NewsBytes – May 2004

Best of BrainShare – calling all those technical folks who didn’t get to the US in March and can’t fly out to sunny Barcelona in September. Places are going fast for the Best of BrainShare 13-15 July, Novell, Bracknell. To book, or for more detailed information on this 3 day event with over 30 sessions to chose from, visit www.mindworksuk.com . Remember there only 100 places in total, so don’t leave it too late to book your place. For any queries, contact Julia Davey at Julia@mindworksuk.com

GroupWise Update by Tim Heywood

There are two items of interest in the GroupWise world right now, both involving Linux. The first is the new EYE campaign, sponsored by Novell and IBM, EYE (Exchange your Exchange) is based on replacing Exchange 5.5 (and 2000) with GroupWise 6.5 on the Linux platform. While technically this might not be anything novel, the fact that IBM, the owners of Lotus Notes, are marketing such a solution speaks volumes about the potential of GroupWise on Linux.

The second item of interest is the cross platform (Java) client. The latest release is able to work in both online and caching mode, communicating directly with port 1677 at a speed not possible with previous versions. Whilst the UI is still Java the method by which Novell have produced such stunning speed is beautiful in its simplicity. Drawing on the past experience of cross platform development and the need for a native Linux client, Novell have ported the core client DLLs to each of the native platforms (MAC and Linux) providing the features and functionality required, with the advantages of the Java front end; simple but brilliant.

On Site Training ….. Have you considered it? By Julia Davey

Peter’s been out and about this month visiting customer sites and the message we’re getting loud and clear is that, whilst there is considerable value in the certified courses, time and money dictate otherwise.

Mindworks’ primary focus is to deliver flexible, high quality bespoke training at sensible prices. Recent requests and training delivered include various combinations and technologies such as: Integration of eDirectory, Active Directory, Oracle, MS SQL; ZENworks training focussing on managing application objects; Active Directory for staff with eDirectory only experience; GroupWise Train the Trainer; iChain & Portal services; Implementing & Supporting Win XP removing AD and replacing it with eDir, to name just a few. If you’re considering training this summer, contact us soon to discuss your requirements.

Our specialised workshops also continue to be popular – whether it’s Linux, NetWare, eDirectory, ZENworks or GroupWise – these courses deal with the real issues you face day in, day out and they evolve from listening to the needs of our customers. Secure Identity Management will be on our Autumn schedule and if you have input on other courses you’d like to see on our Autumn schedule, we’d love to hear from you. Please contact Peter on peter@mindworksuk.com

Currently we are spending as much time as we can visiting customers in order to understand exactly how people are currently using the technologies and their associated issues. Through this we hope to ensure that we meet all of your needs. Remember we deliver what you want, not what we think you should have.

Now we are for it... by Joe Doupnik

Well, it has happened. Linux has finally arrived in people's minds everywhere. It's on our plates too, like it or not. Actually the matter is not about liking but rather of coping.

Novell is offering a taste of its things to come with its present NNLS product package. OpenOffice and SUN's StarOffice are the desktop main course. Apache is salt of the earth, everywhere, but then it was before this started. Trade rags are full of miscellaneous recipes.

Personally I find the drive for the desktop to be curious at this time, given that matters are not really up to it. But the season is upon us. Most of us are concerned first with the server farm, desktops can follow in due course. Yet there is strong coupling being offered in the form of automated desktop packages, ZEN-like but under new names. Thus we managers are being dragged into considering desktop changes too. Oh joy.

Recently I tried to stand back from the fray and think about the environment being offered by the major Linux vendors. Certainly the "distributions" are improving in quality and completeness, much better than those of say two years ago. Yet, that same activity is creating a trap for the inexperienced, and my sampling regarding Linux is today most managers fall into the inexperienced category.

The trap is vendors have increasingly tailored, modified, "improved" applications and many of the core operating systems programme so that they work better, that they offer a more comfortable computing experience, but only in their distribution and done their way. The intentions are honourable, well meant, and involve a heck of a lot of creative work by staff to bring a little order and smoothness to the disparate collection of works composing a typical Linux installation.

I can hear your comments. For gosh sakes, how can that be a trap? Those guys are slaving away to make my life easier. Trap? Rubbish, you say.

Well, consider. You have need of a feature in an application which is not offered in the canned version from your Linux distributor. What are you going to do about it?

