricing with
customers’ pay-
per-use model
Remove
maintenance
windows for HW
and SW changes
Stay competitive
by continuously
modernizing
HW components
Optimize the
VOC hotel
lifecycle and cost
management
StorPool HAD A SOLUTION
FOR ALL OF THAT
Maintain service quality
and optimize prices
99.999% availability
with maintenance included
Align his storage pricing with
customers' pay-per-use model
Pay-per-use with
no cost surprises
Remove maintenance windows
for HW and SW changes
Easy, non-disruptive system
lifecycle management
Stay competitive by continuously
modernizing HW components
Easy adoption of performance and capacity
innovations non-disruptively, at any time
Optimize the VOC hotel lifecycle
and cost management
Systems built with COTS hardware
that easily support his environment
BUT THAT CREATED
A NEW CHALLENGE FOR HINK...
…HE HAD TO ADDRESS ALL THE
"WHY StorPool"
QUESTIONS NEXT…
А NEW MINDSET
FOR RUNNING THE ATOS VOC HOTEL
NEW
Datacenter
design
Always-online
operating model
Way of working -
during normal work
hours, no more 2am
and weekend work
Backend billing model
(OpEx, not CapEx)
Ways to track and
manage service P&L
Way of thinking
about hardware
modernization
(add anytime
approach)
Primary storage
partner
Atos team simply
creates and
manages volumes
Way of working with a
solution provider:
Single data pool
that is always
available
Hyperscaler-like
Principles for
deploying, managing,
monitoring, and
maintaining block
storage
AFTER PUSHING ALL THAT THROUGH,
THERE WAS A FINAL HURDLE:
THE FIRST UPDATE…
THE USUAL ORDER OF OPERATIONS
Notify all your customers that you're doing an update
Plan maintenance downtime for workloads
Plan this for a Saturday or Sunday, or during the night
Have all your people on deck
Be ready to fix everything in case things go wrong
Be ready to roll back in emergencies
WEDNESDAY
AT 18:00
THE UPDATE ROLLED THROUGH,
NODE BY NODE
“The StorPool product is perfect from cost,
flexibility (grow-as-you-go), and performance
point of view. Second - and for us equally
important - is the way the StorPool team
helps us to build the most advanced hosting
platform and their fast and committed
support in case of an issue."
Hink Wiersema
Product Manager
and Architect
WHAT ARE THE OPTIONS?
KEEP WASTING INFRASTRUCTURE, TIME,
PERSONNEL, RESOURCES & MONEY
LET StorPool HELP YOU
EXIT THE STORAGE GAME
ONCE AND FOR ALL
@StorPool
StorPool.com StorPool Storage

More Related Content

Similar to StorPool Storage presenting at Storage Field Day 25pdf

Why Software Defined Storage is Critical for Your IT Strategy
Why Software Defined Storage is Critical for Your IT StrategyWhy Software Defined Storage is Critical for Your IT Strategy
Why Software Defined Storage is Critical for Your IT Strategy
andreas kuncoro
 
Storage Cloud and Spectrum deck March 2016
Storage Cloud and Spectrum deck March 2016Storage Cloud and Spectrum deck March 2016
Storage Cloud and Spectrum deck March 2016
Joe Krotz
 
Emc isilon overview
Emc isilon overview Emc isilon overview
Emc isilon overview
solarisyougood
 
Qnap event v1.6
Qnap   event v1.6Qnap   event v1.6
Qnap event v1.6
Amir Ghorbanali
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
webhostingguy
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
webhostingguy
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
webhostingguy
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
webhostingguy
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
webhostingguy
 
VirtualStor Extreme - Software Defined Scale-Out All Flash Storage
VirtualStor Extreme - Software Defined Scale-Out All Flash StorageVirtualStor Extreme - Software Defined Scale-Out All Flash Storage
VirtualStor Extreme - Software Defined Scale-Out All Flash Storage
GIGABYTE Technology
 
TechTarget Event - Storage Architectures for the Modern Data Centre – Martin ...
TechTarget Event - Storage Architectures for the Modern Data Centre – Martin ...TechTarget Event - Storage Architectures for the Modern Data Centre – Martin ...
TechTarget Event - Storage Architectures for the Modern Data Centre – Martin ...
NetApp
 
MT41 Dell EMC VMAX: Ask the Experts
MT41 Dell EMC VMAX: Ask the Experts MT41 Dell EMC VMAX: Ask the Experts
MT41 Dell EMC VMAX: Ask the Experts
Dell EMC World
 
HPC DAY 2017 | HPE Storage and Data Management for Big Data
HPC DAY 2017 | HPE Storage and Data Management for Big DataHPC DAY 2017 | HPE Storage and Data Management for Big Data
HPC DAY 2017 | HPE Storage and Data Management for Big Data
HPC DAY
 
High Performance Communication for Oracle using InfiniBand
High Performance Communication for Oracle using InfiniBandHigh Performance Communication for Oracle using InfiniBand
High Performance Communication for Oracle using InfiniBand
webhostingguy
 
Oracle Exec Summary 7000 Unified Storage
Oracle Exec Summary 7000 Unified StorageOracle Exec Summary 7000 Unified Storage
Oracle Exec Summary 7000 Unified Storage
David R. Klauser
 
