Cisco hyperflex software defined storage and ucs unite
-
Upload
cisco-canada -
Category
Technology
-
view
61 -
download
4
Transcript of Cisco hyperflex software defined storage and ucs unite
Cisco Storage Overview
Dec 1, 2016
2
The data growth impact is significant…
… along with the business
value.
•Multi-billion dollar business
•High margin business
•Central to the core of many
other business units and
decisions
We witnessing a major shift
Consolidation
Virtualization
IT as a
Service
Hybrid
CloudTRADITIONAL
DATA CENTER
2000 2008 2015
ADOPTION CURVE
The Next 5+ years
We are here
Simplicity
Agility
Standardization
DISTRIBUTED
CLOUD
DATA CENTER
Automation
Efficiency
Speed
IT Transformation: The Tale of Two ITs
Emerging datacenter architecture use-cases
Traditional Shared Storage
- SAN or NAS shared to many servers
- High capacity flexibility
- Can be single point of failure
- Storage management built-in / HW-based
Clustered, Server-side Storage
- Cisco HCI pioneering in segment
- High performance use-cases: NVMe storage in standard
servers in cluster
- Delivers maximum CPU-to-storage efficiency
- High availability / data replication done via Application or OS
- Can achieve highest in performance and application uptime
Storage Trend: Capacity Tier Moving into the Server
All Enterprise Workloads
All Enterprise Workloads
VDI/BURA/BC-DR/IAAS
Scale-Out File/Block/Object
TAM = $5B
CAGR = 31%
TAM = $40B+
CAGR = 0%
TAM = $0.5B
CAGR = 150%
TAM = $6.2B
CAGR = 35%
Pretested and preconfigured bundled solutions,
acting as whole, consisting of compute,
network, storage and management software.
Provide platform to meet customer challenges
and growing infrastructure demands
Consistent, financially efficient, operationally
simplified, independently scalable solution
Rapidly deployable, allows granular upgrades
Customer Challenges &
Infrastructure Demands
Converged Infrastructure
FLEXPOD UCP- SelectVSPEX SmartStackVersaStackVBLOCK
Proliferation of applications; Need for any application on any device
Roll out new and/or business critical database applications quickly & consistently
Single system operation – Reduced Infrastructural and management complexity
Add capacity and storage when needed
Accelerated implementation
Streamlined support
Cisco UCS
Cisco Nexus
Market Trends
2) Hyper-Convergence is evolving,
growing & disruptive to traditional storage
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0
2
4
6
8
10
12
14
16
18
2013 2014 2015 2016 2017
Billio
ns
Integrated Systems Hyper-converged
Fast Deployment (Hours not Days)
Management and Operation Simplicity
(Intuitive UI)
Simple Elasticity (automatic scale)
Cost efficiency (predictable step
growth)
Minimize Risk (Pre-Validated
Designs)
Any workload (Mission Critical Apps)
Flexibility (Virtualized and Bare Metal)
Independent Scale out (building
blocks)
1) Increasing
overlap
The programming that controls the storage is decoupled from the physical hardware
Consists of the Storage Only
Emphasizes storage services such as deduplication or replication, instead of storage hardware
A shared pool that runs on commodity hardware
Utilizes Policy Based Management
Often referred to as Storage Virtualization and is part of a bigger industry trend of Software Defined Datacenter
Definition: What is Software Defined Storage (SdS)
Software Defined Storage
Local Physical Storage
Hosts
Storage Software can run on bare
metal or as a VM on a hypervisor host
New Converged Infrastructure offering utilizing Software Defined Storage
Tight integration of x86 servers for compute and storage, networking and virtualization in all-in-one appliance.
Integration of hypervisors and physical infrastructure
Simple Scale one unit/appliance at a time
Simple Deployment measured in hours
Centralized Management, intuitive UI
Definition: What is Hyper-convergence?
