This document provides an overview of an in-memory database company and its product capabilities. It discusses the company's history and growth, the changing data landscape driving demand for real-time analytics, and how the company's in-memory and hybrid database technologies provide extremely fast transaction processing, high availability, scalability, and flexibility for deploying on-premise or in the cloud. Example customer use cases and implementations are described to demonstrate how the database has helped organizations tackle challenges of high volume data processing and analytics.
3. + Company Overview
Company Description
Provider of in-memory data solutions for real-time access, analysis, and
distribution of high volumes of data in mission-critical environments
Founded 1999
Privately Owned
200+ Employees
4. + Company Overview
Milestones
1991 – Research begins in SK about future effects that RAM will have on
RDBMSs
1999 – Founded as a private in-memory RDBMS provider
2000 – Acquisition of first client: Hyundai
2005:
HDB, a hybrid DBMS with both memory and disk storage, is released
Altibase acquires its 100th client and 500th deployment
2006 – Support for Spatial data (GEOMETRY Data Type) is added
2009 – Complex Events Processing (CEP) engine is developed
2013 – 2014:
Altibase acquires its 500th client and 4000th deployment
Created partnerships will major companies: Dell; Intel; RedHat; Amazon
Listed in Gartner’s Magic Quadrant for Operational DBMSs
XDB, a DBMS optimized for in-memory only, is released
6. + Changing World of Data
Data volumes explode:
Big data: machine generated, continuous
Less structure
Making sense of unstructured data – databases on the TB scale for
metadata
Massive parallelization:
Commodity hardware & appliances: SANs, File Servers
Software: Hadoop, Sharding etc..
OLTP and OLAP
Both are necessary
Performance is key
7. +
Interactivity/Speed
Availability
Scalability
CRM, ERP, SFA
POS, ATM
Custom Apps
Data Warehouse
Data Marts
Reporting Apps
Bandwidth
Customization
Scalability
Transactional & Analytical
OLAP - Analytical Systems OLTP - Transactional Systems
Provides an enterprise with data to act on
Many users
Continuous updates
Tactical activity
Many short transactions
MB - TB of data
Mission critical
Operational Data for day-to-day
Provides an enterprise with answers
Few users
Batch updates
Strategic planning
Long complex lookups
TB - PB of data
Important for Audits
Analytical Data for decision making
8. +
Enterprises have traditionally needed separate data stores
for these technologies. However, as data sizes are growing,
the need for real-time transactional analytics has become a
reality
Databases need to provide applications
with the ability to process data very quickly
and reliably
Databases need to provide applications
with the ability to access large volumes of
data at one time
Transactional & Analytical
OLAP OLTP
9. + Why Companies are turning to Altibase
Extensive knowledge of and
commitment to in-memory databases
Extremely fast response times
Predictable and consistent
Low latency
Throughput scalability
Real time replication
Persistent and durable
Flexible architecture for cloud
deployment
Scalable on commodity platforms
SQL-92 compliant
Rich features interfaces
Proven technology
Highly available and reliable
Drivers for all DB connectivity standards
Hybrid (in-memory + on-disk) architecture removes the need to choose
between speed and size, transactional and analytical. One data source, one
platform, and one license provides developers a tool set that allows for one
central target for both high-speed and analytical platforms
10. + How customers are using Altibase
OCS (Online Charging Systems)
Memory tablespaces enable simultaneous management of call access and customer balances in real
time while permanently storing data
APM (Application Performance Monitoring)
Real-time status monitoring and control between standard web services becomes simple using the
speed of in-memory
EES (Equipment Engineering Systems)
Tracking defects and changing requirements in real time while performing analytics increases yield by
opening up dynamic changes to manufacturing
Location Based Service
Tracking and matching data is completed with ease in-memory. Permanent storage is crucial in the
public sector. HDB provides both.
IP Authentication
Access routing, address assignment, and authentication are important security barriers. High speed and
huge storage is a must for proper operation.
Futures/Options Trading
High and stable performance large amounts of financial data is how we were born
11. + Example Applications
Risk Management
Fraud Detection
Algorithmic Trading
Security Intelligence
Supply Chain Tracking
Telecom / Media Revenue Leakage
Service Delivery
Online Gaming
Inventory Forecasting
Transportations Operations
Management
Software-as-a-Service
Real time Analytics
Profitability Analysis
Global Web Commerce
Sales Incentive Promotions
Management
…. Many More
13. + In-Memory Database Technology
Extremely Fast Transaction processing
Entire database resides in computer’s memory
Powered by special algorithms and data
structures that are highly optimized for in-
memory computing
Hundreds of thousands of transactions per
second
Short and Predictable Response times
Optimized for fastest transactional processing
with the shortest response times measured
in microseconds
The improved response times fuel High
Throughput.