Accelerating Ceph Performance with High Speed Networks and Protocols - Qingch...
Accelerating Ceph Performance with High Speed Networks and Protocols - Qingch...Accelerating Ceph Performance with High Speed Networks and Protocols - Qingch...
Accelerating Ceph Performance with High Speed Networks and Protocols - Qingch...
Ceph Community
 
HP: HP 3PAR - Storage zrodený pre virtualizované prostredie
HP: HP 3PAR - Storage zrodený pre virtualizované prostredieHP: HP 3PAR - Storage zrodený pre virtualizované prostredie
HP: HP 3PAR - Storage zrodený pre virtualizované prostredie
ASBIS SK
 
3PAR and VMWare
3PAR and VMWare3PAR and VMWare
3PAR and VMWare
vmug
 
Make IT Simple, Make Business Agile
Make IT Simple, Make Business AgileMake IT Simple, Make Business Agile
Make IT Simple, Make Business Agile
Huawei Enterprise Hong Kong
 
T12.Fujitsu World Tour India 2016-Your Datacenter‘s backbone
T12.Fujitsu World Tour India 2016-Your Datacenter‘s backboneT12.Fujitsu World Tour India 2016-Your Datacenter‘s backbone
T12.Fujitsu World Tour India 2016-Your Datacenter‘s backbone
Fujitsu India
 

Similar to StorPool Storage presenting at Storage Field Day 25pdf (20)

Why Software Defined Storage is Critical for Your IT Strategy
Why Software Defined Storage is Critical for Your IT StrategyWhy Software Defined Storage is Critical for Your IT Strategy
Why Software Defined Storage is Critical for Your IT Strategy
 
Storage Cloud and Spectrum deck March 2016
Storage Cloud and Spectrum deck March 2016Storage Cloud and Spectrum deck March 2016
Storage Cloud and Spectrum deck March 2016
 
Emc isilon overview
Emc isilon overview Emc isilon overview
Emc isilon overview
 
Qnap event v1.6
Qnap   event v1.6Qnap   event v1.6
Qnap event v1.6
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
 
Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...Virtualization: Doing it right the first time to avoid costly ...
Virtualization: Doing it right the first time to avoid costly ...
 
VirtualStor Extreme - Software Defined Scale-Out All Flash Storage
VirtualStor Extreme - Software Defined Scale-Out All Flash StorageVirtualStor Extreme - Software Defined Scale-Out All Flash Storage
VirtualStor Extreme - Software Defined Scale-Out All Flash Storage
 
TechTarget Event - Storage Architectures for the Modern Data Centre – Martin ...
TechTarget Event - Storage Architectures for the Modern Data Centre – Martin ...TechTarget Event - Storage Architectures for the Modern Data Centre – Martin ...
TechTarget Event - Storage Architectures for the Modern Data Centre – Martin ...
 
MT41 Dell EMC VMAX: Ask the Experts
MT41 Dell EMC VMAX: Ask the Experts MT41 Dell EMC VMAX: Ask the Experts
MT41 Dell EMC VMAX: Ask the Experts
 
HPC DAY 2017 | HPE Storage and Data Management for Big Data
HPC DAY 2017 | HPE Storage and Data Management for Big DataHPC DAY 2017 | HPE Storage and Data Management for Big Data
HPC DAY 2017 | HPE Storage and Data Management for Big Data
 
High Performance Communication for Oracle using InfiniBand
High Performance Communication for Oracle using InfiniBandHigh Performance Communication for Oracle using InfiniBand
High Performance Communication for Oracle using InfiniBand
 
Oracle Exec Summary 7000 Unified Storage
Oracle Exec Summary 7000 Unified StorageOracle Exec Summary 7000 Unified Storage
Oracle Exec Summary 7000 Unified Storage
 
Accelerating Ceph Performance with High Speed Networks and Protocols - Qingch...
Accelerating Ceph Performance with High Speed Networks and Protocols - Qingch...Accelerating Ceph Performance with High Speed Networks and Protocols - Qingch...
Accelerating Ceph Performance with High Speed Networks and Protocols - Qingch...
 
HP: HP 3PAR - Storage zrodený pre virtualizované prostredie
HP: HP 3PAR - Storage zrodený pre virtualizované prostredieHP: HP 3PAR - Storage zrodený pre virtualizované prostredie
HP: HP 3PAR - Storage zrodený pre virtualizované prostredie
 
3PAR and VMWare
3PAR and VMWare3PAR and VMWare
3PAR and VMWare
 
Make IT Simple, Make Business Agile
Make IT Simple, Make Business AgileMake IT Simple, Make Business Agile
Make IT Simple, Make Business Agile
 
T12.Fujitsu World Tour India 2016-Your Datacenter‘s backbone
T12.Fujitsu World Tour India 2016-Your Datacenter‘s backboneT12.Fujitsu World Tour India 2016-Your Datacenter‘s backbone
T12.Fujitsu World Tour India 2016-Your Datacenter‘s backbone
 

More from StorPool Storage

StorPool Presents at Cloud Field Day 9
StorPool Presents at Cloud Field Day 9StorPool Presents at Cloud Field Day 9
StorPool Presents at Cloud Field Day 9
StorPool Storage
 
