|
|
AS OF 7/4/2008 8:33PM EST
|
Guest View: Data Access Middleware Is Important
By Ken North
December 1, 2007 —
Performance and scalability are enterprise-computing requirements for transaction processing, business intelligence, ERP suites, content management and other applications. High-volume Web sites, e-business and Internet applications also require a robust security model. That’s why performance, scalability and security are defining characteristics of top middleware for distributed SQL processing.
Since the favored paradigm for enterprise computing shifted from monolithic to distributed applications, the software industry has partitioned logic into reusable libraries and components. Building complex systems today, like building a Porsche or an F-18 fighter, involves repeating patterns and assembling parts, with success being dependent on quality of parts.
Service-oriented architecture is a variant of the theme, assembling collaborative applications from distributed parts (Web services). Like manufacturing experts, IT managers must be mindful that faulty parts can be the proverbial weak link in the chain.
Web 2.0 applications, ERP suites, data warehouses and other distributed applications typically use middleware components. Middleware plays an integral role in the distributed computing model. It is software, such as network libraries, that operates as a middleman between clients and servers.
For distributed SQL processing, data access middleware provides the data channel from database clients to local and remote database servers. Data access middleware morphed as the database scene evolved to encompass desktop, workgroup, enterprise and Internet applications. Like a faulty part in a Porsche or an F-18, substandard middleware can cause performance problems and system failures.
Distributed and collaborative applications open a door to high-availability computing, scalability, component software and refactoring. Scalability comes from dividing a problem into parts to distribute logic and data across multiple computers. The partitioning of logic and data with a federation of servers is an approach that supports load balancing and failover for high-availability computing. It can scale to handle large numbers of users for enterprise or Web applications. But the distribution of processing and data introduces security issues, data integrity issues, network latency and other performance issues.
Middleware can play a mitigating role if it does more than open a data spigot for SQL queries. Premium-quality data access middleware includes features that help developers deal with problems such as high network latency and frequent round trips to databases.
Network managers and database administrators have a variety of tools for sniffing packets, measuring network latency, detecting intrusions and monitoring query execution. They can, for example, use a diagnostic tool such as the Oracle netstat utility.
Being able to track system events and network traffic statistics is useful for pinpointing problems, such as determining how long Oracle has waited for network packets. Corrective actions can include changing the TCP send/receive buffer size or rewriting SQL to optimize a query, but middleware can play a role. Because SQL protocols such as TNS and DRDA operate at a level above TCP/IP, premium-quality middleware includes options for optimizing database communications on a per connection basis.
Feature Checklist
Best-of-breed middleware should be an essential ingredient in our recipes for robust enterprise and Web applications, but for more reasons than optimizing communication and minimizing round trips to the database. IT managers and enterprise architects will benefit from using a checklist of the data access middleware features that are needed to sustain applications, services and Web sites. For premium-grade data access middleware, certain attributes appear routinely on such a list. They are:
• Multiplatform flexibility. Organizations often have a mix of operating systems (Linux, Windows) and databases (IBM DB2, Microsoft, Oracle). The best-grade middleware is available for different chip sets and multiple operating systems, with support for heterogeneous databases having diverse types and features. It’s also suitable for embedding with diverse applications platforms, ERP suites, integration servers and data warehousing platforms.
• Performance. Caching, connection pooling, multithreaded operation, dynamic adjustment of network packet size, direct support for wire protocols, BIND packages and using stored procedures for prepared queries are features that ensure premium-grade data access middleware will deliver peak performance.
• Scalability. Middleware must align with scalability solutions such as federated servers, load balancing and failover. The best middleware provides alternate server and platform-neutral client load-balancing capabilities.
• Broad connectivity and interoperability. Multidatabase access and interoperable SQL originated when organizations had a dozen types of SQL databases. Today that diversity includes XML document collections, native XML databases and SQL/XML databases. Data access APIs such as ODBC, JDBC and ADO.NET match up with different programming or scripting languages, but many aspects of database client programming are universal. There’s an advantage to using middleware from a supplier whose products cover the spectrum of APIs.
• Security. This is critical, especially since the Internet computing model exposed databases to users outside the firewall. For distributed applications, middleware plays a key security role. It’s part of a multilevel, comprehensive strategy for network security and database security that emphasizes secure communications, secure code and secure databases. Best-of-breed middleware complements three-factor authentication, Kerberos and other solutions for authentication and authorization.
• Reliability and integrity are essential, in particular for middleware that’s part of a mission-critical application. Reliability requires that middleware includes trace tools, spy tools and comprehensive prerelease testing. Integrity includes both data and transactional integrity when queries produce correct answers and updates. When a middleware product is embedded by hundreds of software vendors, such as middleware from DataDirect Technologies, it’s a strong indication of reliability and integrity.
• Compliance with standards is essential because it simplifies software development, promotes interoperability and helps avoid being at the mercy of a specific vendor. Relying on proprietary technology exposes us to price increases, discontinued products, supplier insolvency and other consequences of vendor lock-in. USB and SCSI are but two examples of how standards promote a choice of products instead of being dependent on a single supplier. In the database world, SQL, ODBC and JDBC offer the same benefit of choice, but to the application developer they also represent standard, consistent data access behavior.
Diversity abounds in database development, particularly in mixed platform environments with proprietary and open source tools. A mix of SQL databases, network libraries, scripting and programming languages calls for diverse skill sets. Organizations need specialists for database administration, application development and Web site authoring, partly because each specialty has its own set of tools. Deploying DBMS-specific or application-specific client libraries or middleware exacerbates the support requirement. Proliferating database drivers, data providers and client libraries increases administration, training and application development costs.
Total Cost of Ownership
Discussing the total cost of ownership of “free” drivers and data providers may seem like a paradox, but not when considering the cost of poorly performing applications. A financial services organization, for example, had problems with performance of a key application. It tried tuning the application, queries and database design (at a cost of several hundred thousand dollars). Performance wasn’t acceptable until after switching to a premium-quality database driver. Other costs include opportunity costs, such as losing sales when applications are slow or prone to failure.
To fulfill quality-of-service expectations, we must create systems that are stable, secure and capable of meeting performance and scalability requirements. To maintain a standard of excellence, we must emphasize quality when buying or building the constituent parts. Every part of a system must be robust, exhibit consistent behavior, and support data and transactional integrity. This is a requirement of servers, databases, networks and middleware.
IT organizations that emphasize quality when selecting hardware and networks should not fall into the trap of using substandard middleware components.
Ken North is a consultant, author and industry analyst. Visit The Ken North Computing web site at www.KNComputing.com.
|