Controller
VMVM VM VM
Local Physical Storage
Hypervisor
Controller
VMVM VM VM
Local Physical Storage
Hypervisor
Controller
VMVM VM VM
Local Physical Storage
Hypervisor
• Fast Deployment (Hours not Days)
Management Simplicity (Centralized and intuitive UI)
• Operational Simplicity (Target VM administrator)
• Simple Elasticity (automatic scale)
Cost efficiency (predictable small step-based growth)
What does Hyper-converge promise?
Converged Infrastructure vs. Hyper-Converged
Converged or Integrated Infrastructure – i.e. vBlock, FlexPod, SmartStack, VersaStack, etc• Preconfigured bundles of hardware & software delivered as a single-solution• Basically disparate components acting as a whole with some type of software management layer• Typically includes a compute tier & storage tier connected together with one or more networks• Works with both virtualized and non-virtualized (Bare metal) workloads• Can easily scale one component separate from another since they are separate building blocks –
allows for granular upgrades and tweaks
Hyper-Converged Infrastructure• Compresses the Compute & storage tiers of hardware into a single tier• Deployed as a cluster of nodes with each node having compute/storage resources• Storage is virtualized & shared across entire cluster• Foundational components often include integrated backups, snapshot capability, data deduplication,
inline compression and even WAN optimization• Hypervisor Centric, so often only works with virtualized workloads• Typically a node includes both compute and storage resources, so cannot add one without the other• This federated storage model has several advantages – improved storage utilization, balanced
workloads and ease of storage migration
Customer Criteria Converged HyperConverged
Compute, Storage, Network, Server Virtualization in one Flexible, allows choosing from
Multiple Vendors
Rigid, single vendor offers or
limited choices
Scalability Flexible, can scale up and out
as needed
Has Limitations
Common Management Software Increasingly offered as an
option
Built in by default
Packaged as a Single Unit Possible through SI Offered by default
Pre-Sized and Pre-configured Possible through SI Offered by default
Time to Deployment Days and weeks Hours
Pricing Perceived as high Perceived as low
Performance, Quality, and Uptime High – Designed and validated
for high quality and availability
at various scales and speeds
Low – Tests fail at higher
scales and speeds
Converged vs. HyperConverged
Cisco UCS HyperConvergence Strategy
Customer Needs
Scale
Speed
Simplicity
Technology Options
Integrated Infrastructure
Solutions
Automated
Provisioning
SmartStack FlexPod
UNIFIED ARCHITECTURE
SCALE OUT
UCS M-Series
Modular Servers
UCS
C3160
INTEGRATED INFRASTRUCTUREUCS B-SERIES
UCS MINI HYPERCONVERGED
C-Series
Rack Servers
Software Defined Storage
Solutions
Cisco Strategy
Flexibility of Deployment Options
Best Technology Choice
Comprehensive Validated Designs
Timed with Market Demand
Simplicity, Speed, Scale
+Choice of Storage
Introducing Cisco HyperFlex Systems
Complete Hyperconvergence
Software Defined Compute, Storage & Network
Next Generation Data Management
Always-on Optimization and Flexible Scaling
Future Ready Architecture
Built for Todays Workloads and Emerging Applications
Cisco HyperFlex Configurations
HX240c Cluster
Capacity-heavy 3–8 Node Cluster
(VSI: IT/Biz Apps, Test/Dev)
Per-Node1x1.6TB Cache SSD
up to 23x1.2TB HDDs
SD Card/120GB Back SSD
(Boot/Housekeeping)
HX220c Cluster
Smallest Footprint 3–8
Node Cluster (VDI, ROBO)
Per-Node1x480 GB Cache SSD
6x1.2TB HDDs
SD Card/120GB SSD
(Boot/Housekeeping)
Hybrid Clusters
Compute-heavy Hybrid
(Compute Bound Apps/VDI)
Up to 8 Compute Nodes
Blade or RackB200 M4, C220 M4 or C240 M4 (Supported)
Local Disk, SD Card or SAN Boot
3-8 Node HX220 or HX240 Cluster
4
1
5
2
6
3
7 8 UCSC220 M4
Intel
Inside
XEON
Console!
C220 M4
!