Achieving the ultimate performance with KVM
Achieving the ultimate performance with KVMAchieving the ultimate performance with KVM
Achieving the ultimate performance with KVM
StorPool Storage
 
Public Cloud Performance Measurement Report
Public Cloud Performance Measurement ReportPublic Cloud Performance Measurement Report
Public Cloud Performance Measurement Report
StorPool Storage
 
StorPool Demo Day Presentations
StorPool Demo Day PresentationsStorPool Demo Day Presentations
StorPool Demo Day Presentations
StorPool Storage
 
Optimization_of_Virtual_Machines_for_High_Performance
Optimization_of_Virtual_Machines_for_High_PerformanceOptimization_of_Virtual_Machines_for_High_Performance
Optimization_of_Virtual_Machines_for_High_Performance
StorPool Storage
 
Webinar: StorPool and WHIR - better storage, better business
Webinar: StorPool and WHIR - better storage, better businessWebinar: StorPool and WHIR - better storage, better business
Webinar: StorPool and WHIR - better storage, better business
StorPool Storage
 

More from StorPool Storage (6)

StorPool Presents at Cloud Field Day 9
StorPool Presents at Cloud Field Day 9StorPool Presents at Cloud Field Day 9
StorPool Presents at Cloud Field Day 9
 
Achieving the ultimate performance with KVM
Achieving the ultimate performance with KVMAchieving the ultimate performance with KVM
Achieving the ultimate performance with KVM
 
Public Cloud Performance Measurement Report
Public Cloud Performance Measurement ReportPublic Cloud Performance Measurement Report
Public Cloud Performance Measurement Report
 
StorPool Demo Day Presentations
StorPool Demo Day PresentationsStorPool Demo Day Presentations
StorPool Demo Day Presentations
 
Optimization_of_Virtual_Machines_for_High_Performance
Optimization_of_Virtual_Machines_for_High_PerformanceOptimization_of_Virtual_Machines_for_High_Performance
Optimization_of_Virtual_Machines_for_High_Performance
 
Webinar: StorPool and WHIR - better storage, better business
Webinar: StorPool and WHIR - better storage, better businessWebinar: StorPool and WHIR - better storage, better business
Webinar: StorPool and WHIR - better storage, better business
 

Recently uploaded

Trusted Execution Environment for Decentralized Process Mining
Trusted Execution Environment for Decentralized Process MiningTrusted Execution Environment for Decentralized Process Mining
Trusted Execution Environment for Decentralized Process Mining
LucaBarbaro3
 
Operating System Used by Users in day-to-day life.pptx
Operating System Used by Users in day-to-day life.pptxOperating System Used by Users in day-to-day life.pptx
Operating System Used by Users in day-to-day life.pptx
Pravash Chandra Das
 
Azure API Management to expose backend services securely
Azure API Management to expose backend services securelyAzure API Management to expose backend services securely
Azure API Management to expose backend services securely
Dinusha Kumarasiri
 
5th LF Energy Power Grid Model Meet-up Slides
5th LF Energy Power Grid Model Meet-up Slides5th LF Energy Power Grid Model Meet-up Slides
5th LF Energy Power Grid Model Meet-up Slides
DanBrown980551
 
Generating privacy-protected synthetic data using Secludy and Milvus
Generating privacy-protected synthetic data using Secludy and MilvusGenerating privacy-protected synthetic data using Secludy and Milvus
Generating privacy-protected synthetic data using Secludy and Milvus
Zilliz
 
Letter and Document Automation for Bonterra Impact Management (fka Social Sol...
Letter and Document Automation for Bonterra Impact Management (fka Social Sol...Letter and Document Automation for Bonterra Impact Management (fka Social Sol...
Letter and Document Automation for Bonterra Impact Management (fka Social Sol...
Jeffrey Haguewood
 
Recommendation System using RAG Architecture
Recommendation System using RAG ArchitectureRecommendation System using RAG Architecture
Recommendation System using RAG Architecture
fredae14
 
Presentation of the OECD Artificial Intelligence Review of Germany
Presentation of the OECD Artificial Intelligence Review of GermanyPresentation of the OECD Artificial Intelligence Review of Germany
Presentation of the OECD Artificial Intelligence Review of Germany
innovationoecd
 
GraphRAG for Life Science to increase LLM accuracy
GraphRAG for Life Science to increase LLM accuracyGraphRAG for Life Science to increase LLM accuracy
GraphRAG for Life Science to increase LLM accuracy
Tomaz Bratanic
 
Salesforce Integration for Bonterra Impact Management (fka Social Solutions A...
Salesforce Integration for Bonterra Impact Management (fka Social Solutions A...Salesforce Integration for Bonterra Impact Management (fka Social Solutions A...
Salesforce Integration for Bonterra Impact Management (fka Social Solutions A...
Jeffrey Haguewood
 
Energy Efficient Video Encoding for Cloud and Edge Computing Instances
Energy Efficient Video Encoding for Cloud and Edge Computing InstancesEnergy Efficient Video Encoding for Cloud and Edge Computing Instances
Energy Efficient Video Encoding for Cloud and Edge Computing Instances
Alpen-Adria-Universität
 
