1. FusionCube Competitive Positioning

2. 1 2 3 Contents FusionCube Application Scenarios and Competing Products
Competitive Positioning - Virtualized Hyper-converged Infrastructure 3
Competitive Positioning - Database Hyper-converged Infrastructure

3. FusionCube Application Scenarios and Competing Products/Solutions
Huawei Product
Competing Product/Solution
Virtualization consolidation
VDI
Resource pooling
private cloud
FusionCube 6000
Nutanix NX series
VMware EVO:RAIL (EMC VxRail, Fujitsu PRIMEFLEX)
HP CS250
Simplivity OmniCube
EMC Vblock
NetApp/Cisco FlexPod
Hybrid workloads
High-performance virtualization application
FusionCube 9000
Data warehouse acceleration
database acceleration
database consolidation
Oracle ExaData
SAP HANA
For details, see SAP HANA appliance materials.

4. 1 2 3 Contents FusionCube Application Scenarios and Competing Products
Competitive Positioning - Virtualized Hyper-converged Infrastructure 3
Competitive Positioning - Database Hyper-converged Infrastructure

5. Top Beating Points of Nutanix NX Series
HTB: FusionCube vs Nutanix NX Series
Top Beating Points of Nutanix NX Series
Nutanix NX Series
FusionCube
Poor support for mission critical applications
Support only pure virtual environments. Do not support the physical machine deployment architecture or hybrid deployment architecture. Cannot meet the reliability and performance requirements of mission critical applications.
Support only 10GE/GE. The storage latency is high and cannot meet the performance requirements of mission critical applications.
Supports high-performance InfiniBand networks and hybrid deployment of virtual and physical machines. Provide excellent performance and high reliability, meeting requirements of mission critical applications.
Networks are not converged; poor integration; and complicated installation/deployment
External switches required for all models, low integration, many cable connections, complicated and time-consuming installation/deployment, low reliability.
Built-in switches, high integration, few cable connections, simple installation/deployment, high reliability
Low performance
NDFS-based, efficiency lower than block devices.
Support only GE and 10GE networks and do not support low-latency InfiniBand networks.
Provides block-based storage with an I/O efficiency higher than a file system.
Supports 2 x 56G QDR InfiniBand high-speed networks, delivering a bandwidth 5x+ higher than that of Nutanix NX series.
Small storage capacity in mainstream configurations
Four NX3060 nodes support 24 x 2.5" SATA HDDs.
A FusionCube 6000 chassis with 4 nodes supports 40 x 3.5" SAS/SATA HDDs..
High hardware failure rate
Use OEM low-end servers such as Taiwan SuperMicro servers and support only SATA disks, all delivering high hardware failure rates.
Uses Huawei servers, delivering high reliability.
Supports SAS disks whose failure rates are low.
Slow service response
Management systems and factory services are provided by multiple vendors, resulting in use complexity and low service efficiency.
Good service: Involves only one vendor, delivering high efficiency and convenient local vendor support.
Nutanix NX series: poor support for mission critical applications, no network convergence, difficult deployment, low reliability

6. How to Defend: Nutanix NX Series
Deduplication for better storage utilization:
Support 4 KB granularity fixed-length deduplication.
Compression for better storage capacity utilization: Use Google Snappy compression/decompression algorithms and virtual machine granularity compression. They support the Inline and Post-process compression algorithms.
Nutanix NX products introduce the complex deduplication and compression functions to better utilize hardware capacity but sacrifice system performance.
Deduplication and compression greatly increase the license cost for customers (Nutanix charges 24% of the total price for licenses). In addition, they require many CPU and RAM resources, leaving less resources for applications.
The improvement of storage utilization due to deduplication depends on system workloads: more improvement in scenarios such as VDI but less improvement in other scenarios.