Intel
Inside
XEON
UCSC240 M4
Con
sole
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
1 6 12 18 24
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
UC
S-H
DD
30
0G
I2F
10
5
15
K S
AS
30
0 G
B
!
C240 M4
B200 M4
UCS Integrated HyperFlex Networking
Shipped From Factory With Integrated Networking
High Performance, Flat and Predictably Latent Fabric
Simplified Network Deployments• No specialized protocol
requirements such as IPv6, Multicast
Plug and Play Networking with UCS Unified Fabric
Reduced Complexity, Simplified Decision Making and Deployment
CVM Data Port GroupData vSwitch
CVM Mgmt Port GroupManagement vSwitch
Predefined DC Network Policies
CONTROLLER VM CONTROLLER VM CONTROLLER VM CONTROLLER VM
Integrated Management and
Data Services
Dynamic Data Distribution
Continuous Data Optimization
Independent Scaling of Compute and Capacity
HyperFlex Design Principles
Building on the Right FoundationCisco HX Data Platform
Built From the Ground Up for
Hyperconvergence
Distributed Log-Structured File
System Designed for Scale-out,
Distributed Storage
Advanced Data Services (Snapshots,
Clones) and Data Optimization
(Inline Dedupe, Compression) Without
Trade-offs
Better Flash Endurance and
Disk Performance
Computing, Storage,
Networking, and Hypervisor Integration
No Reliance on Legacy
Filesystems or
Technology
Distributed File system
Local Disks
Unique
Architecture
Local Disks Local Disks Local Disks
Capacity Options
3 Node
HX220c
HX240c
4 Node 5 Node 6 Node 7 Node 8 Node
Assumes: Full HDD population. RF3.
Note: The above calculations are before deduplication & compression. Effective capacity will be higher.
Consult with your Cisco CSE for the latest sizing & design guidance.
6.02 TiB 8.03 TiB 10.04 TiB 12.04 TiB 14.05 TiB 16.06 TiB
23.09 TiB 30.79 TiB 38.48 TiB 46.18 TiB 53.88 TiB 61.58 TiB
Usable Cluster Capacity
Cisco HyperFlex and UCSRunning Non-HyperFlex Workloads In The Same UCS Domain
Cisco UCS Domain
HyperFlex Cluster
+
Non-HX UCS
Systems
Fully Supported Configuration with HX 1.8
UCS Firmware Must Remain at Minimum
Supported Version for HX, or Higher
Useful for Running UCS General Purpose
Workloads Alongside a Separate
HyperFlex Cluster Within the Same UCS
Domain
Example: HX Cluster + C3260 Servers as
a Backup Solution
Cisco HyperFlex and External StorageConnecting HyperFlex Domains To External Storage Arrays
Fully Supported Configuration with HX 1.8 and
Storage Arrays on the Cisco HCL
UCS Firmware Must Remain at Minimum
Supported Version for HX, or Higher
Useful for VM Migrations, RDMs, Backups and
Other Applications
Examples: VM Migrations from External Arrays
RDM from FC Array for Microsoft Clustering
HyperFlex Domain
NFS, iSCSI, FC or FCoE
Direct Connect or use
External Switches
Integrated Management and Data Services
• UCS Manager vCenter Plug-in
• Manage UCS Service Profiles
• Update Server Firmware
• KVM launch of console
• Instant provisioning, cloning and snapshotting of virtual machines from within vCenter
• No separate console, no learning curve
• UCS Manager familiar to 50,000 customers WW
• Server and network deployment settings in pre-configured Service Profiles
• Pointer-based snapshots
• Space-efficient with no performance penalty vs. Vmware Redo Log Snaps
• Fast creations and deletions
• Fine-grained or coarse-grained
• VM-level or VM folder-level
• VAAI-integrated
• Quiesced and crash-consistent
• Use vCenter Snapshot Manager
• Policy-based schedules and retention
Fast and Flexible Native Snapshots
All This Functionality Enabled with the HX Data Platform Filesystem
• Pointer-Based Writeable Snapshots (Instantaneous Clones)
• VAAI integrated
• VM-level granularity