Overcoming the PLG Trap: Lessons from Canva's Head of Sales & Head of EMEA Da...
Overcoming the PLG Trap: Lessons from Canva's Head of Sales & Head of EMEA Da...Overcoming the PLG Trap: Lessons from Canva's Head of Sales & Head of EMEA Da...
Overcoming the PLG Trap: Lessons from Canva's Head of Sales & Head of EMEA Da...
saastr
 
AWS Cloud Cost Optimization Presentation.pptx
AWS Cloud Cost Optimization Presentation.pptxAWS Cloud Cost Optimization Presentation.pptx
AWS Cloud Cost Optimization Presentation.pptx
HarisZaheer8
 
Programming Foundation Models with DSPy - Meetup Slides
Programming Foundation Models with DSPy - Meetup SlidesProgramming Foundation Models with DSPy - Meetup Slides
Programming Foundation Models with DSPy - Meetup Slides
Zilliz
 
Artificial Intelligence for XMLDevelopment
Artificial Intelligence for XMLDevelopmentArtificial Intelligence for XMLDevelopment
Artificial Intelligence for XMLDevelopment
Octavian Nadolu
 
A Comprehensive Guide to DeFi Development Services in 2024
A Comprehensive Guide to DeFi Development Services in 2024A Comprehensive Guide to DeFi Development Services in 2024
A Comprehensive Guide to DeFi Development Services in 2024
Intelisync
 
GenAI Pilot Implementation in the organizations
GenAI Pilot Implementation in the organizationsGenAI Pilot Implementation in the organizations
GenAI Pilot Implementation in the organizations
kumardaparthi1024
 
Monitoring and Managing Anomaly Detection on OpenShift.pdf
Monitoring and Managing Anomaly Detection on OpenShift.pdfMonitoring and Managing Anomaly Detection on OpenShift.pdf
Monitoring and Managing Anomaly Detection on OpenShift.pdf
Tosin Akinosho
 
HCL Notes and Domino License Cost Reduction in the World of DLAU
HCL Notes and Domino License Cost Reduction in the World of DLAUHCL Notes and Domino License Cost Reduction in the World of DLAU
HCL Notes and Domino License Cost Reduction in the World of DLAU
panagenda
 
Serial Arm Control in Real Time Presentation
Serial Arm Control in Real Time PresentationSerial Arm Control in Real Time Presentation
Serial Arm Control in Real Time Presentation
tolgahangng
 

Recently uploaded (20)

Trusted Execution Environment for Decentralized Process Mining
Trusted Execution Environment for Decentralized Process MiningTrusted Execution Environment for Decentralized Process Mining
Trusted Execution Environment for Decentralized Process Mining
 
Operating System Used by Users in day-to-day life.pptx
Operating System Used by Users in day-to-day life.pptxOperating System Used by Users in day-to-day life.pptx
Operating System Used by Users in day-to-day life.pptx
 
Azure API Management to expose backend services securely
Azure API Management to expose backend services securelyAzure API Management to expose backend services securely
Azure API Management to expose backend services securely
 
5th LF Energy Power Grid Model Meet-up Slides
5th LF Energy Power Grid Model Meet-up Slides5th LF Energy Power Grid Model Meet-up Slides
5th LF Energy Power Grid Model Meet-up Slides
 
Generating privacy-protected synthetic data using Secludy and Milvus
Generating privacy-protected synthetic data using Secludy and MilvusGenerating privacy-protected synthetic data using Secludy and Milvus
Generating privacy-protected synthetic data using Secludy and Milvus
 
Letter and Document Automation for Bonterra Impact Management (fka Social Sol...
Letter and Document Automation for Bonterra Impact Management (fka Social Sol...Letter and Document Automation for Bonterra Impact Management (fka Social Sol...
Letter and Document Automation for Bonterra Impact Management (fka Social Sol...
 
Recommendation System using RAG Architecture
Recommendation System using RAG ArchitectureRecommendation System using RAG Architecture
Recommendation System using RAG Architecture
 
Presentation of the OECD Artificial Intelligence Review of Germany
Presentation of the OECD Artificial Intelligence Review of GermanyPresentation of the OECD Artificial Intelligence Review of Germany
Presentation of the OECD Artificial Intelligence Review of Germany
 
GraphRAG for Life Science to increase LLM accuracy
GraphRAG for Life Science to increase LLM accuracyGraphRAG for Life Science to increase LLM accuracy
GraphRAG for Life Science to increase LLM accuracy
 
Salesforce Integration for Bonterra Impact Management (fka Social Solutions A...
Salesforce Integration for Bonterra Impact Management (fka Social Solutions A...Salesforce Integration for Bonterra Impact Management (fka Social Solutions A...
Salesforce Integration for Bonterra Impact Management (fka Social Solutions A...
 
Energy Efficient Video Encoding for Cloud and Edge Computing Instances
Energy Efficient Video Encoding for Cloud and Edge Computing InstancesEnergy Efficient Video Encoding for Cloud and Edge Computing Instances
Energy Efficient Video Encoding for Cloud and Edge Computing Instances
 
Overcoming the PLG Trap: Lessons from Canva's Head of Sales & Head of EMEA Da...
Overcoming the PLG Trap: Lessons from Canva's Head of Sales & Head of EMEA Da...Overcoming the PLG Trap: Lessons from Canva's Head of Sales & Head of EMEA Da...
Overcoming the PLG Trap: Lessons from Canva's Head of Sales & Head of EMEA Da...
 