Connectivity
In-Memory Database
Application
Query Processor
Storage Manager
DRAM
Transactional Log Checkpoint
14. + Persistent In-Memory Database
Persistent and Durable In-Memory DBMS
Full ACID support for all database transactions
Atomicity
Consistency
Isolation
Durability
Durability is achieved via use of transaction
logs and checkpoint images
Fully Recoverable
Multiple Durability Levels to control a balance
between performance and durability
No Durability
Relaxed Durability
Enhanced Durability
Strict Durability
Connectivity
Application
Application
Transactional Log Checkpoint
15. + Highly Available In-Memory Database
High Availability via built-in
Replication Feature
Log-based, TCP/IP Replication
Additional layer of Durability
Adaptive Consistency
Synchronous Replication
Asynchronous Replication
Nonstop Service Architecture
Active-Active
Active-Standby
Near Standalone Replication
Performance
90% in Active-Active
96% Active-Standby
Conflict Detection and Resolution
Offline Replication
Application
Replication
16. + Highly Scalable In-Memory Database
Horizontal Scalability via built-in
Replication Feature
Leveraging TCP/IP protocol
Unlimited nodes
Flexible Load Balancing Architecture
Vertical Scalability
Scales On Commodity Platforms
Increased RAM
Increased CPU
Dynamic sizing of In-Memory database
with no system downtime via
AutoExtend feature
Horizontal Scaling via Replication
Vertical Scaling (CPU, RAM)
18. + XDB – Optimized for In-Memory
Customizable application performance via
Innovative and Rich Interfaces
Conventional client/server protocols TCP/IP and
IPC for compatibility (1)
Direct Access Mode to completely eliminate
network overhead (2)
Direct Access API Mode eliminates not only
network overhead but also query processing
overhead (3)
TCP/IP or IPC
Application
Application
Query Processor
Storage Manager
1
23
Transactional Log Checkpoint
19. + HDB – Superior Deployment Flexibility
Hybrid Architecture
Combines the benefits of in-memory storage
and on-disk storage in a single relational
database
Flexible Deployment Modes
In-Memory Database Only
On-Disk Database Only
Hybrid Database (In-Memory + On-Disk)
Support for different workloads
Real-time access to time critical Hot data
Access to historical Cold data for analytics
Complex transactions through integrated data
Easy bidirectional data migration between Hot
and Cold data zones
Memory Data
In-Memory DBMS
Disk Data
Buffer
Disk DBMS
Disk Data
Buffer
Memory
Data
Hybrid DBMS
Data Size
Speed
20. + Standards Compliant In-Memory Database
Support for SQL Standards
SQL:1999
Support for all common data types
Support Database Connectivity Standards
ODBC (Microsoft 3.5.1 API)
JDBC (Type 2 & 4)
.NET Provider
.NET Entity Framework
OLE DB
Embedded SQL
CLI
Perl DBD
Support for common communication protocols
TCP/IP (IPv4 and IPv6)
Unix Domain Socket
IPC (Shared Memory)
Application
SQL ODBC JDBC
OLEDB.Net CLI
TCP/IP UDS IPC
22. + Open Platform In-Memory Database
Sun Solaris OS
SPARC (64bit)
Intel (64bit)
Intel (32bit) – Client Only
HP HPUX
PARISC (64bit)
PARISC (32bit) – Client Only
IA (64bit)
IA (32bit) – Client Only
IBM AIX
PowerPC (64bit)
PowerPC (32-bit) – Client Only
Linux
Intel (32bit)
Intel (64bit)
Microsoft Windows
Intel (32bit)
Intel (64bit)
23. + In-Memory Database For The Cloud
Altibase DBaaS (Database as a Service) enables:
Large Enterprise Customers for consolidation of data management
Small/Medium Business Customers for outsourcing data management
Altibase DBaaS Benefits
Provisioning
Ease of installation and configuration
Amazon AMI on RHEL 6.4
OpenShift Gear
Docker Container
High Performance
Low Latency
High Throughput
Scalability
Ease of administration
Elasticity
Monitoring/Tuning
Scalability
High Availability
24. + CEP – In-Memory Middleware
ALTIBASE HDB
Or
ALTIBASE XDB
ALTIBASE CEP
Data & Event
Sources
Other Systems
Dropped
Data
ALTIBASE CEP is an In-Memory Middleware for
real-time processing, querying and analyzing data
streams
Data Characteristics
Continuously flowing
Too large to store
Rapidly changing
Time-sensitive
Extremely fast processing/filtering
Continuous Query Processing (CQL)
Continuous execution of registered SQL queries
Bounded ranges such as time windows or tuples
Support for JOIN operations between streaming or
persistent data sets
Publish-Subscriber Model
Tight integration with ALTIBASE HDB and XDB
Data & Event Sources
Transactional Systems
Sensors
Mobile Devices
RFID ……….