7. Nutanix Product Overview
Advantages: Rich Advanced Features and Excellent Compatibility with Virtualization Software
New-Gen DC Network Architecture
High performance: SSD cache, data locality, SSDs as primary storage
Enriched advanced features: deduplication, compression, automated hierarchical storage, backup, disaster recovery, etc.
Scale-out storage architecture: adopt the copy mechanism, supporting 2 to 3 copies; support EC.
Compatible with virtualization software: VSphere, Hyper-V, KVM, self-developed virtualization software (KVM-based)
Flexible automated deployment: fast tool-based installation
Easy to manage: Software and hardware resources can be monitored by a single portal.
Single Storage Pool
Node = independent server
(CPU, memory, SSD, disk, 10GE network)
NDFS cluster

8. Nutanix hardware platform： 2U/4,2U/2, 2U/1 Server

9. Dual 10GbE, 2 x 1GbE, 1 x IPMI (10/100 Mb/s)
Nutanix Specification
NX-1000
Series
NX-3000
Series
NX-6000
Series
NX-7000
Series
NX-8000
Series
NX-9000
Series
Per Node
(4 per appliance)
(2 per appliance)
(1 per appliance)
Compute
Dual core
Sandy Bridge E5-2620v3
2.4/2.6GHz
Dual core
GHz
Dual 12 or 16 core
2.4 GHz
Dual 20 core
Ivy Bridge E5-2680v2
2.8GHz
Dual 8-28 core
Ivy Bridge E5-2680v3
GHz
Dual core
2.5 GHz
Cold
2 x 2/4/6 TB
2 x 2/4/6 TB
5 x 4/6TB
6 x 1 TB
Up to 20 x 1 TB
n/a
Hot
1 x 480GB/800GB/
1.2TB SSD
2 x 480GB/800GB/1.2TB/
1.6TB SSD
1.2TB/1.6TB SSD
2 x 400GB SSD (800 GB)
4 x 400 – 1600 GB SSD
6 x 800 – 1600 GB SSD
Memory GB
128 – 768 GB
32, 64, 128, 256 or 512GB
128 or 256GB
128 , 256 or 512GB GB
Network
Dual 10GbE, 2 x 1GbE
1 x IPMI (10/100 Mb/s)
Dual 10GbE, 2 x 1GbE, 1 x IPMI (10/100 Mb/s)
Power
Redundant 1100W, 110/1620W, 208V
Redundant 1620W, 208V
Please refer to latest Nutanix Datasheet

10. FusionCube: Flexible Architecture for Various Application Scenarios
Nutanix
FusionCube Center
Hypervisor
DB physical computing cluster
DBN
DBN
DBN
Computing cluster 1
Computing cluster 2 … VM VM VM VM … VM
10GE/InfiniBand
FusionStorage
FusionCube
Supports virtual machine deployment, physical machine deployment, and hybrid deployment
Supports only virtualization deployment

11. FusionCube: Shorter I/O Path, Better Storage Performance
Nutanix
User VM(s)
Hypervisor
SCSI Controller
SSD
SSD
SSD
SSD
SSD
High performance due to short I/O path (no bypass because Hypervisor integrates a storage engine)
Poor performance due to long I/O path: controller VM -> Hypervisor -> storage

12. FusionCube: Leading Hardware Specifications and Reliability
Nutanix
FusionCube
R&D capability
OEM SuperMicro low-end server
Independently developed product
Form factor
2U4, 2U2, or 2U1 rack
Huawei FusionServer server
CPU
2 x Intel E5 CPUs
2 or 4 Intel E5 or E7 CPUs
Memory
Up to 16 DIMMs
Up to 24 DIMMs
SSD
Up to 1.6 TB SSD cache
Up to 3.2 TB SSD cache
HDD
4 HDDs per node: less storage capacity, excessive computing resources
2.5-inch non-mainstream SATA disk
12–15 HDDs per node: easy to meet storage requirements
Mainstream 3.5-inch SATA or 2.5-inch SAS disk
Network
GE/10GE
GE/10GE/InfiniBand
Fan module
No redundancy.
Supports single -fan failures.
Management
Not support disk fault management.
Comprehensive hardware and software management mechanism
Per node
Computing: 16 cores, 16 DIMMs
Storage:
Flash: 400/800 GB SSD; HDD: 1 TB 2.5-inch SATA
Network: 2 x 10GE network ports
Controller: Nutanix controller VM