AWS Cloud Cost Optimization Presentation.pptx
AWS Cloud Cost Optimization Presentation.pptxAWS Cloud Cost Optimization Presentation.pptx
AWS Cloud Cost Optimization Presentation.pptx
 
Programming Foundation Models with DSPy - Meetup Slides
Programming Foundation Models with DSPy - Meetup SlidesProgramming Foundation Models with DSPy - Meetup Slides
Programming Foundation Models with DSPy - Meetup Slides
 
Artificial Intelligence for XMLDevelopment
Artificial Intelligence for XMLDevelopmentArtificial Intelligence for XMLDevelopment
Artificial Intelligence for XMLDevelopment
 
A Comprehensive Guide to DeFi Development Services in 2024
A Comprehensive Guide to DeFi Development Services in 2024A Comprehensive Guide to DeFi Development Services in 2024
A Comprehensive Guide to DeFi Development Services in 2024
 
GenAI Pilot Implementation in the organizations
GenAI Pilot Implementation in the organizationsGenAI Pilot Implementation in the organizations
GenAI Pilot Implementation in the organizations
 
Monitoring and Managing Anomaly Detection on OpenShift.pdf
Monitoring and Managing Anomaly Detection on OpenShift.pdfMonitoring and Managing Anomaly Detection on OpenShift.pdf
Monitoring and Managing Anomaly Detection on OpenShift.pdf
 
HCL Notes and Domino License Cost Reduction in the World of DLAU
HCL Notes and Domino License Cost Reduction in the World of DLAUHCL Notes and Domino License Cost Reduction in the World of DLAU
HCL Notes and Domino License Cost Reduction in the World of DLAU
 
Serial Arm Control in Real Time Presentation
Serial Arm Control in Real Time PresentationSerial Arm Control in Real Time Presentation
Serial Arm Control in Real Time Presentation
 