26. + Problems
HP OpenMCM combines several types of monitoring systems, including
Mission Critical Management, Application Process Monitoring, and
Marketing Communication Management. As such, there was significant
load on their monitoring servers, as well as exponentially growing data
storage from these monitoring triggers.
HP’s OpenMCM data processing requirement was a minimum of 30,000 transactions
per second (TPS).
HP then tested another company’s in-memory processing technology which
produced better performance, but still fell short of the mark with only 20,000 TPS.
TCO presented a massive hurdle. System memory limitations of an in-memory only
DBMS necessitated the purchase of a supplemental on-disk DBMS.
Facing such pervasive impediments, HP could not implement key, additional
features, such as real-time analytics.
27. + Solution
Born from ALTIBASE HDB’s core capabilities, speed, and limitless storage size, HP
OpenMCM possesses robust features, including transactional performance of
over 45K TPS and real-time analytics.
Because of the ability to support application loads with significantly less
infrastructure, combined with a more cost effective licensing model, HP gained
tremendously lower TCO.
28. The Company: Korea Telecom
The Problem: Abnormal Traffic
Detection, Analysis, and Control system
29. + Problems
In order to maintain normal operating speeds in their network, KT
used a set of applications that detected abnormal web traffic
patterns (botnets & DDoS, malware), but as their user base
increased, their disk-resident database was unable to keep up with
the detection and logging associated with these services:
The old system configuration was unstable when high volumes of data
were being processed. KT tried third-party solutions with the hopes of
increasing overall performance. While their performance did
increase, it still did not solve the issues, and also introduced instability
that did not exist previously.
Upgrading the system for this type of expansion was difficult to
implement and came with a hefty price tag. The total related costs
were untenable.
30. + Solution
Altibase HybridDB was a perfect fit for this specific set of problems:
The ability to store, write, and retrieve data directly from memory
tables completely resolved all issues regarding lag due to detection and
storage.
With an exponentially larger amount of speed available, KT was able to
expand their applications to detect more kinds of threats in realtime.
HDB’s hybrid functionality also provided KT with the ability to store all
of the gathered data permanently on disk, thus allowing them to use
the historical data to increase preemptive detection based on past
threats
31. + Results
KT was empowered with the ability to analyze data in real-time and react
accordingly.
KT has profited from ALTIBASE HDB’s Hybrid architecture capitalizing on
in-memory storage for frequently accessed, current data while keeping
less active, historical data on-disk.
KT’s reputation, bolstered by unilateral increases in customer
satisfaction, loyalty, attraction and retention, is at an all-time high.
KT was able to move away from batch processing to real-time processing.
ALITBASE HDB’s built-in replication feature and resultant Active-Standby
system allowed KT to deliver real-time and non-stop service.
The number of active servers has been reduced to 4 from over 20 servers.
System resource usage is down by 65%.
By using the parallel architecture model, on-demand system expansion
became a reality.
33. + Problems
Built on an Oracle DBMS backend, KT’s Wired Telephone SMS Service
was overrun with heavy traffic and stability issues. Transactions that
were being processed per second were on the rise, and KT was in
immediate need of replacing their legacy system with a significantly
more robust solution. Three areas called for a rapid overhaul:
The current system configuration restricted usable data management
and flow. Data loss occurred due to a lag in data migration when the
active system failed.
With piling demands on the legacy system, and reactive attempts for
resolutions, total cost of operating and maintaining the system became
untenable.
KT’s system infrastructure found itself obsolete. Data management and
interrelated processes were timeworn and further, associated hardware
was becoming non-responsive.
34. + Solution
KT implemented ALTIBASE HDB In-Memory DBMS into their
Wired Telephone SMS Service in 2010. ALTIBASE HDB’s built-in
support of Geographic Information System (GIS) functionalities
became a natural augment to KT’s business.
ALTIBASE HDB enabled KT to utilize real-time detection and
analysis while maintaining data integrity and high availability with
the built-in data replication features.
35. + Results
The number of active servers was reduced from 30 to 4.
System resource usage decreased by 73%. Active demand on the
system was reduced to 15% from 88%.
KT’s IT spending was reduced significantly, while performance and
system availability was dramatically increased.
KT’s reputation, bolstered by unilateral increases in customer
satisfaction, loyalty, attraction, and retention, is at an all-time high.
KT’s processing speed was increased by over 600%. Before
deploying ALTIBASE HDB, processing speeds were 100 transactions
per second. HDB was able to deliver over 600 TPS.
Even in the event of an unexpected server crash, KT provides
uninterrupted service due to HDB’s built-in replication.
36. The Company: Ministry of
Land, Transport, and Maritime
Affairs
The Problem: National Spatial Data Infrastructure
37. + Problems
MLTM hosts several spatial information systems that are used by
utility providers and other regional government departments.
