© 2013 IBM Corporation
PureData for Transactions Overview
Fraser McArthur ([email protected]) Consulting IT Specialist, PureData Ecosystem Team
2
PureData
© 2013 IBM Corporation
Agenda
Information Technology Challenges PureSystems and PureData Introduction PureData for Transactions Technical Details
3
PureData
© 2013 IBM Corporation
Technology is the leading force for impacting businessIs IT ready for the challenge?
* 2012 IBM Global Data Center Study
68% of IT operating costs in 2013 will be for management and administration
Only 1 in 5 organizations allocate more than
50% of IT budget to new projects
2013
68%29%
1996
* IDC; Converged Systems: End-User Survey Results presentation; September 2012; Doc #236966
New server spendingPower & cooling costsServer mgmt & admin costs
4
PureData
© 2013 IBM Corporation© 2012 IBM Corporation4
Clients struggle to overcome barriers of time, cost and risk
Top Causes of Project DelaysHardware
Troubleshooting and tuning production environment
Integration, configuration and testing of the infrastructure
Installation, cabling and network access for the environment
SoftwareIntegration, configuration and testing of applications
Integration, configuration and testing of middleware
Configuration, build and deployment of applications
45%
45%
29%
41%
35%
34%
Phase Time (days) Budget
Specify/design 73 - 96 14% - 16%
Procure 57 - 112 19% - 21%
Implement 74 – 93 12%
Configure/test 74 – 80 10% - 11%
Cluster & HA 66 – 104 11% - 12%
Backup 44 – 108 10%
Tune 89 – 98 9% - 10%
Management 67 – 110 9 – 10%
34% of new IT projects (US) deploy late
Typical IT Project Time and Budget
From a commissioned study conducted by Forrester Consulting on behalf of IBM
5
PureData
© 2013 IBM Corporation
Let’s view IT from a new perspectiveSystems with integrated expertise and built for cloud
Integrated by DesignDeeply integrate and tune hardware and software
Built-in ExpertiseCapture and automate what experts do
Simplify the Experience
Make every part of the IT lifecycle easier
Get up and running faster: from 4 – 7 months to hours!
Streamline development and operations: months to
minutes (with repeatability)
Optimize ongoing effort: single management console;
no downtime upgrades!
6
PureData
© 2013 IBM Corporation
• Workload optimized performance
• Data load ready in hours
• Integrated management
• Automated maintenance
• Single point of support
Delivering Big Data Platform Services
7
PureData
© 2013 IBM Corporation
Real Time Fraud DetectionCustomer AnalysisE-commerce
Transaction Processing
Reporting and Analytics
Operational Analytics
Different applications rely on different data processing
7
8
PureData
© 2013 IBM Corporation
Transaction Processing
Reporting and Analytics
Scalable Transactional Database
Different data workloads require different solutions
8
Operational Analytics
9
PureData
© 2013 IBM Corporation
• Select model– for Transactions– for Analytics– for Operational Analytics– for Hadoop
• Select size• Load data
Data Platform
Delivering Data Services
Simplicity & Optimizationout-of-the-box
10
PureData
© 2013 IBM Corporation
IBM PureData System for TransactionsOptimized exclusively for transactional data workloads
Delivering data services for transactions
System for Transactions
Speed Industry leading DB2 performance Database node recovery in seconds1
Simplicity Database deployment in minutes, not hours1 Runs multiple database software versions Handles more than 100 databases on 1 system2
No planned downtime for firmware / OS upgrades1
Scalability Scaling up to 30x3
Designed to expand from small to medium to large configurations with no planned system downtime
Smart Supports Oracle Database apps with minimal
change; supports DB2 applications unchanged Clients have experienced cases of 10x storage
space savings via Adaptive Compression4
Footnotes:1. Based on IBM internal tests and system design for normal operation under expected typical workload. Individual results may vary.2. Based on one large configuration.3. Based on the designed minimum and maximum processor and memory resources required for a single database.4. Based on client testing in the DB2 10 Early Access Program.