StorPool Storage presenting at Storage Field Day 25pdf

  • 1.
  • 2. Introduction to StorPool Storage Boyan Ivanov, CEO
  • 4. 1980s and on Client-Server with Local Storage
  • 5. Shared-Disk Storage Software 1990s and on Dual-Controller Shared-Disk Arrays
  • 6. 2000s and on Shared-Disk Storage Software 1990s and on
  • 8. SIMPLE & PREDICTABLE 1-3 RACKS THEY WORK…
  • 9. Dynamic provisioning ? Self-service ? Parallel service for 1,000s of workloads ? Always-on ? 99.999% with maint. included ? No single point of failure ? Consistent low latency ? Linear Scalability ? BUT WHAT IF YOU NEED
  • 11. 200+ compute servers 4,000+ 10+ racks 200+ 1PB+ saved data 20 PB+ 5,000+ 100,000+ VMs АТ LARGE SCALE = SCALE × 20
  • 13. People are simply unhappy No automation, ticket-based IT gets in the way of “the business” Slow to address changes in market dynamics Squeezed profit margins due to org & tech debt ROAD TO
  • 16. AGENDA Introduction to StorPool Storage Boyan Ivanov, CEO Tech Excellence Boyan Krosnov, CTO Helping IT Teams Do Storage Right Alex Ivanov, Product Lead
  • 17. StorPool IN A NUTSHELL Of Flash Storage Deployments 100s Countries of Deployment 30+ Profitable & Growing
  • 19. Dynamic provisioning Self-service Parallel service for 1000s of workloads Always-on 99.999% with maintenance included Consistent low latency Linear scalability (no data silos) No single point of failure
  • 20. HOW DID WE MAKE THIS? Dynamic provisioning Self-service Parallel service for 1000s of workloads Always-on 99.999% with maintenance included Consistent low latency Linear scalability (no data silos) No single point of failure
  • 21. SILICON VALLEY OF THE EASTERN BLOCK
  • 22. KNOW-HOW FROM A DIFFERENT DOMAIN High Performance Networking IT Service Providers
  • 24. StorPool Storage Metrics & Analytics Alerts & Reporting VolumeCare CMP Integrations
  • 25. FULLY MANAGED SOLUTION Parallel multi-node shared-nothing architecture Always-on: Non-disruptive everything Made for large-scale modern IT infrastructure StorPool Storage Software only, hardware-agnostic Linearly scalable Block first, some file
  • 26. IT TEAMS HAVE LONG-TERM RELATIONSHIPS WITH Datacenter providers Hardware vendors Distributors and resellers
  • 27. PROVIDES STORAGE AS A SERVICE UNDER BRING YOUR OWN DEVICE TERMS
  • 32. WE NEED TO CHANGE THE WAY WE THINK ABOUT DATA STORAGE @StorPool StorPool.com StorPool Storage
  • 34. AGENDA Introduction to StorPool Storage Boyan Ivanov, CEO Tech Excellence Boyan Krosnov, CTO Helping IT Teams Do Storage Right Alex Ivanov, Product Lead
  • 36. HOW DO SHARED-DISK ARRAYS FIT INTO THE TYPICAL DATACENTER ENVIRONMENT? ToR SWITCH SPINE SWITCH
  • 37. Slow adoption of innovations X High-cost system lifecycle X Multi-rack availability requirements X Bottlenecks at the controllers X Always-on operating model X Datacenter-scale Ethernet network X Building block concept X BUT WHY IS THE OLD MODEL NOT WORKING? Hard to manage at scale X
  • 38. Uses commodity servers Fits datacenter designs Fits customer workflows Addresses demanding applications Extracts near-100% of HW potential Has high-speed data efficiency Scales linearly Fades into the background THE IDEAL PRIMARY STORAGE PLATFORM
  • 41. StorPool Storage (up to 4,032 nodes) Network – 2x/4x 10-100 GbE per storage node Storage Node Storage Node Storage Node Parallel NVMe/TCP Service Parallel iSCSI Service Highly Available NFS Service Server Server SSD Server SSD SSD SSD SSD SSD SSD SSD SSD iSCSI, NFS NVMe/TCP, iSCSI, NFS NVMe/TCP, iSCSI, SP Block, NFS NVMe/TCP, iSCSI SP Block, NFS VMware Hyper-V block block block KVM Hypervisor StorPool Client block device driver block block block Linux Bare-Metal StorPool Client block device driver
  • 42. StorPool Storage Metrics & Analytics Alerts & Reporting VolumeCare CMP Integrations
  • 43. Parallelism and shared nothing architecture Exceptional latency and throughput Up to 64 sub-clusters managed as a unified large-scale system Linearly scalable large sub-clusters of capacity & performance Datacenter-scale deployments Data management, incl. backup and DR StorPool Storage ARCHITECTURE AND HOW IT WORKS
  • 44. StorPool Framework upstream drivers: - block device - vhost hardware threads, tasks, scheduling, loader memory management huge pages, shm, allocator network protocols: - Ethernet - TCP/IP (ARP, ICMP, TCP) & BGP downstream drivers: - pci device - nvme, linux block - ixgbe, i40e, mlx4, mlx5, bnx2x, bnxt, virtio-net, linux net - rdma - ioat - hsleep, ksleep Linux, KVM integrations StorPool Protocol on-disk format Block Client NVMe/TCP iSCSI Bridge mgmt Server Beacon service discovery service monitoring quorum PMem NVMe SATA/SAS VMware Hyper-V Oracle LVM CLI GUI ARCHITECTURE DEEP DIVE
  • 45. LINEARLY SCALABLE CAPACITY AND PERFORMANCE 453 823 773 1,686 75 106 71 106 4k qd 7x64 70/30[kIOPS] 16k qd 7x64 70/30[kIOPS] 2,000 1,500 1,000 500 0 3 servers, 12 NVMes 4 servers, 16 NVMes 5 servers, 20 NVMes 6 servers, 24 NVMes 7 servers, 28 NVMes 3 servers, 12 NVMes 4 servers, 16 NVMes 5 servers, 20 NVMes 6 servers, 24 NVMes 7 servers, 28 NVMes 4k qd 7x1 100% read[us] 4k qd 7x1 100% write[us] 100 75 50 25 0 125
  • 46. NVMe SSD KVM Virtual Machine KVM Virtual Machine NVMe SSD NVMe SSD NVMe SSD NVMe SSD NVMe SSD NIC server instance 1CPU thread 2-4 GB RAM server instance 1CPU thread 2-4 GB RAM server instance 1CPU thread 2-4 G RAM block instance 1CPU thread 1 GB RAM 25GbE 25GbE Kernel bypass Kernel bypass
  • 47. DATA PROTECTION Storage Node SSD SSD SSD SSD Client Volume/LUN Storage Node SSD SSD SSD SSD Storage Node SSD SSD SSD SSD Storage Node SSD SSD SSD SSD Storage Node SSD SSD SSD SSD Distributed storage and processing Synchronous replication across storage nodes