Fragmentation amongst these systems made integration and
interconnectivity nearly impossible.
Lacking a single interface and corresponding integration, data access was
constrained. As a result, the availability of essential, spatial information was
unreliable.
Existing spatial data was duplicated across multiple systems to facilitate load
balancing and application availability. This redundancy wasted storage space and
created inconsistencies.
The independent systems, with their data overlap, drove operating expenses up
without merit.
38. + Solution
MLTM implemented ALITBASE HDB In-Memory DBMS to serve as a single
system, integrating all spatial information systems into one.
The new system is accessed by various government agencies, including: the
Ministry of Land, Transport, and Maritime Affairs; the Ministry of Public
Administration and Security; regional and municipal government organizations;
academic and research organizations; and the general public.
Information flows freely from the municipal and regional level down to
districts, cities and provinces with ease.
39. + Results
MLTM was empowered with a systems architecture that has functionality similar
to Google Earth.
MLTM has the ability to tap into above-ground and underground data 24/7.
The system integrated 114 unique spatial information databases into
one, leveraging non-redundant data of numerous government agencies across
South Korea.
With the utility and ease of standardized Spatial SQL, MLTM advanced
application development based on ALTIBASE HDB’s integrated system.
ALTIBASE HDB gave MLTM a comprehensive solution while slashing total cost of
operation (TCO).
40. The Company: NH Bank
The Problem: Accounting Processing System
41. + Problems
NH was spending inordinate amounts of excess capital supporting
employee overtime and extended office hours, which resulted in
inefficiencies.
The vast majority of NH’s branches were forced to prolong operating hours on a
regular basis. The culprit was performance issues with its financial accounting
information system.
The system could not effectively manage loading and processing large volumes
of data feeds from its 20 application servers. Representative data feeds included
tasks such as transaction processing, branch teller support, and error
management.
The underperformance wasted valuable resources in the form of needless
overtime pay while sparking an uncontrollable deterioration of company morale
and overall productivity.
42. + Solution
Realizing that the root cause of its performance failures stemmed from the
pervasive limitations of its traditional on-disk DBMS, NH deployed ALTIBASE
HDB’s In-Memory database to enhance its accounting processing system.
ALTIBASE HDB resolved NH’s large volume data management
deficiencies, shortening office hours, increasing worker productivity and
reducing spend.
The Hybrid architecture of HDB provided NH with the ability to process data in
real-time while meeting governmental data storage requirements on disk.
43. + Results
NH reduced the operating hours of its 1,172 branches by an average of over 1
hour per day.
NH’s accounting processing system increased the performance of loading and
processing data feeds from its 20 application servers by 500%.
NH’s accounting processing system seamlessly handles 3,000 TPS and over 50
million transactions per day.
NH provides customers with uninterrupted 24×7 service by leveraging ALTIBASE
HDB’s HA feature that comes out-of-the-box and is deployed with ease.
NH no longer wastes capital on office overhead and overtime pay and has
repositioned itself for high productivity.
45. + Problems
Samsung Securities had three critical areas that suffered from
inadequate DBMS performance and reliability:
■ Customer Retention: Samsung Securities had pinpointed significant decreasing
revenue as well as lost opportunity stemming directly from inadequate retention
of global institutional investors. The problems were uncovered by identifying
customer complaints indicating that they had lost transactions due to insufficient
speed in their futures/options trading.
■ Customer Acquisition: Concurrently, Samsung Securities realized that new client
acquisitions were facing hurdles that originated from the very same problems
with speed.
■ Growth: Samsung Securities was not able to keep up with the increasing trading
volumes in the futures/options market. Simply put, Samsung Securities was
being limited by its speed.
46. + Problems
Samsung Securities, prior to switching to ALTIBASE HDB, was utilizing a Sybase
conventional on-disk DBMS. Growing trading volumes resulted in a significant
increase in the number of database transactions.
This increase put a big burden on the abilities of their existing conventional on-
disk DBMS which quickly became a bottleneck. Even attempts to use caching to
improve performance did not solve the problems, as Samsung Securities could
only process 750 trading orders per minute
In addition, the burden on system resources grew rapidly resulting in up to 60%
CPU consumption to handle database transactions.
47. + Solution
After Samsung Securities deployed ALTIBASE HDB using in-memory only mode to
replace the Sybase DBMS, there was a dramatic improvement in system
performance.
ALTIBASE HDB in-memory DBMS enabled Samsung Securities to process 20,000
trading orders per minute with an average execution time of 3 milliseconds per
order. While delivering extreme speed, ALTIBASE HDB utilized less than 20% of
CPU, resulting in significantly low resource consumption.
