Hyperconverged Infrastructure, It's the Future
Download as PPTX, PDF•4 likes•3,108 views
Hyper-converged infrastructure (HCI) simplifies IT by integrating storage and compute resources, which can reduce complexity compared to traditional setups. This document discusses the efficiency and resilience of HCI, including concepts like erasure coding and deduplication, while comparing various market leaders and their offerings. It also highlights the advantages and limitations of HCI appliances versus software-defined storage solutions.
Hyperconverged Infrastructure, It's the Future
- 1. Hyper-Converged Infrastructures It’s The Future Amsterdam, 24th September 2015Howard Marks Everyone Says So
- 2. • 30 years of consulting and writing for trade press – Now at NetworkComputing.com • Chief Scientist DeepStorage, LLC. – Independent test lab and analysts • Co-Host Greybeards on Storage podcast @DeepStorageNet on Twitter [email protected] Your not so Humble Speaker
- 3. The Question is, Which Future?
- 4. Conventional Infrastructure is Complex • Virtualization simplified the server • But complicates the infrastructure – SAN and NAS replace server RAID – More east-west traffic • Many vendors many UIs – No CLI task is easy once a quarter • Rip, replace, migrate upgrade paths 8 1 2UID41 5 32 SID HP ProLiant DL360 G7 FANS PROC 1 PROC 2 POWER SUPPLY 2 POWER SUPPLY 1 OVER TEMP POWER CAP 1 2 3 4 9 8 7 6 5 4 3 2 1 1 2 3 4 5 6 7 8 9 ONLINE SPARE MIRROR UID 2 1 4 3 6 5 8 7 6 5 4 3 2 1 4 3 2 16 5 PROC 1 PROC 2 POWER SUPPLY 2 POWER SUPPLY 1 OVER TEMP POWER CAP 1 2 3 4 9 8 7 6 5 4 3 2 1 1 2 3 4 5 6 7 8 9 AMP STATUS FANS DIMMS HP ProLiant DL380 G7 CS Array CONSOLE Solid ON = Link Blinking = Activity 49 50 51 5233 35 3734 36 38 39 40 4441 4342 4845 474617 19 2118 20 22 23 24 2825 2726 3229 31301 3 52 4 6 7 8 129 1110 1613 1514 Summit X460-48t TM STACK NO. Stack1MGMT FAN PSU-1 PSU-2 2 3 4 Management USB 10G Pow erEdge M1000e 9 10 1 2 11 12 3 4 13 14 5 6 15 16 7 8 0 1 00 1 0 0 1 00 1 0 0 1 00 1 0 0 1 00 1 0 0 1 00 1 0 0 1 00 1 0 0 1 00 1 0 0 1 00 1 0 CONSOLE Solid ON = Link Blinking = Activity 49 50 51 5233 35 3734 36 38 39 40 4441 4342 4845 474617 19 2118 20 22 23 24 2825 2726 3229 31301 3 52 4 6 7 8 129 1110 1613 1514 Summit X460-48t TM STACK NO. Stack1MGMT FAN PSU-1 PSU-2 2 3 4 Management USB 10G Rack Servers SAN Switches Storage Array Ethernet Switches Blade Servers
- 5. Hyperconverged Infrastructure • Simplify infrastructure by combining storage and compute • Storage management in VM or Hypervisor kernel process – Pools SSD & HDD across multiple servers – Shared nothing, scale-out storage • Must protect data against node failure • Can be sold as hardware or software – Storage is software defined/driven
- 6. Storage Efficiency and Resilience • Dual controller, RAID-6 – Standard external storage today – Can survive 1 controller failure plus – 2 disk failures before data loss – 71% (5+2) to 86% (12+2) efficient • Most HCI uses replication – 2-way can survive 1 controller failure OR 1 drive • 50% efficient – 3-way can survive 1 controller failure AND 1 drive • 33% efficient • Erasure coding improves efficiency – Especially for large clusters – Can negatively impact random I/O UID 1 2 DL120 G5 ProLiant HP UID 1 2 DL120 G5 ProLiant HP X X X 50% 60% 70% 80% 90% 100% Number of Appliances Erasure Coding Efficiency
- 7. Two Locations and RAID • Creating a Virtual Storage Appliance (VSA) • VSAs great solution for ROBO, SMB • Local storage, may require RAID controller • Publish as iSCSI or NFS • Example: Stormagic – Basic iSCSI VSA 2 nodes $2500
- 8. Isn’t Deduplication The Answer? • Deduplication and compression take CPU and RAM • CPU cycles on HCI Host are expensive – Typical server $8-15,000 • 10 2.6Ghz cores, 256GB RAM $10K – Software • vSphere Enterprise+ $6990 • S&S 3yrs $5244 • Windows Data Center $6155 • Big storage VMs mean less resources to run your VMs • Deduplication on spinning disk creates read bottlenecks
- 9. The Hyper-Converged Appliance • Most appliances are just servers – High density cabinets optional – Simplivity adds ASIC based dedupe – NVDIMMs a possibility • Software provides storage smarts
- 10. HCI Appliance • Better qualification – Possibly Tighter integration • Faster time to value • Simplified support • Generally higher price • Limited selection of nodes – 4:1 Compute/CPU • Limited capacity Software Only • Greater flexibility – Range of hardware • E3 to 24 drives – Deployment models • Storage on bare metal • Maintain server vendor • Every system a snowflake Reference architectures & meet in channel models blur the line
- 11. • HCI not as attractive at scale – 20 servers and storage cost less than – 24 servers with HCI – vVols make external storage easier • Virtualization admins may not be paranoid enough
- 12. HCI Market Leaders • Nutanix – Appliances, direct and via Dell – Storage based on GFS – vSphere or Acropolis (KVM) w/Prism management • Simplivity – Appliances, direct, Cisco, Lenovo – Hardware assisted dedupe – Integrated multi-site, S3 backup – NFS for external servers
- 13. VSAN/EVO:RAIL • Runs in vSphere kernel • Hybrid or all flash* • Requires n-1 witnesses for n replicas – So 5 nodes for FTT=2 • New in 6.1 – 2 node with external ESX witness – Metro synchronous replication • EVO:RAIL – OEM spec for 4 node/2u – Generally overpriced
- 14. Smaller, More Specialized Solutions • Scale Computing, Nimboxx – Use open source KVM w/management GUI – 3 node total cost under $30,000 • Pivot3 – Erasure coding, 12 drive bays • Good for orgs with lots of cold data • Breqwatr Cloud Appliance – HyperConverged Openstack
Editor's Notes
- #9: Total $18,389 Dell R730 2xE5-2660 (10 core 2.6 Ghz) 265GB RAM Etc $10,000 Vmware E+, Windows