  • 48. PERFORMANCE CONSISTENCY, FOR REAL! Average 95th 99th 99.9th Read latency 133µs 224µs 358µs 498µs Write latency 66µs 81µs 99µs 293µs NEW APP ADDED: Transactional workload with 4KB read/write SYSTEM LOAD: 50,000 IOPS per TB Usable 4KB 70/30 read/write 500,000 IOPS total
  • 49. MULTI-RACK SETUP StorPool Storage can be deployed “horizontally” in multiple racks StorPool Cluster
  • 51. HARDWARE REFRESH CYCLE Without service interruption Replace servers one-by-one No maintenance windows No complex data migrations
  • 52. Integrations: Kubernetes CSI, OpenStack Cinder, CloudStack, OpenNebula, OnApp StorPool VolumeCare (enables snapshot-based backup and DR services) StorPool API Kubernetes OpenStack CloudStack OpenNebula OnApp Bespoke/custom integrations INTEGRATIONS & AUTOMATIONS
  • 53. SCOPE OF MANAGED SERVICES Continuous Service Improvement DESIGN DEPLOY FINE-TUNE MONITOR MAINTAIN
  • 54. WHAT ENABLES OUR MANAGED SERVICES AT SCALE Мetrics, analytics Еvents and alarms, multi-channel alerting Аutomation tooling Unification of everything possible Cross-immunization Hyperscaler-like approach to infrastructure
  • 55. LESS HASSLE & FEWER PEOPLE INVOLVED Higher overall cloud efficiency Higher performance Higher density Fewer servers => => =>
  • 56. & 60% more applications per rack Fewer compute servers needed for applications 60% less electricity consumption They're paying less $ even though electricity prices went up! USA electricity prices are 2x higher.
  • 57. Parallel, always-on, no silos, automated. No toil, no hassle. @StorPool StorPool.com StorPool Storage
  • 58. Helping IT Teams Do Storage Right Alex Ivanov, Product Lead
  • 59. AGENDA Introduction to StorPool Storage Boyan Ivanov, CEO Tech Excellence Boyan Krosnov, CTO Helping IT Teams Do Storage Right Alex Ivanov, Product Lead
  • 60. “The hyperscalers have long proven that commodity hardware is good enough to run industrial-grade workloads. The trick is in applying the right software layer to get the most out of the available hardware.”
  • 61. StorPool Storage Metrics & Analytics Alerts & Reporting VolumeCare CMP Integrations
  • 62. Dynamic provisioning Self-service Parallel service for 1000s of workloads Always-on 99.999% with maintenance included Consistent low latency Linear scalability (no data silos) No single point of failure
  • 63. Issues with performance consistency Infrastructure sprawl ENTER THE STORAGE GAME Complexity Endless cycles High-cost workarounds
  • 64. HARD & EXPENSIVE SYSTEM LIFECYCLE MANAGEMENT License tied to specific hardware Systems rarely add performance during lifecycle Modernize, and you need a new license Time-consuming HW modernization Fixed overhead Perfect for prophets Pay-as-you-go? Typically a fixed monthly cost Hiked MSRPs Enable fictional 60% - 90% discounts Support and drives Tagged as % of MSRP ZOOMING INTO THE STORAGE GAME
  • 65. VENDORS NEED ROI FOR EACH SYSTEM THEY DEVELOP WHY?
  • 66. THE IMPORTANCE OF BLOCK STORAGE PERFORMANCE Slower block storage performance Slower application response times Reduced user productivity Delayed time-to- market Reduced revenues and margins
  • 67. PRODUCTIVITY PLUMMETS WITH SLOW APP RESPONSE TIMES 80%+ for experts 75%+ for average 95%+ for novices Transaction rate 4,000 3,500 3,000 2,500 2,000 1,500 1,000 500 0 0.25 0.5 0,75 1.0 1.25 1.5 Expert Average Novice Application response time (seconds)
  • 68. SLOW APP RESPONSE TIMES COST MONEY, ESPECIALLY AT SCALE $150k per month for 50 users $300k per month for 100 users $9 million per month for 3,000 users 1 2 3 Potential monthly savings from rapid application response for varying numbers of simultaneous users 10 9 8 7 6 5 4 3 2 1 0 0 300 Users 200 Users 100 Users 50 Users X $100,000 per month Application response time (seconds)
  • 69. IMPACT OF FASTER APP RESPONSE TIMES AT SCALE? Faster response times More productivity More savings 1 2 3 Potential monthly savings from rapid application response for varying numbers of simultaneous users 10 9 8 7 6 5 4 3 2 1 0 0 300 Users 200 Users 100 Users 50 Users X $100,000 per month Application response time (seconds)
  • 70. CULPRITS FOR BLOCK STORAGE PERFORMANCE DEGRADATION Usable capacity utilization ≥ 50% Storage drive failures, large drive rebuilds Data migration between data silos Inefficient storage software stack Cleanup processes Synchronous replication Write penalties due to write algorithms Network/system interconnect mismatch
  • 71. TRADITIONAL APP RESPONSE TIME WORKAROUNDS Faster, bigger direct-attach storage SSDs, NVMe SSDs, Persistent Memory, etc. Faster, better server-storage interconnect Higher bandwidth, NVMe-oF Application splitting & database sharding On-prem or in the cloud More, faster, newer storage systems Come with faster drives and interconnect At large scale? Not so much... Easy to do at small scale Storage, servers, software, network, racks, etc. Leads to infrastructure sprawl Leads to orphaned storage, a.k.a. siloed storage App instances require separate I/O Splitting, patching, troubleshooting, updating, load balancing... Waste costs across the board
  • 72. TRADITIONAL APP RESPONSE TIME WORKAROUNDS LEAD TO EXCESSIVE 2X TO 3X MORE EXPENSIVE INFRASTRUCTURE TCO!
  • 73. MORE, FASTER, NEWER STORAGE ARRAYS HELP, TO A POINT Faster and bigger SSDs More arrays enable more app splitting & DB sharding Better overall performance @ premium cost Some scale-out options with active- active pairs
  • 74. BUT, WHAT HAPPENS AT SCALE, ESPECIALLY AT LARGE SCALE?