ALTIBASE HDB, as a full-featured and standards-compliant DBMS, made it easy
for Samsung Securities to migrate existing database objects and data. All existing
Sybase tables and stored procedures were converted to ALTIBASE HDB by four
technical staff within two months using familiar programming languages and
standard SQL.
48. + High Availability (HA)
■ Samsung Securities implemented our native replication feature based on Active-
Standby HA architecture. In this architecture, an up-to-date backup of the
database is maintained on a second system. If the master server unexpectedly
becomes unavailable, service immediately resumes from an identical database
on an alternate server. This provides a nonstop operating environment with
improved reliability and fault-tolerance. This architecture ensures that Samsung
Securities’ mission-critical data remains uncompromised. Unplanned downtimes
(system crash), malfunctions or planned downtimes are issue-free due to
patches or upgrades to its DBMS.
49. The Company: LG Display
The Problem: Real-Time Flaw Detection
50. + Problems
LG Display suffered from inadequate quality control. Its inability to track defects
and change requirements in real-time was at the core. Data stored on multiple
DMBS’s perpetuated LG Display’s incapacity to monitor and react to these
issues, lowering yield and leading to high scrap. Limited to tracking only lots, the
automated system could not exploit the invaluable effects of pinpointing, per
unit, imperfections1.
Numerous workstation reliance on multiple databases, led to substantial performance
degradation, triggering uncontainable defects.
Vital quality control protocols went unanswered. Specifically, composition ratios of
active substrates and color filters were unidentified until it was too late.
Quality control was relegated to 12-hour old data delivered in 10 minute intervals.
LG Display found itself in a quandary. Increasing customer demand caused further
erosion in quality control, prompting unmanageable blows to profitability and
reputation.
1) The ability to detect even the slightest flaw in a single plate or deposited layer is vital to the quality control process.
51. + Solution
LG Display implemented ALTIBASE HDB in-memory DBMS in 2008. ALTIBASE
HDB captured defect data immediately and continuously. LG Display’s quality
control personnel could identify and remedy defects at the point and time of
occurrence.
LG Display deployed ALTIBASE HDB hybrid technology (in-memory and on-disk)
on a 20-core HP Superdome Server.
LG Display received real-time defect data on a micro-level per product. Data
was further broken down by station, time of process deviation, and composition
layer.
Prior to implementing ALTIBASE HDB, LG Display’s legacy system processed 21
million transactions per hour consuming 25%-45% CPU usage. With ALTIBASE
HDB, 12 million transactions are processed per half hour with only 10%-20%
CPU usage.
52. + Results
LG Display possesses an industry leading quality control system.
Precise product manufacturing processes are wed with exacting defect
detection.
Personnel isolate problems instantly and act on accurate data, reforming
communication.
LG Display relies on stringent data streams, quickly identifying root causes of
defects1.
LG Display implements practices to eliminate recurring breakage points.
LG Display’s revenues, profits, reputation, customer satisfaction and loyalty are
bolstered by a superior product.
53. + Technical Details
LG Display achieved outstanding results with their new Advanced LCD
Processing Control system by leveraging the key features of ALTIBASE
HDB, specifically in the areas of high-performance, high availability and
flexibility.
ALTIBASE HDB ALTIBASE HDBReplication
APC
Ahead Processing Control
Active-ActiveMonitoring &
Control
LG Display
ALTIBASE HDB
Active-Active Architecture
And
Application Interfaces
54. + Performance, Flexibility and High Availability
Before LG Display implemented ALTIBASE HDB, each manufacturing process was
managed by a dedicated conventional on-disk DBMS. LG Display desired to
manage all manufacturing processes by using a single DBMS for more efficient
administration of the monitoring systems and to have an integrated view of all
manufacturing processes. However the idea of a centralized DBMS system was
not possible due to poor performance of the conventional DBMS. LG Display’s
performance requirements of 5800 TPS (transactions per second) for INSERT
statements and 2100 records/hour for data migration were not met by the
conventional DBMS.
Besides the performance issues, the need to maintain a dedicated DBMS for
each manufacturing process created major challenges in the areas of application
development and maintenance, as well as database administration and tuning.
The developers had to deal with an extremely complex application
infrastructure. The database administrators had a difficult job of administering a
large number of DBMS’s, for software updates and data integration tasks. The
challenging administration tasks also resulted in numerous unplanned and
planned system outages.
55. + Performance, Flexibility and High Availability
ALTIBASE HDB allowed LG Display to consolidate all existing DBMS’s into a single
high performing DBMS to monitor all manufacturing processes in real-time.
Taking advantage of both the unique hybrid functionality and the built-in
replication feature, LG Display implemented ALTIBASE HDB based on a 2-
node, Active-Active architecture. In this architecture, both ALTIBASE HDB
instances were configured in hybrid DBMS mode; in-memory DBMS + on-disk
DBMS.
