Wednesday, February 22, 2012

5 Facts that SAP won't tell you about HANA


SAP has been touting HANA as an innovative, breakthrough technology and the next “big thing”. They are aspiring to be the #2 database vendor riding on the HANA hype. Well, time to dig deeper and bring forward 5 facts that SAP won’t tell you about HANA.

#1: HANA is an In-Memory database. So, where’s the innovation?
SAP has positioned HANA as the newest, most innovative category defining offering. However, HANA is an in-memory database which may be a new category for SAP but has existed in the market for years. A quick Wikipedia search on in-memory databases reveals that in-memory databases have been around since the 1990s and today there are 40+ such independent offerings of which HANA is one. Oracle alone has 3 in-memory database offerings with successful products like TimesTen, Berkeley DB and MySQL.  Introduced in 1990’s, Oracle’s TimesTen remains an early innovator and a leader in this space. HANA, introduced in 2011 is the youngest member of the group.

#2: HANA adoption is growing rapidly. So, where’s the growth?
SAP will show numbers like FY 2011 revenues of $200M and 100+ customers to underscore HANA’s rapid customer adoption. Putting these numbers in perspective; Vertica, the largest independent in-memory database vendor before being acquired by HP in 2011, was on track to deliver revenues around $100M with 200+ customers and over 100% YoY growth rate. Oracle remains the leader in data warehouse platform market with FY 2010 revenues of close to $3B and thousands of customers. Given Oracle & Vertica’s impressive performance, HANA’s numbers while good are hardly “rapid”.

#3: HANA is enterprise ready. So, where’s the manageability and reliability?
It takes years to develop and perfect a complex product like database management system. Oracle database has been perfected over 30+ years and billions of dollars in R&D investment. TimesTen has been around for 15 years and is still being aggressively developed and perfected. SAP would like you to believe that HANA is enterprise ready from day one but dig deeper and you’ll find that HANA lacks basic features like clustering, high availability, file system persistence and ACID style transaction integrity support. HANA lacks referential integrity support so there is NO means to ensure the integrity of data stored in a HANA database. HANA’s support for locks and transaction isolation is primitive so multi user concurrency is an issue. Hopefully, you get the picture that HANA is an immature version 1 DBMS which is far from being ready to support mission critical enterprise applications.

#4: HANA is non disruptive. So, where’s plug and play?
HANA has limited support for standard ANSI SQL.  In fact, HANA requires applications to be custom written for it using non-standard SQL. In my view this is a major show stopper. In this day and age where every vendor is working diligently to support openness and application integration via support for services oriented architecture and web services in comes HANA with SAP’s age old vision of closed system with no access to underlying data structures. HANA takes vendor lock in to new levels by limiting your choice of applications, reporting & analysis tools to a few offered by SAP.

#5: HANA is an appliance. So, where’s ease and speed of deployment?
Wikipedia defines computer appliances as consisting hardware and software pre-integrated and pre-configured before delivery to customer, to provide a "turn-key" solution to a particular problem. Benefits of appliances include ease and speed of deployment with lower risk and faster time to value. With HANA you buy hardware, software, networking switches and storage from different vendors. There isn’t a single point of support and with different vendors having markedly different development and upgrade cycles it’s excruciatingly hard to test, configure, certify and update the joint solution.

In conclusion, SAP’s larger than life solution HANA definitely underscores the strategic importance of data management and analysis to organizations but due to limitations highlighted above HANA is far from being ready to support mission critical enterprise applications. Customers should consider mature technologies like TimesTen based Oracle Exalytics and Oracle Exadata for their in-memory analytics needs. 

Tuesday, February 7, 2012

Big Data Analytics – The Journey from Transactions to Interactions


Big Data Defined

Enterprise systems have long been designed around capturing, managing and analyzing business transactions e.g. marketing, sales, support activities etc. However, lately with the evolution of automation and Web 2.0 technologies like blogs, status updates, tweets etc. there has been an explosive growth in the arena of machine and consumer generated data. Defined as “Big Data”, this data is characterized by attributes like volume, variety, velocity and complexity and essentially represents machine and consumer interactions. 

Case for Big Data Analysis

Machine and consumer interaction data is forward looking in nature. This data available from sensors, web logs, chats, status updates, tweets etc. is a leading indicator of system and consumer behavior. Therefore this data is the best indicator of consumer’s decision process, intent, sentiments and system performance. Transactions on the other hand are lagging indicators of system or consumer behavior. By definition leading indicators are more speculative and less reliable compared to lagging indicators; however, to predict the future with any confidence a combination of both leading and lagging indicators is required. That’s where the value of big data analysis comes in, by combining system and consumer interactions and transactions, organizations can better predict the consumer decision process, intent sentiments and future system performance leading to revenue growth, lower costs, better profitability and better designed systems.

So, which business areas will benefit via big data analysis? Think of areas where decision-making under uncertainty is required. Areas like new product introduction, risk assessment, fraud detection, advertising and promotional campaigns, demand forecasting, inventory management and capital investments will particularly benefit by having a better read on the future.

Figure 1: Combination of big data and transactional data delivers better insights and business results


Big Data Analytics Lifecycle
The big data analytics lifecycle includes steps like acquire, organize and analyze. Big data or consumer interaction data is characterized by attributes like volume, velocity and variety and common sources of such data include web logs, status updates and tweets etc. The analytics process starts with data acquisition. The structure and content of big data can’t be known upfront and is subject to change in-flight so the data acquisition systems have to be designed for flexibility and variability; no predefined data structures, dynamic structures are a norm. The organization step entails moving the data in well defined structures so relationships can be established and the data across sources can be combined to get a complete picture. Finally the analysis step completes the lifecycle by providing rich business insights for revenue growth, lower costs and better profitability. Flexibility being the norm, the analysis systems should be discovery-oriented and explorative as opposed to prescriptive.

Getting Started
Oracle offers the broadest and most integrated portfolio of products to help you acquire and organize these diverse data sources and analyzes them alongside your existing data to find new insights and capitalize on hidden relationships. Learn how Oracle helps you acquire, organize, and analyze your big data by clicking here.


Figure 2: Oracle’s engineered system solution for big data analytics