13. HTB: FusionCube vs EVO:RAIL
Top Beating Points of EVO:RAIL
EVO:RAIL(EMC, Fujitsu)
FusionCube
Poor support for mission critical applications
Supports only pure virtual environments. Do not support the physical machine deployment architecture or hybrid deployment architecture. Cannot meet the reliability and performance requirements of mission critical applications.
Support only 10GE/GE. The storage latency is high and cannot meet the performance requirements of mission critical applications.
Supports high-performance InfiniBand networks and hybrid deployment of virtual and physical machines. Provides excellent performance and high reliability, meeting requirements of mission critical applications.
Support only vSphere, limited application scenarios
Closed system, mainly bound with VMware. Not support other Hypervisor software such as KVM.
Open and compatible, various application scenarios: Supports multiple Hypervisor software including FusionSphere, VMware, and KVM. Supports physical deployment of databases such as Oracle databases, SAP HANA, DB2, and Sybase Iq.
Low performance
File system-based architecture, delivering an efficiency lower than block devices; supports only GE and 10GE networks, not low-latency InfiniBand networks.
Provides block-based storage with an I/O efficiency higher than a file system.
Supports 2 x 56G QDR InfiniBand high-speed networks, delivering high bandwidth and low latency.
High license fee
Supports only the VMware Enterprise Plus license.
Supports all VMware software licenses.
Fixed configurations,
high costs
At least four nodes to start up. Supports expansion only by four
nodes each time.
Three nodes to start up. Supports
single-node expansion.
Bad scalability
Supports a maximum of 64 nodes.
Supports a maximum of 256 nodes.
Slow service response
Slow factory service response: Involves products of three vendors (VMware, server vendor, and switch vendor), resulting in low factory service response.
Good service: Involves only one vendor, delivering high efficiency and convenient local vendor support.
EVO:RAIL does not support mission critical applications and supports only VMware virtualization.
It offers limited scenarios and low performance.

14. EVO:RAIL Overview Virtual SAN EVO:RAIL: Unique VMware Inside Ecosystem
Uses standard hardware in the industry. Local HDDs and SSDs are configured as a storage resource pool that functions as a virtual SAN, enabling storage sharing between VMs.
Application scenario: VDI, ROBO, DR-Target, development and testing
Pricing mode: Sold by CPU; prices lower than low-end and medium-range storages
VMware cooperates with EMC, Dell, Fujitsu, Inspur, SuperMicro, HP, HDS, and NetApp. Use high-density servers to integrate vSphere and VSAN. Hardware vendors provide differentiated backup and disaster recovery.
Application scenario: Small-sized virtualized DC, VDI, ROBO, and private cloud
Business model: integrated in and sold with hardware from partner hardware vendors (VMware Inside)
Virtual SAN: VM-level policies for simple management
EVO:RAIL Value: Easy to Use and Expand
Embedded in vSphere. Supports 3 to 64 nodes.
Provides copy mechanism to ensure reliability.
Uses SSD Cache to provide high performance. Allows SSDs to function as primary storage.
Automated VM granularity policy management.
Version 6.2 supports advanced features such as deduplication, compression, and EC.
All use VMware software.
Easy to use: 15-minute delivery at best
Low initial investment and easy to expand: Requires only one 2U high-density server to start up. Supports a maximum of four 2U high-density servers (each contains four 2-socket x86 servers).