This architecture was the perfect fit for eliminating performance, administration
and reliability issues of LG Display’s older systems. ALTIBASE HDB in-memory
DBMS delivered extreme speed for real-time monitoring of all manufacturing
processes by delivering mind-blowing 50,000 TPS performance for INSERT
statements. ALTIBASE HDB on-disk DBMS was used as the repository for 12-hour
and older manufacturing process data. Migrating data from memory to disk was
a simple task attributable to ALTIBASE HDB’s unique MOVE SQL statement. With
MOVE statement, LG Display was able to migrate data from memory to disk at a
10,000 TPS performance.
Active-Active HA architecture was realized using the built-in replication
feature, allowing synchronous replication in a shared-nothing configuration with
a zero downtime topology. With the new architecture, day to day tasks for both
application developers and database administrators were greatly simplified.
The world of data has been changing very rapidly in the last few years. We agree that data volumes are explodingWe see parallelization, and big-data movement with Hadoop etc.But in the database world – the battle has always been between OLTP and OLAPOLTP are the operational systems and OLAP are the analytical systems.
Now let’s take a look at how real-time technology is having an impact on two very important component of our business : Within the IT frameworkAnalytical systems such as your data warehouse, data mart, reporting applications deliver the informationAnd operational systems such as CRM, SFA, POS, and ERP applications provide the reach; the interaction with customers, partners, and end users Analytics are carried out by on-line analytical processing – OLAP systemsOperational systems are on-line transaction processing – OLTP systemsTraditionally;OLAP systems; deliver to few users, answers to long complex questions, from large datasets to enable strategic decisions. These systems are typically updated in batch and they are important in the enterpriseOLTP systems; Support many users, processing a lot of transactions, with smaller operations and datasets, to ensure that the business can tactically operate. They are typically real-time systems and they are not merely import ant; they are mission critical for the enterprise.Today – the lines are getting blurred about OLTP and OLAP systems.
High performance is the core competency backed by a full-featured and mature RDBMS and long R&D history in in-memory computing domain.There are a lot in-memory databases out there including big names like Oracle Timesten and IBM SolidDB. But none offers the wealth of features Altibase provides. Because these companies see in-memory as a front-end to their on-disk DBMS thus they put features good enough to solve specific problems and rely on their conventional disk DBMS to do the rest. Altibase has everting in the in-memory database and it does not need yet another backend DBMS to be complete.
As already mentioned we have more than 500 customers using Altibase solutions. Here are only a few examples of different industry applications that make use of Altibase solutions.
Examples of utilization of in-memory computing abound in multiple vertical sectors and geographies. In some cases dramatic business innovation has been enabled by a wholehearted adoption of "in memory" technology as an architectural foundation for all, or at least most of, the application software running the business. Industries such as on line gaming or global software-as-a-service could not exist without in memory computing technologies. The scalability, performance and continuous availability required to be successful in those markets would not be achievable using traditional computing models and patterns. However these are probably just the tip of the iceberg of a much wider array of in-memory enabled applications that the industry will find out about over the next five years and beyond.
The entire database resides in memory.The techniques used in in-memory computing are far more advanced than the traditional on-disk databases which bottleneck at I/OExtreme performance achieving SELECT query performance over 1, 000,000 TPS (maxed out at 1,420,000 TPS during in-house tests).In ALTIBASE HDB, the focus is on short and predictable response times that naturally result from the fact that the data is already in memory. Predictable response times are a natural attribute of in-memory databases since there is no I/O activity (except for database recovery purposes). The improved response times also fuel high throughput.
One of the misconceptions about in-memory databases is their reliability. ALTIBASE HDB guarantees reliable transactional processing by implementing a database server that satisfies all ACID (atomicity, consistency, isolation, durability) requirements.Atomicity requires that database modifications must follow an “all or nothing” rule. Each transaction is atomic. If one part of the transaction fails, the entire transaction fails and the database state is left unchanged.Consistency ensures that any transaction that the database performs can take it from one consistent state to another.Isolation refers to the requirement that other operations cannot access data that has been modified during a transaction that has not yet completed. The question of isolation occurs in case of concurrent transactions (multiple transactions occurring at the same time).ALTIBASE HDB supports the isolation levels defined in the SQL-92 standard.MVCC is a concurrency control method which basically aims to avoid Writers blocking Readers and vice-versa. The problem of Writers blocking Readers can be avoided if Readers can obtain access to a previous version of the data that is locked by Writers for modification. HDB keeps only the latest version of data in the database, but reconstruct older versions of data dynamically as required by exploiting information within the Write Ahead Log.Durability is the ability of the DBMS to recover the committed transaction updates against any kind of system failure (hardware or software). Durability is the DBMS guarantee that after the user has been notified of a transaction's success, the transaction will not be lost.ALTIBASE HDB adheres to WAL (write-ahead logging) protocol. Before overwriting an object to non-volatile storage such as disk with uncommitted updates, the log records relative to such updates should beforced to the non-volatile log space (UNDO information is gathered).Before committing an update to an object, the log records relative to such an update should be forced to the log on non-volatile storage (REDO information is gathered).The durability level controls how ALTIBASE HDB handles transaction logging. The HDB server supports different levels of durability from most relaxed, to most strict. Each of these levels guarantees durability to a different extent and realizes different performance characteristics. Relaxed durability yields the best performance; where as strict durability eliminates loss of transactions.