11
PureData
© 2013 IBM Corporation
Three Configurations Available
Configurations T1500-96Small
¼ Rack
T1500-192Medium½ Rack
T1500-384Large
Full Rack
Chassis 1 1 2
Compute Nodes(16 cores per node)
6 12 24
CPU Cores 96 192 384
# of clusters/ databases 3/ 30 6/ 60 12/ 120
Memory 1.5 TB 3.1 TB 6.1 TB
V7000 Storage Unit(each unit has: 18 x 900GB
HDD, 6 x 400 GB SSD)
1 2 4
V7000 Storage Expansion(each unit has: 18 x 900GB
HDD, 6 x 400 GB SSD)
1 2 4
User Capacity Raw SSD Storage Raw HDD Storage
18.6 TB4.8 TB
32.0 TB
37.2 TB9.6 TB
64.0 TB
74.4 TB19.2 TB128.0 TB
Upgrade Upgrade
12
PureData
© 2013 IBM Corporation
Components – PureSystem Manager (PSM)
PureSystem™ Manager (PSM) Provides Integrated management for all
system resources Built-in Pattern Deployment 2 PSMs per rack for redundancy via active /
passive mode and virtual IP-Address Hosts the Optim® Performance Manager VM
42U41U40U39U38U37U36U35U34U33U32U31U30U29U28U27U26U25U24U23U22U21U20U19U18U17U16U15U14U13U12U11U10U09U08U07U06U05U04U03U02U01U
7189 PDU
7189 PDU
7189 PDU
7189 PDU
PSM 2PSM 1
System Console
BNT Top of Rack Switch
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
BNT Top of Rack Switch
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
System Console System Console to operate the PureSystem
Manager component
Large Rack
13
PureData
© 2013 IBM Corporation
Components – Flex System & Compute Nodes
Flex System™ Chassis 1 for Small/Medium and 2 for Large configuration
Each holds up to 12 Compute Nodes (see below)
Back plane with fully redundant Network (En) and Storage connectivity
Flex System™ Compute Nodes Up to 24 Compute Nodes per rack, hot
swappable
Serve as nodes for DB2 pureScale running as CFs or DB2 Nodes with:
Red Hat Enterprise Linux® 6.2 IBM DB2 ESE 10.1 FP1
Storage optimization and WLM DB2 pureScale TSM for backup and recovery
42U41U40U39U38U37U36U35U34U33U32U31U30U29U28U27U26U25U24U23U22U21U20U19U18U17U16U15U14U13U12U11U10U09U08U07U06U05U04U03U02U01U
7189 PDU
7189 PDU
7189 PDU
7189 PDU
PSM 2PSM 1
System Console
BNT Top of Rack Switch
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
BNT Top of Rack Switch
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Large Rack
14
PureData
© 2013 IBM Corporation
Components - Storage
42U41U40U39U38U37U36U35U34U33U32U31U30U29U28U27U26U25U24U23U22U21U20U19U18U17U16U15U14U13U12U11U10U09U08U07U06U05U04U03U02U01U
7189 PDU
7189 PDU
7189 PDU
7189 PDU
PSM 2PSM 1
System Console
BNT Top of Rack Switch
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
BNT Top of Rack Switch
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Storage System Storewiz v7000 controller and
expansion units
Up to 192 disks in large configuration.
Storage Network Use of multiple switches & physical
connections
Balanced HDD & SSD Setup 48-disk modules (12 SSD + 36 HDD) to
optimize for performance & cost
Speed & Redundancy Setup via RAID 10
4.8 TB SSD and 32.4 TB HDD capacity in a pair
Large Rack
15
PureData
© 2013 IBM Corporation
Components – Network Connectivity
42U41U40U39U38U37U36U35U34U33U32U31U30U29U28U27U26U25U24U23U22U21U20U19U18U17U16U15U14U13U12U11U10U09U08U07U06U05U04U03U02U01U
7189 PDU
7189 PDU
7189 PDU
7189 PDU
PSM 2PSM 1
System Console
BNT Top of Rack Switch
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
V7000 Expansion
V7000 Controller
BNT Top of Rack Switch
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Compute ITE
Compute ITE Compute ITE
Network Connectivity Dual 10Gb Ethernet Switches for external
communication.