15. FusionCube: Leading Performance
EVO:RAIL
Distributed Object-based RAIN10 Architecture
Server
Server
Server VM VM VM VM VM VM
Volume 1
Volume 2
Volume 3
FusionSphere
FusionSphere
FusionSphere
Client
Client
Client
Objects indicate vmdk files, snapshots, swap partitions, and namespaces. Each of them is called an object.
FusionStorage
VSAN uses a RAIN architecture or redundant array of independent nodes. Currently VSAN is a RAIN-10 architecture at object level rather than a disk level. This means it creates copies of objects stripped across host in the cluster. VSAN can tolerate a failure of a HDD, a network or a host.
5.1.1 Number of Failures to Tolerate
This property requires the storage object to tolerate at least NumberOfFailuresToTolerate of concurrent host,
network or disk failures in the cluster and still ensure availability of the object.
If this property is populated, it specifies that configurations must contain at least NumberOfFailuresToTolerate + 1
replicas and might also contain an additional number of witness objects to ensure that the object’s data is available
(to maintain quorum in cases of split-brain), even in the presence of as many as NumberOfFailuresToTolerate
concurrent host failures. Therefore, to tolerate n failures, at least (n + 1) copies of the object must exist and at
least (2n + 1) hosts are required.
NOTE: Any disk failure on a single host is treated as a “failure” for this metric. Therefore, the object cannot
persist if there is a disk failure on host A and another disk failure on host B when you have
NumberOfFailuresToTolerate set to 1.
FusionCube: Segment physical disks into several partitions (usually thousands of partitions) based on the striping algorithm. Data on a logical disk is stored in multiple partitions to maximize I/O concurrency among multiple disks.
EVO:RAIL: The distributed storage software VSAN adopts the RAID 10 mechanism to divide an object into 1–12 blocks (1 block by default) and then use the copy mechanism to copy blocks between two hard disks.

16. FusionCube: Flexible Architecture for Mission-critical Applications
EVO:RAIL
FusionCube Center
FusionCompute
DB physical computing cluster
DBN
DBN
Computing cluster 1
Computing cluster 2
DBN … VM VM VM VM … VM
10GE/InfiniBand
FusionStorage
FusionCube
Supports VM deployment, physical machine deployment, and hybrid deployment
Supports only VMware virtualization deployment

17. Top Beating Points of HP CS250
HTB: FusionCube vs HP CS250
Top Beating Points of HP CS250
HP CS250
FusionCube
Not support mission critical applications.
Supports only pure virtual environments. Do not support the physical machine deployment architecture or hybrid deployment architecture. Cannot meet the reliability and performance requirements of mission critical applications. Supports only 10GE/GE. The storage latency is high and cannot meet the performance requirements of mission critical applications.
Supports high-performance InfiniBand networks and hybrid deployment of virtual and physical machines. Provides excellent performance and high reliability, meeting requirements of mission critical applications.
Low reliability
No hardware sub-health check: Hard disk issues cannot be handled promptly.
Provides hard disk sub-health checks (read restoration, UNC, link intermittence, etc.).
Hard disk issues can be handled promptly.
RAID 5 or 6 must be rebuilt for a single node. If a hard disk fault occurs, the rebuilding takes a long time, affecting the performance of the entire cluster.
Data rebuilding involves all nodes and takes less than 30 minutes/TB.
Low performance
Distributed storage architecture based on disk RAID arrays.
Supports only GE and 10GE networks and do not support low-latency InfiniBand networks.
Does not support E v3 or above high-performance processors (due to server heat dissipation causes).
Block-based distributed storage architecture, P2P routing technology
Supports full-series high-performance processors and high-performance application requirements. FusionCube delivers 3x-5x performance than CS250 with the same hardware configuration.
Small storage capacity
A system with four nodes supports 24 x 2.5-inch disks, and the maximum capacity of a single node is 7.2 TB.
A single node supports 12 x 3.5-inch disks and a maximum capacity of 72 TB.
Compatible with SAS/SATA disks.

18. CS250 Distributed Storage Software: HP StoreVirtual Architecture (1)
RAID10
RAID5
Ethernet
Centralized management console: centralized management platform
Managers: cluster management module; manages data distribution and data re-distribution upon node faults. A management group contains 3 to 5 managers.
A management group can manage multiple clusters.
Volumes belong to clusters. Different volumes require different copy redundancy policies such as two copies, three copies, RAID 5, and RAID 6.
Standard ISCSI interfaces are provided for external entities.
Manager
RAID6
Cluster1
(SAS pool)
Manager
Manager
Management Group
Volume 1
Volume 2
Volume 3
• Control data replication. They are not directly in the data path and do not participate in I/O.
• Manage communication between storage systems in the cluster.
• Resynchronize data when storage systems change states.
• Coordinate reconfigurations as storage systems are brought up and taken offline.
Cluster2
(SATA pool)
Centralized management console
vol1

