I am busy designing and building  infrastructure for various “as a Service” service offerings

Web as a Service (IIS)
Data as a Service (SQL and Oracle)
Middleware as a Service

So I thought I would float a few questions

What is High Availability?
When do you need it?
How do you create it?
How do you measure it?

DISCUSS !!

[UPDATE]

Here is my two penneth,

Historically many bespoke application are developed on the same servers the prodcution application runs on. Often this server was built by the developers with little regards for consistancy. After all developers develop applications for a living, not build server infrastructure.

Developing an application on the production server can lead to many problems, particularly if that server build is not quite right or the developer is a little wet behind the ears and does not considure scaling issues and hard coded resource refernces. Then of course it is always possible to cripple the production server by testing rogue application code at the whim a developer. I could witter on for ages about this, bit i want to get on towards what needs to be done to consolidate these applications to central managed infrastructure. Putting to one side the issues of supporting the, often extremely large, consolidation platform, many problems need to be over come to ensure the successful migration / consolidation of bespoke applications in the enterprise.

<UPDATE>
It is worth pointing out that these applications should be consolidated to a well defined, well managed platform, where each appliaction exists in a logical container where ALL interfaces and features are defined and controlled by the infrastructure. Hopeful reducing or even eliminating many of the future gotchas

It should also be noted the technology is only a small part of the battle. Accurate recording of succesfull migration, ownership, and reporting of KPIs also need to be considered. i am not talking a few spreadsheets, it needs to be a fully operation CMDB. spreadsheets are not searchable on MASS, so without a proper CMDB you still only have a huge quantity of unstructure data which is be out of data before you can use it
</UPDATE>

 So this is my list of check points and Gotchas

Information Required

• Who owns the application (business owner, not a Project manager) ?
• Do they have resource to manage coding changes that may be required ?
• Who Need access to develop the application ?
• Who supports the APP ?
• Does the app have a offical supoprt team ?
• Does the app have approval from the corporate security guys?
• How many users ?
• Does it need access to employ data / GAL ?
• .net version requirements
• Database Requirements (oracle and SQL), linked Servers
• Database requirements (un supported and  will need to migrate)
• Downstream dependencies
• Upstream dependencies
• LOB implications (many corporates have to segragate their LOBs BT Openreach for example)
• Any 3rd party access customers (via portals)
• Any 3rd party access offshore users
• Any 3rd party access offshore support
• Any 3rd party access vendors COTS software
• Quantity of SMTP traffic generated
• Database growth expectations
• File Space growth expectations
• Any confidential data requirements
• Any encryption Key management requirements

Things that may need to change

• Db connect strings
• Tnsnames migration to non tns connection strings
• SMTP parameters
• Cold fusion with ldap queries
• Hard coded File paths
• Hard coded Server names
• Hard coded data (other)

Gothchas

• Oracle client compatibility
• File system access by application
• 3rd party components
• Has it been used to running with elevated privileges
• Do any DTS/SSIS need access to file systems
• Does SQL run as a service account, What perms does this have
• Self referencing http calls such as web services (can break on hardware load balanced farms)
• SMTP components
• Non web based components, windows exe, scripts or services
• Use of MS Office components
• Use of MSACCESS
• Sql optimisation (query and indexes)
• Cold fusion version (4.x , 5 code not 100% compatible with MX series)
• Cold fusion with AD  authentication (will need code change if using bluedragon CFML engine)
• Asp/asp.net auth against local accounts
• .net data components and memory usage
• SSL
• Session management
• Web gardens
• .net assemblies in bin folder
• .net assemblies in GAC
• .net trust requirements
• Frontpage (Yuck)
• WebDAV (Security implications)
• Application session variables cannot be seen by other servers in the farm
• Do the support guys have SA access to the current SQL server
• Do the Support guys have console access to the current  SQL server
• Do the support guys have console access to the web current  server
• Date time issues on code developed on servers with wrong regional configuration

Following on from Outbreak of wordpress, it must be contagious and WordPress MU in a Corporate environment

WordPress MU is now a Live service inside BT, for all employees to use !

