|Technology Focus: The Hidden Costs of Downtime Make HA a Serious Business Necessity|
|System Administration - High Availability / Disaster Recovery|
|Written by John Ghrist|
|Saturday, 18 April 2009 19:00|
The need for HA is no laughing matter in the current economic climate. Check out your HA software options for the System i.
Keeping your System i servers processing 100 percent of the time was, for many years, not considered a serious requirement for any but the largest of companies. The intrinsic reliability of the System i and its predecessors was good enough for most SMBs, which relied on the platform to handle processing during normal business hours. Add to that the cumbersome implementation of early high availability (HA) systems for the platform in the late '90s and into the current decade, and you had a market situation such that, for the majority of System i shops, HA was either a luxury most enterprises had no room for in their IT budgets or the potential makings of a standup comedy routine at a company party.
System Uptime Situation Gets More Ticklish
Some of these attitudes have persisted until fairly recently, restricting the demand for HA solutions. This situation likely contributed to the major consolidation among System i HA providers in 2007, when Vision Solutions bought such longtime HA vendors as iTera and Lakeview Technology, and IBM purchased DataMirror Corporation.
But the increase in such phenomena as trading with international companies whose workers are awake when ours are normally asleep, the proliferation of commerce over the Internet on a 24x7 basis even within North America, and the growing awareness of unplanned downtime as more than just an inconvenience is changing such attitudes. Particularly in an economic climate where corporate income counts more than ever, the hidden costs of downtime make HA a serious business necessity that makes more economic sense for a growing slice of the System i universe.
Paying to Play
HA isn't free, unfortunately, and enterprises must evaluate for themselves when HA solution prices justify the investment. That can be done only by a sober assessment of the real costs of downtime, many of which are hidden. An obvious cost is the loss of actual business while systems are down, but cloaked in that concept are intangibles that are hard to calculate. These include, for example, the opportunity loss of potential customers who turned to a competitor and formed a loyalty bond with them because your Internet site was unreachable, and future losses because of the negative impression potential customers formed when they couldn't transact the business they wanted to settle quickly. Other costs can include loss of productivity on the part of employees whose salary and benefits must still be paid even if they can't do much of their normal work while systems are down. Other factors may include penalties for failure to complete contracted work or deliveries, the costs of associated discounts or "make-good" actions, and revenue postponements due to billing delays. Nothing funny about the money here.
While it's possible to create a homegrown HA system using multiple servers and such services within the OS as journaling, the costs of building and maintaining such a system, assessing and maintaining its reliability, and using internal IT resources in the event of problems must all be taken into account for an accurate cost assessment. For many SMBs, going with a packaged solution backed by the vendor's resources is the best alternative.
Vision Solutions, one of the HA vendors whose products are profiled below, has summarized many of the justifications for SMBs to invest in HA in a white paper, "Five Reasons Why Small Organizations Should Consider High Availability," which is available free at the MC Press White Paper Center. Although it's understandably pro-HA adoption, it summarizes many of the considerations you should review before making a decision about purchasing any HA product.
Have You Heard the One About Clustering?
Clustering is the practice of using multiple servers, called nodes, that are linked together via LAN, WAN, or the Internet to provide internal services such as processor load balancing, parallel processing for complex applications and problems, and grid computing. Because a cluster's nodes don't have to be in the same geographical location, clustering is also another means of providing not only an HA environment, but a disaster recovery (DR) solution as well. Replicating objects and transactions to a nonlocal server provides coverage if some calamity befalls your primary data center.
Although a detailed discussion is beyond the scope of this article, if you're considering HA, you might want to also consider clustering. If your enterprise could make use of some of clustering's other functions, it might be easier to justify the expense of HA.
Getting to the Punchline
What follows is a quick overview of the major players and software products available in the System i HA market space. This article doesn't cover HA services, mainly because going with that kind of solution ties you to whatever methodology and products the service company you contract with prefers, making an overview of other solutions moot. This article will also focus only on solutions for the System i OS because there are still not that many System i machines running other OS options.
Each product includes the vendor name, the product name, a link to more information about each product, and a brief description. These descriptions are in no way complete information about the products; they are just summaries of major features to help you decide where to focus your own research efforts first.
Bugbusters Software Engineering, Inc.