Native VM Clones for Rapid Provisioning
• Batch creation GUI
• Apply unique names
• Use customization spec to apply IP
• Powerful tool to rapidly setup a large set of VMs using just VC (without scripting or View composer); Up to 256 clones in parallel per job
• Golden/Base VM can be a template, powered on or powered off
Cloud-Scale Changes need for Storage
Core Enterprise Workloads Cloud Scale
Traditional Storage
ServerSingle
Server
Unstructured
Date Store
Single
Application
SCM ERP/
Financial
Legacy CRM Email Social Media Email
Video
Backup
Archive
Cloud
Storage
SAN/NAS
Use Cases for the UCS C3000 Server Family
Service Providers
Public Sector
Service Providers,
Enterprise
Service Providers,
Enterprise
Software Defined
Storage
Object Store
Media Streaming and
Content Distribution
Video Surveillance
Exchange
Backup and Archive
(as Service)
Big Data and
Analytics
Service Providers,
Enterprise, Cloud
Cisco UCS C3160/3260
UCS Compute PortfolioPerformance Optimized for Bare Metal, Virtualized, and Cloud Applications
Cloud Scale
UCS C460 M4
Mission-Critical, 4-Socket
Server for Large, CPU-
Intensive Applications
UCS B420 M4
Enterprise Class, 4-
Socket Blade for Large,
Memory-Intensive Bare
Metal
and Virtualized
Applications
Rac
kB
lad
e
UCS B260 M4
Mission-Critical, 2-Socket
Blade for Large, CPU-
Intensive Bare Metal
and Virtualized Applications
UCS B460 M4
Mission-Critical, 4-
Socket Blade for Large,
CPU-Intensive Bare
Metal
and Virtualized
Applications
UCS C240 M4
Ideal Platform for Big Data, ERP,
and Database Applications
UCS B200 M4
Optimal Choice for VDI, Private
Cloud, or Dense Virtualization/
Consolidation Workloads
UCS C220 M4
Versatile, General Purpose Enterprise
Infrastructure, and Application Server
UCS C3000 Series
Ideal Capacity-Optimized
Platform for Large Object
Storage at Scale
Intensive/Mission CriticalEnterprise
Performance
UCS C3260M4 Chassis ArchitectureDrives
4 Rows of Hot-Swappable HDD
4TB/6TB/8TB/10TB with up to 2 Rows
of 400GB/800GB/1.6TB/3.2TB SSD
Total Top Load: 56 drives
FAN8 Hot-Pluggable
Fans
Server NodeUp to (2) Based on Intel V4 CPUs, LSI
12G RAID, Up to 512GB DDR4 RAM
(1024GB Post-FCS), and NVMe
Optional Second NodeServer Node or Drive Expansion
or PCIe Expansion
Up to (4) 120GB/480GB/1.6TB SSDsHW RAID, Hot-Plug, OS/Boot
System I/O Controller (SIOC)Up to (2) Cisco VIC 1300 on Chip
Power Supply4 Hot-Pluggable PSUs
*Shown with Single Server Node
and IO Expander
C3260M4 Chassis Configurations
Dual-Node 3260M4 Single-Node 3260M4
Single-Node 3260M4with IO Expander
Single-Node 3260M4with Disk Expander
Server Node 2
Server Node 1 Server Node 2
Empty
Server Node 2
IO Expander
HDD Expander
Server Node 1
HDDSSDPCIe
Flash
NVMeNVDIMMsDIMMs
Storage Is Moving Ever Closer to the CPU
Cache
• Cisco a leader in enabling storage inside the server – Key Differentiation:
1. UCS Manager provides unique and powerfully-streamlined management experience
2. Advanced BIOS and Fan control algorithms provide highest density with highest and consistent performance
• Performance combined with competitive price-points causes PCIe interface to increase quickly (NVMe is standard driver set)
UCS and Server-side Storage
/NVMe
2016-2018 server-side storage technologiesShipping now
HDD10,000 µs
1,000 µs
SAS/SATA SDD (NAND)
100 µs
PCIe/NVMe SSD (NAND; Add in Card or
2.5” Small Form Factor)10 µsPCIe/NVMe Xpoint (New Media, not
NAND; various Form Factors)
$1/GB
$2/GB
$3/GB
$6/GB
• Streaming/Sequential
• Backup
• Big Data/Analytics
• DRAM Extension
• In-Memory DB
• Caching
1 µs
• OLTP DB (most)
• Some Analytics (Hadoop, SAS)
• Virtual Servers
Flash $/GB
decreasing quickly
- IOPS +
Latency Planned for CY2017
Why is NVMe run over PCIe?