− 48 ports SFP+, 4 ports QSFP+
Increased Redundancy and Bandwidth through the use multiple switches and physical connections
− Failure of cable or switch leads to reduced bandwidth - not reduced service
Communication Networks are separated for different duties:
− Customer Data Network used by customer applications to connect to database
− Customer Management Network for remote management access
− DB2 pureScale Cluster Interconnect
− Internal PureData Management Network
Large Rack
16
PureData
© 2013 IBM Corporation
Cluster Elasticity Predefined cluster deployment patterns to handle
different transaction requirements
PureData system supports multiple pureScale clusters in a single system– Up to 3 pureScale clusters for the small configuration– Up to 6 for medium and 12 clusters for large
Shrinking of clusters is currently not supported (e.g. shrinking from 6 to 4 nodes)
Up to 10 DBs in a cluster.
* CFs and DB2 nodes are co-located and therefor cannot grow to a 4 node cluster.
cluster Size
# of Compute
Nodes
CFs DB2 Nodes# CPU
Cores# CPU
Cores
LARGE 6 2 32 4 64
MEDIUM 4 2 32 2 32SMALL 2 2* 16 2* 16
17
PureData
© 2013 IBM Corporation
PureData for Transactions: Disaster Recovery
• Disaster Recovery is provided at the Database Level • Replication can be from one source to multiple targets• Source and target machines are not required to be identical• Data synchronization is maintained for an entire database or subset of objects• Data Synchronization Server (Q-replication or CDC) is running on a separate
machine
Data Synchronizationvia
Q-Replication or CDCData Synchronization Server
OR
Any server running DB2
PureData forTransactions
PureData forTransactions
© 2013 IBM Corporation
Simple System Management
19
PureData
© 2013 IBM Corporation
Patterns and Deployment
Pattern: pre-defined configuration that can be easily deployed
Cluster Patterns− Predefined database cluster architecture and configuration− Automatically creates, configures, and deploys
a cluster topology with built-in redundancy and high performance
− Operating system + all required software components
− Fully integrated, configured, and tunedfor a particular purpose
Database Patterns Configuration and best practices applied to a
database for a specific context Built-in pattern: Transactional (OLTP)
DB2 pureScale cluster options
20
PureData
© 2013 IBM Corporation
Simplified and Integrated System Management
Single console to manage all resources and work running on the system– Consistent console across
IBM PureSystems
Role-based securityand tasks – Management– Monitoring – Maintenance
Easy integration with broader enterprise monitoring tools and processes
Two main views– System Console– Workload Console
IBM PureData System for Transactions IBM PureData System for Transactions
IBM PureData System for Transactions – Monitoring – Database IBM PureData System for Transactions – Monitoring – Database
21
PureData
© 2013 IBM Corporation
System Console
System configuration Hardware monitoring
− down to a component level Console authentication
users and groups Maintenance functions
21
22
PureData
© 2013 IBM Corporation
Workload Console
Managing DB2 pureScale clusters & databases
23
PureData
© 2013 IBM Corporation
11
22
33
44
Creating a database cluster
24
PureData
© 2013 IBM Corporation
11
22
33
44
Creating a database within the cluster
25
PureData
© 2013 IBM Corporation
DeployDeploy Softwareapplication
HA Proxy
Caching Server Database
Application Server
HA Proxy
Application Server
Caching Server
Database
Simplified deployment across systems
Do you require High Availability
deployment ?
Do you require High Availability
deployment ?