  • 75. WHEN YOU HAVE TENS OR HUNDREDS OF ARRAYS? USABLE AND PERFORMANCE CAPACITY CHAOS Continuous workload monitoring Constant rebalancing between data silos Performance degradation due to capacity consumption Labor-intensive data migrations for system lifecycle management
  • 76. Why? Vendors need ROI for each system they develop. IN SHORT, ENTER THE STORAGE GAME HARD & EXPENSIVE SYSTEM LIFECYCLE MANAGEMENT License tied to specific hardware Systems rarely add performance during lifecycle Modernize, and you need a new license Time-consuming HW modernization Fixed overhead Perfect for prophets Pay-as-you-go? Typically a fixed monthly cost Hiked MSRPs Enable fictional 60%-90% discounts Support and drives Tagged as % of MSRP
  • 77. ENDING THE STORAGE GAME Variable cost that follows revenue Easy for business Pay-per-use Pricing aligns with end-user offerings No pricing fugazis With usage-based volume discounts 24/7 support Included with the StorPool license EASY, NON-DISRUPTIVE SYSTEM LIFECYCLE MANAGEMENT Why? Systems built with cheap read-intensive storage drives Adding drives or servers scales capacity and performance Modernize as vendors introduce new server generations Add new hardware generations non-disruptively
  • 78. WOULDN'T BE POSSIBLE WITHOUT... Consistent performance even under massively parallel I/O Easy workload consolidation with less infrastructure
  • 80.
  • 81. “The StorPool product is perfect from cost, flexibility (grow-as-you-go), and performance point of view. Second - and for us equally important - is the way the StorPool team helps us to build the most advanced hosting platform and their fast and committed support in case of an issue." Hink Wiersema Product Manager and Architect
  • 82. ATOS VIRTUAL ORACLE COMPUTING HOTEL LAUNCHED IN 2013 Hinks' goals for what should be offered to end customers Flexibility Decrease Oracle licensing Low Total Cost of Ownership Running workloads for Banks Insurance companies Hospitals In the Netherlands
  • 83. HARD & EXPENSIVE SYSTEM LIFECYCLE MANAGEMENT License tied to specific hardware Systems rarely add performance during lifecycle Modernize, and you need a new license Time-consuming HW modernization Fixed overhead Perfect for prophets Pay-as-you-go? Typically a fixed monthly cost Hiked MSRPs Enable fictional 60%-90% discounts Support and drives Tagged as % of MSRP HINK‘S GOALS Flexibility Low Total Cost of Ownership Decrease Oracle licensing
  • 84. DIFFICULT SYSTEM LIFECYCLE MANAGEMENT License tied to specific hardware Systems rarely add performance during lifecycle Modernize, and you need a new license Modernize to next-gen HW? Migration Fixed Overhead Perfect for prophets Pay-as-you-go? Typically a fixed monthly cost Hiked MSRPs Enable fictional 60%-90% discounts Support and Drives Tagged as % of MSRP HINK‘S GOALS Flexibility Low Total Cost of Ownership Decrease Oracle licensing CRASHED STRAIGHT INTO THE STORAGE GAME
  • 85. WORKING WITH HIS TEAM, HE GOT TO A POINT WHERE HE FACED 5 SIGNIFICANT CHALLENGES Maintain service quality and optimize prices Align his storage pricing with customers’ pay- per-use model Remove maintenance windows for HW and SW changes Stay competitive by continuously modernizing HW components Optimize the VOC hotel lifecycle and cost management
  • 86. StorPool HAD A SOLUTION FOR ALL OF THAT Maintain service quality and optimize prices 99.999% availability with maintenance included Align his storage pricing with customers' pay-per-use model Pay-per-use with no cost surprises Remove maintenance windows for HW and SW changes Easy, non-disruptive system lifecycle management Stay competitive by continuously modernizing HW components Easy adoption of performance and capacity innovations non-disruptively, at any time Optimize the VOC hotel lifecycle and cost management Systems built with COTS hardware that easily support his environment
  • 87. BUT THAT CREATED A NEW CHALLENGE FOR HINK...
  • 88. …HE HAD TO ADDRESS ALL THE "WHY StorPool" QUESTIONS NEXT…
  • 89. А NEW MINDSET FOR RUNNING THE ATOS VOC HOTEL
  • 90. NEW Datacenter design Always-online operating model Way of working - during normal work hours, no more 2am and weekend work Backend billing model (OpEx, not CapEx) Ways to track and manage service P&L Way of thinking about hardware modernization (add anytime approach) Primary storage partner Atos team simply creates and manages volumes Way of working with a solution provider: Single data pool that is always available Hyperscaler-like Principles for deploying, managing, monitoring, and maintaining block storage
  • 91. AFTER PUSHING ALL THAT THROUGH, THERE WAS A FINAL HURDLE: THE FIRST UPDATE…
  • 92. THE USUAL ORDER OF OPERATIONS Notify all your customers that you're doing an update Plan maintenance downtime for workloads Plan this for a Saturday or Sunday, or during the night Have all your people on deck Be ready to fix everything in case things go wrong Be ready to roll back in emergencies
  • 94. THE UPDATE ROLLED THROUGH, NODE BY NODE
  • 95. “The StorPool product is perfect from cost, flexibility (grow-as-you-go), and performance point of view. Second - and for us equally important - is the way the StorPool team helps us to build the most advanced hosting platform and their fast and committed support in case of an issue." Hink Wiersema Product Manager and Architect
  • 96. WHAT ARE THE OPTIONS?
  • 97. KEEP WASTING INFRASTRUCTURE, TIME, PERSONNEL, RESOURCES & MONEY
  • 98. LET StorPool HELP YOU EXIT THE STORAGE GAME ONCE AND FOR ALL @StorPool StorPool.com StorPool Storage