Download Choosing a Computing Architecture

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Tandem Computers wikipedia , lookup

Transcript
Choosing a Computing
Architecture
Chapter 8
Architectural
Requirements
Scalability
Manageability
Flexibility
User
Budget
Availability
Extensibility
Integration
Business
Technology
Strategy for Architecture
Definition
Obtain existing architecture plans
Obtain existing capacity plans
Document existing interfaces
Prepare capacity plan
Prepare technical architecture
Document operating system requirements
Develop recovery plans
Develop security and control plans
Create architecture
Create technical risk assessment
Hardware Architecture
Involve all experts
New technology
Old technology
Networking
Hardware Architectures
Robust
Available
Reliable
Extensible
Scalable
Supportable
Recoverable
Parallel
VLM
64-bit
Connective
Open
Hardware Architectures
SMP
Cluster
MPP
NUMA
Hybrids use SMP and MPP
Evaluation Criteria
Determine the platform for your needs
SMP
Clusters
NUMA
MPP
High
Scalability
Low
High
Maturity
Low
Parallel Processing
Parallel daily
operations
Shared resources
- Memory
- Disk
- Nothing
Loosely or tightly
coupled
Database
Application
Hardware
Operating system
Making the Right Choice
Requirements differ from operational
systems
Benchmark
- Available from vendors
- Develop your own
- Use realistic queries
Scalability important
SMP
Communication by shared memory
Disk controllers accessible to all CPUs
Proven technology
CPU
CPU
CPU
CPU
Common bus
Shared memory
Shared disks
SMP
Benefits:
- High concurrency
- Workload balancing
- Moderate scalability
- Easy administration
Limitations:
- Memory (cluster for improvements)
- Bandwidth
NUMA
Loosely coupled shared memory
CPU
CPU
CPU
CPU
CPU
CPU
Shared bus
Shared
memory
Disk
Nonuniform
memory access
Shared
memory
Disk
NUMA
Benefits:
- Fully scalable, incremental additions to
disk, CPU, and bandwidth
- Performs better than MPP
- Suited for Oracle server
Limitations:
- The technology is new and less proven
- You need new tools for easy system
management
- NUMA is more expensive than SMP
Clusters
Node 1
CPU CPU CPU
Node 2
CPU CPU CPU
Node 3
CPU CPU CPU
Shared
memory
Shared
memory
Shared
memory
Common high-speed bus
Common high-speed bus
Clusters
 Shared disk, loosely coupled
 Dedicated memory
 High-speed bus
 Shared resources
 SMP node
 Benefits:
- High availability
- Single database concept, incremental
growth
 Limitations:
- Scalability, internode synchronization needed
 - Operating system overhead
MPP
CPU
CPU
CPU
CPU
Memory
Memory
Memory
Memory
Disk
Disk
Disk
Disk
MPP
A shared nothing architecture
Many nodes
Fast access
Exclusive memory on a node
Low cost per node
Scalable
nCUBE configuration
MPP Benefits
Unlimited incremental growth
Very scalable
Fast access
Low cost per node
Good for DSS
MPP Limitations
Rigid partitioning
Cache consistency
Restricted disk access
High memory cost per nodes
High management burden
Careful data placement
Windows NT
Architecture based on the client-server model
 Benefits:
- Include built-in Web services
- Scalability
- Ease of management and control
 Limitations:
- Not as secure
- Cannot execute programs remotely
- Lack linear scalability beyond four processors
- Addressing space for applications is limited to
two gigabytes
Architectural Tiers
Tiered structures:
- Modular
- Logical separation
Distributed structures:
- Two-tier
- Three-tier
- Four-tier (and more)
Middleware
Technologies for integration
Gateway
Database Server
Requirements
Robust
Available
Reliable
Extensible
Scalable
Supportable
Recoverable
Parallel
Parallelism
Database
Query
Load
Index
Sort
Backup
Recovery
Further Considerations
Optimization strategy
Partitioning strategy
Summarization strategy
Indexing techniques
Hardware and software scalability
Availability
Administration
Server Environments
Operational
servers
•Open DBMS
•Network, relational,
hierarchical
•Mainframe
proprietary DBMS
•Oracle, IMS, DB2,
VSAM, Rdb, Non
Stop SQL, RMS
Warehouse
servers
•Open DBMS
•Relational
•General purpose and
warehouse-specific
DBMS
•Oracle, Informix,
Sybase, IBM DB2,
NCR/AT&T Teradata
Red Brick
Data mart
servers
•Open DBMS
•Relational and
multidimensional
•General purpose
and warehouse
specific DBMS
•Oracle, Oracle
Express, Arbor
Essbase, MS SQL
Server, NT
Parallel Processing
A large task broken into smaller tasks:
Concurrent execution
One or more processors
Elapsed time
Processor 1
Parallel
Processor 1
Processor 2
Processor 3
Processor 4
Not parallel
Parallel Database
Increased speed
Improved scalability
Performance gains
- Availability
- Flexibility
- More users
Processor 1
Processor 2
Processor 3
Processor 4
Parallel
Parallel Query
SQL code split among server processes.
SubQuery
Query
SubQuery
SubQuery
Parallel Load
Bypass SQL processing to speed throughput.
Parallel Processing
Reduces the time to create
Allocates memory in cache efficiently
Runs simultaneously from any node
- Offline
- Online
Recovery Runs simultaneously from redo logs
Summaries Uses the CREATE TABLES AS
SELECT statement
Index
Sort
Backup
Summary
This lesson discussed the following topics:
Outlining the basic architecture
requirements for a warehouse
Highlighting the benefits and limitations of
all the different hardware architectures