One of the apparent fundamentals in our industry today is that Linux

and Windows are antithetical from almost every angle (implementation,

administration, development and philosophy). However, in spite of

this, the reality is that almost every computer based shop has some of

each (although usually with a bias one way or another).

This talk will investigate the development and philosophy divide with

a case study (and various anecdotes) of the SteelEye replication

product for Windows. SteelEye, in fact, has two replication products:

one for Linux being open sourced and upstream in-kernel (based on

md/nbd) and one for windows which is closed source and implemented as

a filter driver. A while ago it was noticed that the windows product

was performing badly in comparison to the Linux product on WAN

connections which were high latency, low bandwidth. Our solution to

this was to take the engineers responsible for both the Linux and the

Windows product and put them all on to the task of improving the

Windows filter driver. The result, a few months later, was a Windows

replication product that matched the Linux product in terms of speed

and bandwidth utilization. This talk will follow the trials and

tribulations of the combined team, answering the questions

• How did they work together (and how transferable were the skill sets)

• What were the problems and how did they solve them (and what role did Linux skills play in the solution)

• Is there a problem of GPL contamination in moving from Windows to Linux (and vice versa).