The platform is currently a single, yet rather large linux VM running WordPress MU 1.3.3 with the LDAP auth plugin into the Corprate AD.
as the need and used grows this will be fronted by a cookie insert load balancer and more apache/php instance

• The front page includes some blog guidelines and a few a few MU plugins
• Recently update blogs
• New blogs
• Top Bloggers (quantity not quality)
• Global Tag cloud

We have also …

• Disabled the self registration system, as by default wordpress MU gives you a blog name based on your username( a meaning less 9 didgit number in our case).
• Enforced logon for comments on all blogs, and modified the PHP to prevent anyone changing that
• Pre loaded the WP_users table with ALL employees, so no registration is required for a User account (require for comment), LDAP to the AD deals with password policies
• User Blog are (Currently Manualy) created using the email address prefix, replacing “.” with a “_” for john.smith@ get a blog called john_smith
• Group Blogs what ever name they wish (within reason)

Life is full of tools, tips, tricks and technology. However most people fail on the “When to” and “When NOT to” use a particular piece of technology.

For example you would not use a pair of scissors to cut the lawn, or a lawn mower for a hair cut , yet they are both cutting / triming devices !

So why it is we see some people use a full on SQL database to manage the population of a TEN item drop down box and other use MSAccess (and the MSACCESS GUI not a web app) to run a million pound revneue generating mission critical business application for 100 users over 20 sites

I read an article some while ago about the police had just caught on to the fact the sat navs area good place to “hide” SD or other small memory cards, will the sorn off USB cable is even better

Hide a ultra small USB stick inside a broken USB cable, and who is even going to try and plug it in

http://www.reghardware.co.uk/2008/04/17/sawed_off_usb_key/

Just in case any one wondered what I actually do for a living, I build world class web hosting platforms. We are talking up to a Terabyte of content and 8-10 Million hits per day. So when a friend of a friend said he was having trouble with a site getting 190,000 hit per day, i did a detailed analysis and made some recomendations.

 Then I hear they have been and done something completely different, and possibly daft.

 Well I say good luck to him

I am slowing gaining some good experience in coding with visual studio 2008 and VB.NET and some C# for ASP.NET

But I fell into a corker this week

I have a simple form with a dropdown list box, the items in this drop down list box are populated from a data string obtained from an external webservice during the page load, Nice solutions or so I thought

1st problem a page load occurs during a post back, so the dropdown list was getting populated twice, once during the initial page load, and AGAIN during a validation post back on a textbox

So I decided to clear the dropdown list before adding the items

Some hours of frustration later I discovered the clearing and reloading the list items was not such a good plan, Pageload also goes off during a form submission which then nukes the selectedvalue data for the dropdown list

End result, test the dropdown list item count before to see if I have already added any item, of not add them

Sorted
1. My selected values work properly
2. My postbacks are quicker as the webservices call only occurs on initial load

Sat in a Server 2008 session at the UK 2008 product launch,

and they have just announced “Hyper-V goes RC0 today”

Just been pondering the internet and downloads etc and this news item sprung to mind www.theregister.co.uk/2008/02/22/burnham_dcms_filesharing/  in which the government are threatening to put legislation so that ISPs stop their customers taking part in illegal downloads. This is like asking the police to stop every drink driver by stopping every vehicle on a busy motorway and breath testing every driver and all the passengers just to be sure.

The ISPs cant even do a decent job of stopping SPAM either from or to their networks, what chance to they stand of stopping peer2peer activities.

About a month ago I was in an infrastructure meeting discussing backup strategy. The conversation was around disk based backups such as near store and Virtual tape libraries.
So I asked the question “do we know how these disk based backup system cope with earth quakes? So as yesterday data could be on the same disk that is being access today, a disk level corruption could be ‘interesting’”
As the point in time I got laughed at, “Earthquakes, yeh right !”

Does not seem such a silly question now does it !

From BBC “The biggest earthquake in the UK for nearly 25 years has shaken homes across large parts of the country.”