3. 33
High Performance Databases
Microsoft SQL ServerOracle Databases
Majority of databases can be virtualized
But there are times when you might not virtualize
• Licensing costs
• Virtualized limits for max DB size (Server Maximums)
Need: Support bare metal workloads with scale-out high-performance and
highly available storage from Nutanix
4. 4
Run your 3-tier App entirely on Nutanix
Repurpose Nutanix for bare metal DBs
• Utilize Nutanix better
• Protect server investment
High Performance & Highly Available
• Better price-performance than purpose built All
Flash systems
Eliminates traditional Block Storage Complexity
• Simple Operations
Web
Servers
App
Servers
iSCSI
ABS
3-Tier Applications
5. 55
• Utilize existing server investments to run workloads while receiving
scale-out high-performance storage from Nutanix
• Legacy & bare metal applications that cannot be virtualized can
continue to live out their life until the application is rewritten, but still
receive first class treatment on storage
Investment Protection
6. 6
How ABS Works
BLOCK STORAGE
iSCSI
Traffic Load Balanced Across Nodes
Eliminates the need for iSCSI MPIO
Client intervention NOT required when the
Cluster scales out
Inherits All Nutanix features
Run VMs
Hyperconverged
While serving out
storage externally
7. 77
Starts with Volume Groups
Nutanix Container
HostA
Target: “iqn...VGA”
Attach: “Iqn...HostB”
Volume Group B
vDiskD, vDiskE
Attach: “Iqn...HostA”
Volume Group A
vDiskA, vDiskB, vDiskC
HostB
Target: “iqn...VGB”
A B C ED
Disk0 Disk1 Disk2 Disk0 Disk1
Virtual Disks
LUNs
9. 99
Showing Legacy the Nutanix way
• Bare-metal apps can now access all capabilities of the Nutanix
scale-out storage fabric
• Extremely simple configuration using Volume Groups
• Unique high availability scheme without complicated multipath
IO (MPIO) setup and maintenance overheads
10. 1010
Traditional Dual Controller SAN
SAN Storage
Controller 1
10.4.1.1
Controller 2
10.4.1.2
Disks
Path 2
Application Server
LUN 0
LUN 1
LUN 2
iSCSIInitiator
MPIODriver
Path 1
• Redundant paths to dual
controller
• MPIO driver monitors
path health and triggers
failover
• Any topology change on
storage requires client
reconfiguration
• MPIO configuration and
behavior is client specific
Not suitable for a scale-out distributed
storage backend
(Active)
(Failover)
11. 1111
Simple Client Configuration with ABS
Application Server
VG-
SQL
Disk 0
VG-
SQL
Disk 1
VG-
SQL
Disk 2
iSCSIInitiator
Node A
10.4.1.1
Node B
10.4.1.2
Node C
10.4.1.3
Data Services IP
10.4.1.100
Login VG-SQL
Redirect 10.4.1.2
1. Configure discovery portal 10.4.1.100
2. Login to VG-SQL using data services IP 10.4.1.100
Login VG-SQL
Session Established
VG-SQL
12. 1212
High Availability
Application Server
VG-
SQL
Disk 0
VG-
SQL
Disk 1
VG-
SQL
Disk 2
iSCSIInitiator
Node A
10.4.1.1
Node B
10.4.1.2
Node C
10.4.1.3
Data Services IP
10.4.1.100
Login VG-SQL
Redirect 10.4.1.1
Node B failed or brought down for maintenance
VG-SQL
Session Disconnect
13. 1313
High Availability of Data Services IP
Application Server
VG-
SQL
Disk 0
VG-
SQL
Disk 1
VG-
SQL
Disk 2
iSCSIInitiator
Node A
10.4.1.1
Node B
10.4.1.2
Node C
10.4.1.3
Data Services IP
10.4.1.100
Node C hosting the Data Service IP failed
VG-SQL
14. 1414
Showing Legacy the Nutanix way
• Bare-metal apps can now access all capabilities of the Nutanix
scale-out storage fabric
• Extremely simple configuration using Volume Groups
• Unique high availability scheme without complicated multipath IO
(MPIO) setup and maintenance overheads
• A single application can utilize all cluster resources
15. 1515
Load Distribution using Virtual Targets
Application Server
VG-
SQL
Disk 0
VG-
SQL
Disk 1
VG-
SQL
Disk 2
iSCSIInitiator
Node A
10.4.1.1
Node B
10.4.1.2
Node C
10.4.1.3
Data Services IP
10.4.1.100
VG-SQL
16. 1616
Load Distribution using Virtual Targets
Application Server
VG-
SQL
Disk 0
VG-
SQL
Disk 1
VG-
SQL
Disk 2
iSCSIInitiator
Node A
10.4.1.1
Node B