Should you be especially brave, or have some experience, or both, you would go to the authors and obtain the source code distribution for building locally. That is what "Real Managers" have done for decades.

But we discover the app interacts with other applications too, the dreaded dependency problem. In the end we can figure out how to build our local version and keep it separate from the vendor's material. But it may not be nearly as easy as it looks, particularly when source code changes may be (often are) needed to adapt the code to this particular Linux. Recall, the word Unix is an oxymoron;
no two issues are quite alike. This takes us into code and system details which may be quite foreign to all but gurus. Yet, folks have done this for decades and lived to tell the tale. Um, I guess that is one of these self-fulfilling prophecies, considering who tells those tales.

As more applications are added to servers and our clients demand more services from them, we face the daunting challenge of building apps ourselves from original sources, and maintaining them. It turns out that we learn most tricks fairly quickly, plus we can raid & plunder the Linux vendor's source RPM material for hints and its patches. We become clever, devious, and experienced.

We need to become this experienced, so that we can offer the services we must. Vendors are targeting the desktop where gloss and glitter are everything and folks are not expected to tinker with internals. That's where the market is. They are not designing servers, not yet anyway. Our newly arising freedom of choice is turning out to be literal: lots of choices, lots of freedom to go our own way at our own expense and become thoroughly lost. Or use a desktop box with vendor prebuilts and pretend it is a server.

The lesson we draw from these experiences is we need to have in-house knowledge of how the operating system works and how various applications can be built. Hackery won't carry the day. The public airwaves won't solve many of our problems, the signal to noise ratio is poor on popular channels. We just need
to figure out most matters locally and get on with building more sophisticated systems than we have had before, and luckily we _can_ do this. We combine our tailorings with offerings from the major vendors, build what we need, and add to it next year. The knowledge base grows with it, we have a new set of skills.

Thus a first step is to sample the food on our plate. It tastes peculiar at first, we make a face and get used to it. Soon we may ask for second helpings and a doggy bag. But first we must figure out how to eat the stuff without gagging or making
a mess. Experience is needed (don't use your fingers, you clod) and acquiring it becomes our next goal.

Note
Joe Doupnik will be in the UK from June this year on a year’s sabbatical delivering on site training for MindWorks and developing a number of academic projects. If you’d like to tap in to Joe’s expertise please don’t hesitate to contact us. peter@mindworksuk.com

New Technical Workshops – 1 day July 16th , Novell, Bracknell
Visit www.mindworksuk.com to book and for more detailed information.

Anyone booked on Best of BrainShare will be entitled to £25 discount off the £150+vat cost.

Virtual Office Demo

Do you want to experience Virtual Office which is now available in NetWare 6.5 and Nterprise Linux Services? Check out the Virtual Office flash demonstration created by BrainStorm, Inc.

http://www.novell.com/coolsolutions/nnlsmag/features/trenches/tr_brainstorm_vo_demo_nls.html

HKCU User Shell

David Brightman’s HKCU User Shell can be downloaded free from: http://www.novell.com/coolsolutions/tools/1878.html

Linux and Web Services by Rob Hebron

This article contains a brief overview of the justifications, aims and methods of delivering resilience, performance and flexibility in the serving of web content and services web services. It will also introduce the use of Linux and open source projects to achieve these aims.

Terminology

In the article, when web services are referred to, the serving of static and dynamic content over the HTTP and HTTPS protocols is, more specifically, what is being referenced.

The justification

The first question that should be asked when considering beefing-up your web services architecture is a simple one – is it worth it? To reach your end result will require a great deal of time and effort. It would simply not be worth doing this in a production environment for a web site whose usage or content simply does not warrant it. Is it an internal site that is in use, but not critical? If so, then look at backup and recovery rather than the web services architecture.

If content delivery is of such key importance that measures need to be taken to ensure that the content is always available then one (or both) of two aims become relevant: reliability and scalability.