ALTIBASE HDB combines several key services such as high-availability, fault tolerance and load balancing with its built-in replication feature (a standard component)ALTIBASE HDB’s log-based replication architecture, while providing very speedy performance, imposes very little overhead on computing resources since it only transforms transaction logs into logical logs and sends them to remote servers for processing.ALTIBASE HDB replication feature maintains an up-to-date backup of the database on an active server, and in the event of that server unexpectedly becomes unavailable, immediately resumes services again from an identical database on an alternate server, providing a non-stop operating environment.ALTIBASE HDB replication provides users with a choice of two commonly used modes of replication. Asynchronous mode (meaning master server does not wait until a remote server is done applying a transaction) and Synchronous(meaning a master server commits a transaction only after it has received conformation from the remote server). Async mode focuses on high performance while Sync mode focuses on data integrity and consistency.In replication environments, it is quite common to have data conflict issues. ALTIBASE HDBprovides a built-in audit feature that can discover and auto-resolve data conflict issues due toreplication.Other replication features include TCP/IP based 32-way replication, table level replication, network failure detection, support for heterogeneous platforms, offline replication, and so on.The computing environment can maintain near standalone performance during replicationoperations (90% Active-Active and 96% Active-Standby).
Horizontal scaling, or also known as scale-out, refers to multiple independent computer nodes working together to process workloads. Altibase allows horizontal scaling up to 32 nodes over TCP/IP networks.Vertical scaling, or also known as scale-up, refers to the ability to extend processing capability by additional DRAM and processor power within the same computer node. Altibase in-memory database scales very well vertically. In addition, it allows dynamic resizing of database without any service down time or interruption.
In house performance test numbers. The Select operation performance at a whopping 1,417,164 TPS. Please note that Altibase’s performance scales very well with the increment of client connections. Some of the competition like TimesTen and SolidDB perform poorly as the number of client connections increase, especially after 4 clients.
Altibase recognizes that one size does not fit all when it comes to application performance.There are different application architectures with unique requirements.Altibase, in addition to the standard client/server protocols such TCP/IP and IPC, also offers two additional interfaces for applications to achieve extreme processing performance.The methods called Direct Access Mode and Direct Access API mode are suitable for applications that can run on the same machine as Altibase in-memory database.Direct Access Mode allows use of standard interfaces such as ODBC and JDBC. Using this mode applications can achieve significantly better performance than conventional client/server applications due to the fact that all database operations are executed directly from application process’s space without any inter-process or network communication.Direct Access API mode takes this to a whole new level by eliminating the need to go through the query processing overhead. This mode provides highest performance possible.
There is a direct correlation between Data Size and Speed. As data gets bigger speed gets slower.Pure disk-resident databases allow nearly unlimited amounts of storage, but their performance is dominated by disk access. Pure in-memory databases are fast, but strictly limited by the size of memory. With In-memory databases access speed can be up to 20 fold especially with INSERT, UPDATE and DELETE (DML).The disk resident databases can improve performance via buffer pools however managing the buffer pool requires substantial memory and CPU cycles, and the solution under performs as compared to an in-memory database. A typical in-memory SELECT is still 5x faster even when using buffer pools.But, the in-memory database is not necessarily the best cure for all problems either. The benefit gained from the memory-centered processing is sensitive to work loads, usage scenarios and clearly not a good fit for large data volumes. Some databases are so large they will never fit into an IMDB.Almost since they were first developed, in-memory databases have been used in conjunction with disk resident databases to create a hybrid database infrastructure that could take advantage of the faster execution speeds in-memory databases. Most database vendors achieve this with an architectural concept known as a dual-engine DBMS by integrating their in-memory engine with disk-based engine. The IBM InfoSphere Change Data Capture (InfoSphere CDC) technology is responsible for replicating the databetween the back (DB2) end and front (SolidDB)end to ensure that each database is performing transactions on the same data. However, there are only a small number of database vendors that offer a pure hybrid database architecture in which both in-memory and on-disk objects are handled by a single engine. With a pure hybrid database, the tables with large volumes of data reside on disk, and the smaller tables that are frequently accessed reside in memory while a single engine is responsible for processing all data objects. Hybrid architecture also enables customers for lower TCO since once server replaces two servers, one software license for both memory and disk, and of course greatly simplified application development and database adaministration.ALTIBASE HDB is also very flexible when dealing with temperature of data.Data in a database can be classified according to its temperature. The temperature of data is based on how often it is accessed, its volume, how volatile it is, and how important the performance of the queries that access the data is.Hot data is frequently accessed and updated, and users expect optimal performance when accessing this data. Cold data is rarely accessed and updated, and the performance of the queries that access this data is not essential.Identifying and characterizing data into temperature tiers can allow optimization of the performance of the queries that matter most while helping to reduce overall cost.