19. CS250 Distributed Storage Software: HP StoreVirtual Architecture (2)
RAID 5 (SAS disks) or RAID 6 (SATA disks) is created locally for a server (storage node).
Provides standard ISCSI interfaces. Each cluster provides a virtual target IP address. Client initiators connect to virtual target IP addresses and automatically redirect them to the corresponding physical IP addresses (based on the minimum number of connections).
For Windows clients, HP's special multi-pathing software selects storage nodes based on the block load balancing rule, similar to VBS.
Block size: 256 KB

20. How To Beat: Vblock/FlexPod vs. FusionCube
Top N Beating Point
EMC Vblock/FlexPod
FusionCube
High initial configuration and expensive, but poor applicability
Adopts a traditional architecture integrating blade servers, SANs, switches, and virtualization software. The minimum configuration is 24U or 42U with high specifications and high price, unsuitable for SMB and branch office scenarios.
Adopts a hyper-converged architecture that requires a minimum height of 4U, which is low configuration. Easy to expand, cost-effective, easy to manage, wide application scenarios.
Poor storage performance
Adopts the scale-up architecture with 8G FC ports, shared cache, capacity usually less than 100 GB, IOPS usually less than 1 million, and throughput usually less than 20 Gbit/s.
Adopts the scale-out distributed storage architecture that delivers good network performance (56G IB network with a latency less than 2 uS). The cache capacity increases linearly with the node quantity. The storage performance is outstanding: The IOPS per cabinet is 7.2 million, and the throughput per cabinet is 192 Gbit/s.
Difficult management
Split architecture, large number of components, complicated networking; multiple portals, complex management.
In-depth integration of computing, storage, network, virtualization, and management. Easy to configure and manage.
Bad scalability
Different types of storage and computing hardware is used for different specifications, making linear expansion impossible. The poor scalability also increases original investment waste risks.
Supports flexible hardware and software configuration, and hitless linear capacity expansion, protecting existing investments.
Slow service response
Components are provided by multiple vendors, making problem locating difficult and factory service response slow.
Good service: Factory services are provided by a single vendor, delivering high efficiency. Provides local vendor support.
Vblock/FlexPod: higher initial configuration, higher price, poor applicability, low performance, and complicated management

21. Vblock Overview Dominates the market with powerful partners
Various industry-specific solutions
VCE was established in 2009 with the investments from Cisco, EMC, VMware, and Intel. In 2015, EMC bought the shares of Cisco, and VCE became one department of EMC.
Vblock series: Cisco UCS network and server+EMC SAN+VMware virtualization product, covering all different market segments
The Vblock 2015 Q1 revenue reaches 330 million USD (IDC data), realizing a 29% year-on-year increase. Vblock has the No.1 share of the converged infrastructure market.
Customer benefits: simplified infrastructure installation, deployment, and management.

22. 24U and 42U Vblock 12U FusionCube 9000 4U FusionCube 6100
Vblock: High Initial Configuration, Space-Consuming, High Environment Requirements, Poor Adaptability
24U and 42U Vblock
12U FusionCube 9000
4U FusionCube 6100

23. Vblock: Split Architecture, Low Integration, Complex Networking
Huawei FusionCube
Vblock 340
2*10Gb/s
2*10Gb/s
IP Network
10Gb/s
Uplink 2*10G
Uplink
2*10G
2*10G
2*10G
10G
CX310
CX310
CX310
CX310
40GE
Stack
40GE
Stack …
10G
10G
Blade
Blade
Blade
Blade
Blade
Blade
FusionStorage
Chassis #0
Chassis #7
Vblock: separated storage, independent FC switches, low integration, and complex networking.