Or
Direct to CPUSERVER
- Peripheral Component Interconnect Express (PCIe) the single chosen interconnect for
NVMe Storage
- NVMe is the first standard for addressing storage designed for Solid State Media – SCSi
was designed for spinning storage media (HDD)
- Fusion io pioneered the PCIe Storage market but…
- Now, with NVMe
- Dramatically lower pricing
- In-box drivers / very easy to deploy
UCS – In-Server Storage Options
SATA
SSDHDD
SAS
SSDHDD
NVMe
SSD
RAID Controller
RAID ControllerPCIe - Direct
to CPU (just
like Fusion io)
- Performance
+
Small Form Factor (SFF) – 2.5” SFF – 2.5”HHHL
SFF – 2.5”
Mainstream Computing
Scale Out
Cisco UCS: Platform for All Architectures
HyperconvergedInfrastructure
Converged Infrastructure
Unified ManagementSingle Control Plane
Single API
Edge Cloud
ROBO
UCS C3000
Series
Fourth Generation
UCSHyperFlex Systems
UCS Mini
E-Series
C-Series
Rack Servers
UCS Integrated
Infrastructure Solutions
Single Operational Model
UCS Manager UCS Director Enterprise Cloud Suite
Closing Thoughts
Orchestration
Granularity
Programmability
Scalability
Physical, Fixed Ratio,
MonolithicAbstracted,
Variable Ratio,
ComposableNexus UCS
ACISystem Link
(M-Series)
CPU CPU CPU CPU CPU
Memory
Local Storage
Network I/O
SAN I/O
Local Storage
Network I/O
SAN I/O
Network I/O
SAN I/OSAN I/O
Memory
Local Storage
Network I/O
Memory
Local Storage
Memory
SAN I/O
Network I/O
Memory
Local Storage
Mode 1 Mode 2
Multi-CoreServer
VirtualizationStorage
VirtualizationSANPC Server
NetworkVirtualization
IT, Apps and Infrastructure are Evolving
Mainstream Applications
Converged Infrastructure
Pre-Assembled
VM VM VM VM VM
SDS
ComposableInfrastructure
APP
Cloud Native & Scale Out Apps
Rack & Multi-node Servers
Dynamically Assembled
Microservices & Containers
Object & Distributed Storage
DevOps Style IT
Traditional IT Models
Analytics
Hyper Converged
Virtualized Apps
Manually Assembled
Cisco Storage Innovation
Industry-leading Integrated Infrastructure Solutions
Storage-optimized UCS Servers
UCS HCI & SDS Solutions
Storage Protocol Optimized I/O
Simplified, Policy-driven Storage
Consumption with UCS Storage
Profiles and UCS Director
Big Data
and
Analytics
OrchestrationServer-Based StorageIntegrated Infrastructure
UCS Hyperconverged
UCS Software-Defined Storage
UCS C3000
Cisco Virtual
Interface Card
1 2 3
Cisco Infrastructure for Future Proofed IT
Traditional IT DevOps Style ITAutomation
Scale up & Scale OutVirtualized & Bare Metal
Converged Infrastructure
Hyper Converged ComposableInfrastructure
+ + +
Cisco Unified Computing System
Virtual SAN
Cisco System Link TechnologyUCS M-Series
Extensible Control Plane&
Comprehensive API
Consistent Infrastructure Management
&Broad Interoperability
Architectural Flexibility&
Future Proofed IT
DevOps Style IT
Traditional IT Models