ALTIBASE HDB is a standards-compliant relational database supporting SQL and the standard ODBC , JDBC, OLE DB and .Net programming interfaces. Most application design, programming, data model design and system administration paradigms used with other database systems are directly applicable with ALTIBASE HDB.SQL - Many features of SQL-92 and some of SQL-99 and SQL-2003 standards are also supported.ODBC - The ALTIBASE HDB ODBC Driver conforms to the Microsoft ODBC 3.5.1 API standard. Itis distributed in the form of a library. The HDB ODBC Driver supported functions are accessed with HDB ODBC API, a Call Level Interface (CLI) for HDB databases, which is compliant with ANSI SQL CLI.JDBC - ALTIBASE HDB JDBC driver is a type 4 driver (100% pure Java implementation) that conforms to JDBC 3.0 standard. We are currently working on a new version that will conform to 4.0 specification.
Another misconception about in-memory and hybrid databases is their completeness. ALTIBASE HDB in terms of tools and utilities is as complete as any DBMS.iSQL is a command-line tool for ALTIBASE HDB. It allows users to connect to HDB, and issue SQL statements. It also allows authorized users to perform database control tasks such DB startup, shutdown, backup and recovery.iLoader is a command-line utility to extract or load data from/to an ALTIBASE HDB Server. Typically used for database migration or backup operations. AExport is a command-line utility to support automated data migration between Altibase HDB Servers (including support for different HDB versions).Audit is command-line utility to manage the data synchronization between the servers in the replication environment. Compares local and remote servers on a table-by-table basis and provides information about data inconsistencies.It also bi-directionally resolves the inconsistencies according to the synchronization policy set in the environment file.AdminCenter is an intuitive, graphical tool to help both developers and database administrators to manage their Altibase HDB Servers. AdminCenter for DBA focuses on database monitoring functionality. AdminCenter for Developers provides rich set of productivity features for the database application developers. AdminCenter is a Java/Eclipse based GUI application, and it is platform independent.AltiProfile is an online query analyzer to evaluate and optimize DML statements. Provides detailed statistics and execution plan information in the form of text files for ease-of-use.And of course also, 3rd party integration with popular database tools and utilities via standard interfaces.
Altibase is an open platform runs on all common platforms. Mainframe is not supported. We are working on a Mac port.
- The amount of data in the world is constantly increasing, and with it the demand for real-time data processing.- As the list of entities that produce and consume data grows, the quantity of data is explosively increasing. There is a lot of information. And it is on the move.- And most of this information is machine created unlike old days when humans created the data.- Data volumes are higher though decision cycles are shrinking.- This creates a new challenge. You have the data that has the value while in-flight but has very little or no value after fact.The rise of ubiquitous technology is expected to trigger an increasingly distributed environment, changes in the attributes of data, and enormous increases in the amount, complexity and heterogeneity of data. - To effectively handle such data, powerful computing skills are not the only requirement: completely new data-handling methods suitable for this computing paradigm are required.Continuous QueryA query registered, then continuously evaluated over the data. CQL is a subset of standard SQL and provides familiar interface to application developers.Window allows temporarily saving a set of rows to maintain a history of data stream.Count-based: “Select * from CPUdata keep 100 rows”Time-based: “Select * from CPUdata keep 4 seconds”Publish/Subscribe is a messaging pattern where senders (publishers) of messages do not program the messages to be sent directly to specific receivers (subscribers). Rather, published messages are characterized into classes, without knowledge of what, if any, subscribers there may be. Subscribers express interest in one or more classes, and only receive messages that are of interest, without knowledge of what, if any, publishers there are.This decoupling of publishers and subscribers can allow for greater scalability and a more dynamic network topology.
1. Average of 20 days from creation and delivery of customer bills.
1. Average of 20 days from creation and delivery of customer bills.
1. Average of 20 days from creation and delivery of customer bills.
1. Average of 20 days from creation and delivery of customer bills.
1. Average of 20 days from creation and delivery of customer bills.
1. The new system is 5.3 time faster than the legacy system with the added benefits of High Availability (HA) and replication.2. Per-second charging scheme saves SK Telecom customers an average of 14.5 million USD per annum.3. SK Telecom acquired Hanaro Telecom in 2007 to offer internet services with SK Broadcom division.