Bugbusters Remote Software Facility (RSF) is a veteran System i communication product that the company continues to reinvent. The latest version makes a quantum leap into HA functionality, providing replication services to a hot backup machine. Designed to be inexpensive and simple, RSF-HA is Web-downloadable and installable without training; provides an i5/OS-based Work With Synchronization Attributes panel for monitoring and controlling HA activity; and can replicate user profiles, authority changes, system values, and network attributes.
A fruit of IBM's acquisition of DataMirror, IBM iCluster uses i5/OS journaling to replicate data and objects to other machines in a cluster. It includes modules for HA and load balancing, as well as a specially designed one for SMBs. The product can scale to handle up to two billion transactions an hour, includes a GUI-based administrator that runs on a PC to control the system, and facilitates regular swapping of primary and backup processors.
IBM Power HA for i
Formerly known as the High Availability Solutions Manager, Power HA for i not only provides HA services, but also works with IBM's System Director Navigator for i and DS6000/DS8000 hardware replication products to offer an asynchronous DR solution. It supports mirroring between sites, web-based cluster management for HA and DR operations, encryption capabilities and other security features for online data and backup media, and auditing to support compliance requirements.
Power Systems Capacity BackUps (CBU)
IBM's CBU offerings facilitate HA/DR operations between any two System i machines and interact with some of the other HA vendor offerings mentioned elsewhere in this article, one of which IBM recommends customers additionally purchase. Simpler than iCluster or Power HA for i, CBU is a built-in feature for selected System i machine models and Blade Servers that is specifically designed to support in-house DR operations. Special features include transferability of System i, 5250, and user entitlements between supported servers.
Maximum Availability offers three levels of HA products, of which SENTRY is the most basic, but it's easy to upgrade to one of the others. Relying on i5/OS journaling, it requires installation only on the backup machine, unlike most HA solutions. It uses menu navigation and also transfers data queues, user profiles, and spooled files.
DEFENDER includes all of SENTRY's features and adds self-diagnosing autonomics; advanced auditing; replication of auxiliary storage pools and logical partitions; role swapping and failover capabilities; and transfer of data areas, dynamic database changes and nearly 40 other object types.
GARRISON is the top-of-the-line offering, which offers DEFENDER's features and adds transfer capabilities for IFS and MQSeries objects.
France-based Trader's offers Quick-EDD/HA, which provides an HA environment that emphasizes real-time replication between production and backup systems using i5/OS journaling. However, it also includes an application synchronization tool that lets apps with security requirements that are incompatible with journaling be replicated between systems even if "out-of-synchronization" conditions occur. Quick EDD/HA replicates data, application and system objects, and IFS and spooled files.
Vision Solutions, Inc.
iTERA HA provides basic HA capabilities at a reasonable price, mirroring objects and transactions to a backup machine and enabling smooth switchover of users and processing to the backup machine as necessary. It copies programs, data areas, data queues, user profiles, device configurations, triggers, and constraints, as well as IFS and WebSphere files and other objects.
MIMIX HA comes in two versions: MIMIX ha lite for SMBs and MIMIX ha dr1 for larger enterprises with more complex environments. In addition to HA services such as those provided by iTERA HA, MIMIX emphasizes administrative tools that simplify product installation and monitoring. Individual modules automate operations such as product configuration and deployment, integrity issue reporting, determination of single- or multiple-threading for application processes, data recovery from any point in time, and between-machines switching.
Orion Solutions HA
Orion Solutions HA emphasizes HA for high-volume and complex HA environments, being capable of handling millions of transactions per minute. It offers two versions: Orion Professional and Orion Enterprise. Professional uses either remote or local journaling to provide an HA environment for medium-sized businesses. Enterprise adds support for advanced technologies such as MQSeries, three-tiered applications, clustering, and flexible replication technologies (e.g., one-to-many, many-to-one, many-to-many).
Vision Cluster1 is a cluster-management environment that operates with or without replication and can manage multiple clusters. It's designed to support storage area networks and cross-site mirroring. Features include autonomic monitoring with color-coded alerts, single-screen status displays for HA system elements, automated switching between systems, and optimized system configuration and setup.
|Last Updated on Sunday, 19 April 2009 11:27|