10.4.1.2
Node C
10.4.1.3
Data Services IP
10.4.1.100
• Single VG-SQL exposed as multiple virtual iSCSI targets
• Each virtual target can be independently redirected to a different node
• Client still manages a single volume group
VG-SQL-TGT0
VG-SQL-TGT1
VG-SQL-TGT2
17. 1717
Showing Legacy the Nutanix way
• Bare-metal apps can now access all capabilities of the Nutanix
scale-out storage fabric
• Extremely simple configuration using Volume Groups
• Unique high availability scheme without complicated multipath IO
(MPIO) setup and maintenance overheads
• A single application can utilize all cluster resources
• Dynamic load balancing based on real-time load analysis and
cluster topology changes
18. 1818
Load Distribution using Virtual Targets
Application Server
VG-
SQL
Disk 0
VG-
SQL
Disk 1
VG-
SQL
Disk 2
iSCSIInitiator
Node A
10.4.1.1
Node B
10.4.1.2
Node C
10.4.1.3
Data Services IP
10.4.1.100
VG-SQL-TGT0
VG-SQL-TGT1
VG-SQL-TGT2
Node D
10.4.1.4
• Node C overloaded
• Session migrated to idle Node D
19. 1919
All data reduction features supported
(Compression, Deduplication, Erasure Coding)
Thin Provisioned
SCSI UNMAP (TRIM) support
CHAP Authentication
Standard Block Storage Features and Storage Efficiency
20. 2020
• Exchange + iSCSI
• Shared Storage
Clustering
• MS SQL Clustering
• Oracle RAC
• Bare Metal
• Scale-Up
Performance
Use Cases & Deployment Modes
Host Cluster
Nutanix Storage Container
HostA
Target: “iqn...VGA”
Attach: “Iqn...HostA”
“Iqn...HostB”
VGA
vDiskA, vDiskB, vDiskC
HostB
Target: “iqn...VGA”
A B C
Disk0 Disk1 Disk2 Disk0 Disk1 Disk2
22. 2222
READ
Nutanix Blog: Related Content Published Soon!
Nutanix Bible: Acropolis Block Services
USE
Nutanix Sizer: Effectively Size your Target Environment !
Like What You Just Heard… There’s More!
Current list of clients we will qualify include RHEL 6+, Oracle Linux 6+ and Win 2012 R2/2008 R2
We will additionally qualify Oracle RAC and SQL
As you can see conceptually, an external server will load balance across all nodes within our cluster.
If a node fails, the client simply redirects traffic to other nodes in the cluster, so failover on failures is seamless
Cool IP around optimal path selection, load balancing, less client intervention, etc.
Inherits all Nutanix feature set
- With this feature, we are showing legacy the Nutanix way of managing datacenter infrastructure.
– Just because it is legacy, it does not need to be complicated or boring. Just like all other aspects o finfrastructure management on the Nutanix fablic, we have now made the complexity of provisioning and managing block storage invisible.
External applications can now access all the features and capabilities of the Nutanix scale out storage fabric that were previous available to hyper-converged applications.
Mike’s demo gave a preview of how easy it is to configure LUNs for an application using the volume groups construct.
To further simplify provisioning, we have created a unique high availability scheme that completely eliminates the need for multipath IO. (switch to next slice for MPIO example followed by the VIP diagram).
- With this feature, we are showing legacy the Nutanix way of managing datacenter infrastructure.
– Just because it is legacy, it does not need to be complicated or boring. Just like all other aspects o finfrastructure management on the Nutanix fablic, we have now made the complexity of provisioning and managing block storage invisible.
External applications can now access all the features and capabilities of the Nutanix scale out storage fabric that were previous available to hyper-converged applications.
Mike’s demo gave a preview of how easy it is to configure LUNs for an application using the volume groups construct.
To further simplify provisioning, we have created a unique high availability scheme that completely eliminates the need for multipath IO. (switch to next slice for MPIO example followed by the VIP diagram).