24. Vblock: Many Management Components, Difficult to Manage

25. Detail Comparison

26. Detailed Comparison: Product Specifications
Feature
Nutanix NX Series
Simplivity OmniCube
VMware EVO:RAIL
HP CS250
Huawei FusionCube
Supported
virtualization platform
vSphere, Hyper-V KVM, Acorpolicy
vSphere, Hyper-V
vSphere
vSphere, FusionSphere
Support separated deployment or not
Yes No
Support physical deployment or not
Nodes per device
1, 2, or 4 1 4
4, 8, or 16
Equipment dimensions
2U high-density or rack server
2U rack server
2U high-density server
4U high-density server FusionCube U blade server FusionCube 9000
Minimum number of nodes
3 nodes
1 node
4 nodes
3 nodes　
Maximum number of nodes
64 nodes
One data center: 8 blocks or 32 nodes
32 nodes
256 nodes
CPU
E5-2600
E5-2600, E7-4800
Storage
SAS, SATA, SSD
SAS, SATA
SAS, SSD
SAS, SAS+SSD
SATA, SAS, SSD
Cache
SSD cache for read SSD cache for write
SSD cache for read
SSD cache for write
4 GB battery-backed write cache per node
Network
10GE/GE
10GE/GE/IB
Capacity expansion mode
Expand one node at a time .
 Expand one node at a time.
Expand only one block (4 nodes) at a time .
Expand one node at a time.

27. Detailed Comparison: Storage System Features
Nutanix NX Series
Simplivity OmniCube
VMware EVO:RAIL
HP CS250
Huawei FusionCube
Architecture
Distributed file system
RAID + Copy
Distributed block storage
Storage protocol
NFS, SMB3, iSCSI
NFS
Internal protocol
iSCSI
SCSI, iSCSI
Number of copies
2 or 3 (configurable)
2 or 3
2-N (configurable)
Hardware RAID and software multi-copy mechanisms
Deduplication
Supports online deduplication.
Supports online deduplication, implemented by PCIe cards. No
Compression
Yes
Supports compression. All data can be compressed offline for a PCIe card.
Erasure code
Storage QoS
Support multiple storage clusters or not

28. Detailed Comparison: Data Protection and Service Continuity
Feature
Nutanix NX Series
Simplivity OmniCube
VMware EVO:RAIL
HP CS250
Huawei FusionCube
Snapshot
Yes
Backup:
local backup
remote backup
DR:
Host replication
synchronous storage replication No
DR: asynchronous storage replication
DR: stretch cluster
Cloud access

29. Detailed Comparison: Network Connection
Network Plane
Nutanix NX Series
Simplivity OmniCube
VMware EVO:RAIL
HP CS250
Huawei FusionCube
Storage plane
10GE or 10GE
10GE
10GE, 56 Gb/s InfiniBand
Management plane
10GE or 1GE
VM service plane

30. Detailed Comparison: Installation and System Management Functions
Feature
Nutanix NX Series
Simplivity OmniCube
VMware EVO:RAIL
HP CS250
Huawei FusionCube
GUI
CLI
Multi-cluster management
Supported
Not supported
Supported　
Vcenter plug-in
Supported. CLI commands are still available. No
API
REST API, Java SDK and Python bindings
REST API
REST API, CLI
Automatic installation
Installation time
(4 nodes)
15-60 minutes
1 hour or less
15 minutes or less
11 minutes at best

31. 1 2 3 Contents FusionCube Application Scenarios and Competing Products
Competitive Positioning - Virtualized Hyper-converged Infrastructure 3
Competitive Positioning - Database Hyper-converged Infrastructure

32. How To Beat: Oracle ExaData vs. FusionCube
Top N Beating Point
Oracle ExaData X5-2
FusionCube
Poor openness, few application scenarios
Supports only Oracle RAC 11g and 12c. Earlier versions are not supported.
Supports mainstream databases: SAP HANA, Oracle database, DB2, SQL server, Sybase IQ, etc.
Supports all Oracle database versions. Oracle RAC is not mandatory.
Application and the configuration are not flexible, and customer investments cannot be maximized.
Databases do not support virtualization deployment. Other applications cannot be deployed on ExaData.
Databases do not support virtualization installation. Other applications can be deployed in the same system (supports hybrid deployment).
Supports only four models 1/8, 1/4,1/2, and Full Rack. Database nodes can only be 2-socket servers not 4-socket servers. Does not support CPU and memory flexible configuration. Supports only 42U cabinets.
Supports flexible configurations based on customer requirements. Supports 4-socket database servers. Provides different storage, capacity, and other configurations based on CPU, memory, and network quantities and specifications. Supports 42U or 12U cabinets to meet different equipment room requirements.
Database licenses must be purchased and reused based on specifications.
Flexible license policies: You can purchase and reuse database licenses as required.
Low integration
42U height at least, low integration
12U height at least, high integration, saving over 50% footprint
Low price/performance ratio
FusionCube is 20% to 40% cheaper than ExaData with the same specifications and performance.
Oracle ExaData: few application scenarios, complex configuration, and low cost effectiveness

