Navigating the software-defined storage world shaping tomorrow with you René Huebel "Software-defined storage Crossing the right bridge"
SDS the model and the promises Control Abstraction The promises More flexibility Better utilization Higher scalability Lower costs Infrastructure
Driving forces behind the trend Blurring the hardware borders Continuous performance growth Acceptance of virtualization Portability and flexibility Popularity of cloud technologies Demand for reasonable storage
Which approach gives you better results? Proprietary SDS Software Agile and SDDC model Commodity hardware Tunable, flexible performance Software vendor Lock-in Open-source SDS + commercial support Agile & software-defined model Commodity hardware Unpredictable storage growth Compose storage services via API Classic Arrays + virtualization Data center stability Predictable performance Predictable data growth HW & SW lock-in penalty Do it Yourself SDS based on open-source Max flexibility No look-in Limited vendor support Low cost but high risk
Virtualization and Mainstream SDS Central management Pools and virtualizes capacity across devices Advanced, comprehensive storage functions Comfortable provisioning of storage to users Strength Configuration & operation is no longer dependent on hardware functions Flexibility and freedom of choice by selecting the physical storage platform
A new design: Hyper-Converged (HCI) VM VM VM VM VM VM VM Consumers Administration Hypervisor Abstraction & Control Software SSDs HDDs SSD HDDs SSD HDDs Resource layer
in hyper-converged IT (HCI) The hyper-converged approach Server hardware in a double role Deep integration in virtualization Fast scalability Challenges Less suitable for mixed storage workloads Complex integration work / lifecycle Additional storage silo
Windows 2016: Spaces Direct Hyper-Converged SDS Compute and storage together Scale and managed together Small to medium sized deployments Converged SDS Compute & storage separate Scale and managed independently Larger scale-out deployments
PRIMEFLEX for Spaces Direct About Fujitsu and Microsoft Spaces Direct: The solution gives us excellent performance, reduces cost for storage, floor space, power and cooling. The flexible platform design allows us to serve both our SMB and enterprise customers and provides us with the investment protection for future growth. Adam Lantz, Sales Director, Make IT
VMware VSAN VM VM VM vsphere Virtual SAN VM VMware VSAN provides data management functionality embedded in the ESX hypervisor Aggregates disks and SSDs of the underlying servers into virtual and clustered data store Includes quality of service and high-availability services Virtual SAN Clustered Datastore Ideal for data services in large and growing VMware environments
Virtual SAN scalability Scale OUT Scale UP Add more disks Add more nodes 64 Hosts 6400 VMs 6M+ IOPS Elastic, granular and non-disruptive scalability
Vertical: Travel vsan Version: 6.x Type: PY RX25xx series Clustersize: 100+ nodes Customer comments 40% lower costs per VM FJ convinced in terms of Datacenter technology, performance & energy efficiency Need for massive cost reduction
Vertical: Automotive VSAN Version: 6.x type: PRIMERGY RX2540 series Clustersize: Starting with 5 s for VDI Customer comments: Focus on cost reduction & consolidation Optimizing VDI landscape FJ provided best reliability
Vertical: Chemical industry VSAN Version: 6.x / PRIMERGY RX2540 series Remote Office / Branch Office (ROBO) Data Center Infrastructures Customer comments: Massive cost reduction Centralized management and upgradeability Reliable solution
Integrated System: PRIMEFLEX PRIMEFLEX for VMware VSAN VMware VSAN Certified Fujitsu x86 PRIMERGY Servers Flash-based Boot Device Pre-Installed VMware Software Single point of contact for support Flexible Licensing Integrated Systems reduce the TCO of Hyper Converged IT.
Game changers
2005 2007 2009 2011 2013 2015 2017 2019 How to stop the DATA avalanche? 40 35 30 25 20 15 10 5 0 44+ ZB in 2020 Requirements: Cloud storage Unlimited scalability 100% automation Lowest costs on the basis of /TB Resistance and longevity Distributed Processing Software
Open Source: Do It Yourself vs. Supported Risks Own Efforts Costs Skills Softwaresupport Updates, maintenance, troubleshooting Lifecycle Management Keeping HW and SW updates in sync Do all yourself Maintenance contract for storage SW Operative Verwaltung of HW and SW components Integration of Software with Hardware, Network Investment Use a complete SDS solution
Complete SDS Solution ETERNUS CD10000 Object Access Block Access ETERNUS CD10000 Unified Management Red Hat Ceph -Software File Access (TechPreview) Hyperscale SDS Cloud Based on Red Hat Ceph open-source technology Scalability up to 100 PB Appliance approach combining HW & SW & services ETERNUS CD10000 storage nodes Unified management of HW & SW
ETERNUS CD10000 Architecture Gateways / Clients Object Block File Unified Access Management of ETERNUS CD10000 CEPH Cluster Distributed / GW management SW deployment Cluster management Network management Rolling upgrades Log-File management SNMP CEPH OS CPU RAM Disks NW CEPH OS CPU RAM Disks NW CEPH OS CPU RAM Disks NW CEPH OS CPU RAM Disks NW CEPH OS CPU RAM Disks NW CEPH OS CPU RAM Disks NW Ceph SW Linux OS Hardware Management Monitor 1 2 3 4 N
Place, Replicate, Recover, Migrate D Pseudo-random distribution
Place, Replicate, Recover, Migrate D D Pseudo-random distribution Transparent creation of data copies for zero downtime
Place, Replicate, Recover, Migrate Automatic recreation of lost redundancy D D D Pseudo-random distribution Transparent creation of data copies for zero downtime
Place, Replicate, Recover, Migrate Automatic recreation of lost redundancy D D Pseudo-random distribution Transparent creation of data copies for zero downtime Automated tech refresh makes the system immortal
Place, Replicate, Recover, Migrate Automatic recreation of lost redundancy D D Pseudo-random distribution Transparent creation of data copies for zero downtime Automated tech refresh makes the system immortal
Vertical: Cloud Service Provider 2x ETERNUS CD10000 in a DR configuration Use case: Object Store Multiple services in parallel Customer comments Fully automated operations Ensuring reliability SLAs Seamless scalability FJ acts as a SPOC
To avoid beginners' mistakes Software-defined depends on the right hardware foundation
Taking the right bridge New usage scenarios drive the emergence of SDS SDS shows its strength in areas of large and fast scalability needs Understand the needs for your application environment Identify which SDS model is the best fit Do evaluations to justify the relevance in your scenario Keep an eye on software functions and pricing Fujitsu has experience and offerings in all categories
Crossing the right bridge Is a great chance to accelerate the change!
What are the differences? Block SAN interface Structured data Thousands of files and many TB Block granularity Specific location on disk No metadata Complex management File NFS/ CIFS Structured/ Semi/ Unstructured Millions of Files and 1 Petabyte File granularity Fixed locical hierachy Very limited metadata Complex management Object HTTP interface Semi/ Unstructured Billions of Objects / Multi-EB Object granularity Location independent Rich metadata Simple management