The aims

Reliability
A reliable web service is one that’s offers non-stop service, with full functionality available at all times. Typically reliability is compromised by one of three factors:

• software failure (including the OS, custom code, third party applications, security holes & patching)
• hardware failure (specifically the platform on which the web service runs)
• infrastructure failure (including LAN, WAN and power)

When addressing reliability, I normally approach it in using the following steps:

1. Consider whether the software is reliable first. Reliability can only really be tracked through testing. If problems arise can they be traced and resolved properly? If any part of the software were unreliable I would consider it to be a waste of time to address other areas affecting reliability.
2. Conduct a holistic risk analysis of hardware and infrastructure to determine the problems most likely to occur and the impact of them. The results will be determined by factors such as your own experience with the whole environment, contract terms with third parties and the record of delivery against those terms. What are the expected fix times – are these acceptable?
3. Of the risks identified with the highest likelihood of occurring, which can be addressed most effectively?
4. How will the steps taken to address them affect the functionality and manageability of the web service?

The output of these steps will hopefully be a list of steps that can be taken that take into account both software and hardware, and will have a positive impact on reliability. It is extremely rare, however, that all risks can be addressed.

Clustering is typically considered at an early stage to improve expected reliability. Specifically, clustering with fail-over functionality is used. The cluster is made up of two or more computers (nodes) that monitor each other - if there is a failure of one node, then another takes over the provision of the service. Commonly, service provision requires files stored on disk and for this reason clusters often include shared storage, such as a SAN.

Fail-over clustering alone isn’t a panacea – it only addresses one particular source of potential problems, which may not even be the most likely to occur. For example a service may become unavailable on a server through an intermittent software bug. The server stays up and the web service appears to continue running, so the intended failover to a backup server never takes place. Alternatively, a service may become unavailable because of a WAN link failure, or because the ISP has made an error while reorganising your DNS zone records. Fail-over clustering won’t fix reliability if these problems occur.

Achieving reliability for a web service requires investigation of all factors that may impact upon it, together with a consideration of the risk of failure and what can be done to address the failures should they occur.

Scalability
A scaleable web service is designed and implemented in such a way that as usage outgrows the capacity of a single server, additional servers can be added and client requests load balanced across all servers. There are two main ways of achieving this:

1. Layering the design of the web service. This involves separating major blocks of code and functionality and distributing them across multiple servers. This reduces the load on each individual server, but introduces additional risks with intra-server communication. A classic 3-layer model separates presentation, business rules and data storage.
2. Load balancing across multiple servers running the same code. Load balancing can be achieved through round-robin DNS or through more advanced techniques such as layer 4 switching. The latter may also assist in improving functionality since many layer 4 switches are able to monitor the availability of servers and redirect client requests away from a failed server. This in turn leads to other risks, such as loss of user session data stored in memory (should the web service use this) on an individual server when the request is redirected to another. In addition, advanced shared storage may be required to enable all servers to read from and write to the same data storage. This may be a database or a replicated file system. Of course, reliability of the shared storage then becomes an issue

Advanced application architectures, such as Java 2 Enterprise Edition, are designed to enable services to be distributed across multiple servers running the same code. However this may be overkill for many implementations and desired results may be obtained more easily be layering an application across high-spec servers and introducing fail-over clustering for each layer.

For large scale critical web services the two methods can be combined, with multiple servers in each layer and layer 4 switching load balancing requests between layers.

How Linux can help

Open source code running on Linux can provide the means to meet all of the aims outlined above. Of course, commercial alternatives also exist, but it is well worth considering open source alternatives if only in a test-bed environment where the intention is to investigate and test different strategies.

Linux itself and high quality projects provide a stable environment for a web service. These projects include the Apache Webserver (http://httpd.apache.org), application environments including: scripting (PHP http://www.php.net, Perl http://www.perl.com/); object-oriented (Python http://www.python.org/, Ruby http://www.ruby-lang.org/en/); non-open source but free (Java http://java.sun.com) and emerging (Mono http://www.go-mono.com/), For java, numerous projects exist that enable it for web services, including Tomcat(http://jakarta.apache.org/tomcat/index.html) and the JBOSS J2EE application server (http://www.jboss.org/index.html).

A range of projects exist that can provide fail-over clustering on Linux. I’ve personally used linux-ha (http://www.linux-ha.org/) in two-node clusters for over 4 years. For more on Linux clustering see http://lcic.org

Linux drivers exist for many common HBAs used to access SANs. In addition, projects listed at http://lcic.org provide alternatives such as file system replication and “network RAID” virtual devices.

The Linux Virtual Server project at http://www.linuxvirtualserver.org provides full layer 4 switching functionality. In addition, iproute2 (already installed with most modern Linux distributions) allows advanced routing to be implemented, including WAN link aggregation and intelligent source-based routing.