33. How to Defend: FusionCube vs. ExaData
Description
FusionCube does not support SQL uninstallation technologies such as SmartScan, StorageIndex, and HCC.
SQL uninstallation applies only to data warehouses and analysis-type queries. It does not improve most OLTP performance.
FusionCube provides SSD storage and 56G FDR distributed storage and delivers much better I/O performance than ExaData. In certain scenarios such as full table scanning that fully depends on I/O capability, FusionCube performs much better than ExaData.
FusionCube does not support compression.
The compression technology is used to improve database capacity, but a high compression ratio also greatly affects the performance of update operations. It offers little help in actual application and is seldom used.
ExaData supports storage QoS.
Storage QoS works only when multiple databases are integrated. It provides limited value in OLAP and OLTP scenarios.

34. Oracle X5-2 Overview

35. HANA FusionCube: Open for More Databases and OSs
Exadata supports only Oracle 11g and Oracle 12c.
FusionCube supports mainstream commercial databases:
Oracle database, IBM DB2, and Microsoft SQL Server
SAP certification, world's fastest HANA platform
Shared platform, not bound to a specific database
HANA
FusionCube supports more database types, database versions, and OS types

36. FusionCube: Flexible Configuration
CPU: 2P 4P
Cost
E5-2697V3
E5-2643V3
E7-4890V2
E5-2699V3 OR …
Exadata database appliance
E5-2667V3
E5-2630V3
E5-2660V3
Memory:
512 GB
1024 GB
256 GB
128 GB
FusionCube capacity expansion curve
SSD:
IT scale
2.4T
3.2T 2 4 6
FusionCube
Flexible configuration: Exadata provides only fixed configuration, but FusionCube allows flexible configuration of computing, storage, and CPUs. For example, you can use E high-frequency low-core CPUs to reduce license fees.
Flexible expansion: FusionCube supports flexible expansion by computing blade, storage blade, or cabinet. However, Exadata supports only fixed-configuration expansion.

37. FusionCube: Virtualization and Mixed Load
ExaData
FusionCube
Database
Oracle DB
SAP HANA, Oracle database, DB2, SQL Server, Sybase IQ
Application deployment
Not supported.
Supports the deployment of other applications.
Virtualization deployment
Supports application virtualization deployment.
Supports database virtualization deployment.
FusionCube: All in one box
Applications and middleware
running on VMs
Hosting physical database servers
FusionCube supports more enterprise applications and offers higher integration.

38. How To Beat: Server+FC SAN vs. FusionCube
Top N Beating Point
Server+FC SAN
FusionCube
Low integration
Split architecture
Convergence of computing, storage, and network. The 12U FusionCube appliance meets diverse requirements of most small- and medium-sized applications.
Poor storage performance
Adopts the scale-up architecture with 8G FC ports, shared cache, capacity usually less than 100 GB, IOPS usually less than 1 million, and throughput usually less than 20 Gbit/s.
Adopts the scale-out distributed storage architecture that delivers good network performance (56G IB network with a latency less than 2uS). The cache capacity increases linearly with the node quantity. The storage performance is outstanding. The IOPS per cabinet is 7.2 million, and the throughput per cabinet is 192 Gbit/s.
Low reliability
Many connection cables, low reliability
Converged architecture, few connection cables, high reliability
Difficult management
Separation management of computing, storage, and network, complex management, difficult fault locating
Unified management of computing, storage, and networks. Fault locating is easy, and the locating method is easy to learn.