© 2013 IBM Corporation
DB2 pureScale for High Availability and Performance
27
PureData
© 2013 IBM Corporation
Software ArchitectureDriving principles for the design of PureData System for Transactions Extreme Capacity
− Designed for OLTP− Grow easily with your business− Move from small to medium to large
Application Transparency− Avoid the risk and cost of changing applications
because of the database topology− Automatic Work Load Balancing, Automatic Client
Reroute, and Client Affinity
Continuous Availability− Maintain service across unplanned events− Maintain service across planned events using
maintenance
28
PureData
© 2013 IBM Corporation
Technology Overview
DB2 engine runs on several host computers• Co-operate with each other to provide coherent access to the database
from any DB2 node
Data sharing architecture • Shared access to database• DB2 nodes write to their own logs• Logs accessible from another host (used during recovery)
Cluster caching facility (CF) from STG• Efficient global locking and buffer management• Synchronous duplexing to secondary ensures availability
Low latency, high speed interconnect• Special optimizations provide significant advantages on RDMA-capable
interconnects (10 Gigabit Ethernet (RoCE) or InfiniBand)
Clients connect anywhere…… see single database
• Clients connect into any DB2 node• Automatic load balancing and client reroute may change
underlying physical DB2 node to which client is connected
Integrated cluster services• Failure detection, recovery automation, cluster file system• In partnership with STG (GPFS,RSCT) and Tivoli (SA MP)
Leverages IBM’s System z Sysplex Experience and Know-How
Single Database View
DB2 DB2 DB2
Shared Storage Access
PrimaryCF
SecondaryCF
Log Log Log Log
Database
DB2
DB2 Node DB2 Node DB2 Node DB2 Node
CSCS CS
CS CS
CS
29
PureData
© 2013 IBM Corporation
Scale with Ease
Without changing applications Efficient coherency protocols
designed to scale withoutapplication change
Applications automatically andtransparently workload balancedacross DB2 nodes
Without administrativecomplexity No data redistribution required
Single Database View
DB2 DB2
Shared Storage Access
PrimaryCF
SecondaryCF
Log Log Log Log
Database
DB2
DB2 Node DB2 Node DB2 Node
DB2
DB2 Node
30
PureData
© 2013 IBM Corporation
Advantages of RDMA – An Example
RDMA over low latency fabric• Direct memory access• Enables short response time :
~10-15 microseconds Silent Invalidation
• Informs DB2 nodes of page updates, requiring no CPU cycles on those nodes
• No interrupt, No IP Socket Calls, No context switching, or other message processing
• Increasingly important as cluster growsHot pages available without disk I/O from GBP memory
• RDMA and dedicated threads enable read page operations in ~10s of microseconds
Node 1
Bufferpool
Node 2
Bufferpool
501
501
Node 3
Bufferpool
501
Node 4
Bufferpool
501
501
CF
GBP
GLM
501
Node 1 requests lock on page 501
Lock granted!
501
501
Silent Invalidation
31
PureData
© 2013 IBM Corporation
Online Recovery
DB2 pureScale design point is to maximize availability during failure recovery processing
When a DB2 node fails, only in-flight data remains locked until node recovery completes In-flight = data being updated on the
failed node at the time it failed
Target time to row availability <20 seconds
32
PureData
© 2013 IBM Corporation
Client Connectivity and Workload Balancing
Run-time load information used to automatically balance load across the cluster − Shares design with system z Sysplex− Load information of all DB2 nodes kept on each node− Piggy-backed to clients regularly− Used to route next connection (or optionally next transaction) to least loaded DB2 node− Routing occurs automatically (transparent to application)
Failover− Load of failed DB2 node evenly distributed to surviving nodes automatically
Fallback− Once the failed DB2 node is back online, fallback does the reverse
© 2013 IBM Corporation
Security Model
34
PureData
© 2013 IBM Corporation
Authentication
User authentication and group membership External (LDAP) – preferred
− Enforce password rules and policies− Centralized account management− Requires an external LDAP server (e.g. MS Active Directory)
− Internal (OS)− Functional IDs exempt
from policies Mix and match
35
PureData
© 2013 IBM Corporation
System-level Security
Access to System Console and Workload Console Separation of duties
− Controls access to specific console objects, properties, and actions− Workload and database configuration management− Administration of resources: Workloads, cloud groups, hardware, audit, security− Assign “Read-only” or “Full Permission” to users or groups
System Console System Users, User Groups
36
PureData
© 2013 IBM Corporation
Instance-level Security Controls what users/groups have access to instance and databases within Grant instance access to console users via Workload Console
− Menu Databases DB2 pureScale instances select an instance− cluster owner user (SYSADM) is created during instance deployment− “Access granted to” field: defines who can manage the instance− “Instance available to” field: defines who can deploy database in that instance
Create more users via Workload Console− Menu Databases DB2 pureScale instances select an instance Manage− Creates users and groups in OS or LDAP− Can grant SSH access to Compute Nodes
36
37
PureData
© 2013 IBM Corporation
Database-level Security Instance owner has SECADM and DBADM authorities.
− SECADM is revoked from instance owner and only default DB user has this right after database is deployed.
Create users and groups via Workload Console− Database Databases <select database> Manage− Created in OS or LDAP− Can grant SSH access to compute nodes
Grant/revoke permissions as usual in DB2 − DB2 CLP or IBM Data Studio
37
38
PureData
© 2013 IBM Corporation
Thanks!
Questions?Please feel free to reach out to me anytime, if questions come to mind:
Fraser McArthur ([email protected])