Evolution 2.0

Novell Evolution 2.0 will be available in the third quarter as part of Novell's Linux desktop. For more information about Evolution, visit http://www.novell.com/products/evolution . The Evolution Connector for Microsoft Exchange Server source code can be found at http://ftp.ximian.com and developer information about Evolution can be found at http://www.gnome.org/projects/evolution . The Connector code is now available to the public along with the rest of Evolution under the terms of the GNU General Public License (GPL). Beginning May 14, current Evolution users will be able to download Connector for Microsoft Exchange Server for no charge at http://www.novell.com/products/connector/download.html .

File System Factory 1.2

Novell has announced the availability of Novell(R) File System Factory(TM) 1.2 to automate the provisioning and management of user and group storage by leveraging identity information and policies. With File System Factory, IT administrators can effortlessly meet the storage demands of their organization while reducing hardware and support costs. Novell File System Factory 1.2 is currently available on Novell NetWare(R), and future releases will support Windows* and Linux* to help customers upgrade servers and migrate from Windows to Linux.

Novell File System Factory expands the provisioning capabilities of Novell Nsure(TM) through the automation and management of storage on networks with dynamic user communities using the policies and roles defined in Novell eDirectory(TM). When a user or group is created in the directory, File System Factory verifies the relevant management policies and automatically allocates a designated amount of personal or group storage space on the correct server. Likewise, if users move in the directory their data automatically follows them, and as users leave the network, File System Factory can delete their stored data in a specified time period or transfer it to a manager.

"With the amount of user data skyrocketing, the ability to migrate it to a different server reduces the frequency with which I need to do server upgrades," said George A. Raetzke, senior systems programmer for Northern Illinois University. "File System Factory frees us from maintaining a lot of custom account management code, and I have been able to migrate tens of thousands of users during the day without impacting their access or productivity."

With File System Factory, the creation and management of storage can also be integrated with an identity provisioning solution for further automation and control. This allows organizations to use provisioning technologies like Novell Nsure Identity Manager in concert with File System Factory to manage personal and group storage based on a user's identity and role throughout the lifecycle of their relationship with the organization.

"Network storage is fundamental to the productivity of every user within an organization. Yet, providing and maintaining that storage for users can become quite a headache for administrators who are trying to keep up with employee turnover and the growing need for more storage space," said Alan Nugent, Novell CTO. "With File System Factory, Novell is extending the benefits of its provisioning solution beyond managing security to managing storage resources based on a strong identity foundation."

New features in Novell File System Factory 1.2 include:

• Workflow that enables a manager to review, reassign or vault employee data when an employee leaves the organization.
• Web-based administration dashboards monitor and track trends in managed storage for both users and groups.
• Policy-based management allows help desk personnel to increase and manage a user's storage space without providing access rights to the whole file system.
• Auditing and reporting capabilities that allow administrators to see who has rights to a given piece of storage, how they got those rights and the amount and percentage of space available for each user.

GroupWise Messenger Plug-in for Trillian Pro

Support for GroupWise Messenger was recently added to Trillian Pro 2.0 via a messenger Plug-in for the Trillian Pro client. One of the key benefits to using Trillian Pro is that in addition to GroupWise Messenger you can also communicate with AOL, MSN, and Yahoo users all within the same client. Find out more.

http://www.novell.com/coolsolutions/gwmag/features/trenches/tr_trillian_plugin_gw.html

Mono 1.0

Novell has announced the availability of Mono 1.0 Beta 1, an open source version of the Microsoft .NET development platform that enables Linux and UNIX developers to build and deploy cross-platform .NET applications. Sponsored by Novell and led by Miguel de Icaza, the Mono project provides a productive development environment, stable APIs and a well-defined technology roadmap, plus the benefits of .NET for Linux and UNIX.

Mono 1.0, due later this quarter, incorporates key .NET-compliant components, including a C# compiler, a Common Language Runtime just-in-time compiler and a full suite of class libraries. Other benefits of the Mono project include:

• A superior development environment for writing Linux applications with unprecedented productivity.
• Ability for developers to write rich client, Web services and server-side applications and deploy them on Linux, NetWare, Mac OS and Windows.
• Tools that facilitate the creation of product APIs and SDKs that are language independent across multiple operating systems.

For more information on the Mono project and to download the Mono 1.0
Beta 1, visit http://www.go-mono.com .