logo

I've talked about this before. In many large enterprises, the result is dedicated silos of computing with little sharing for applications. For example, edge server applications (web, file, print) have their own operating systems and hardware, while mission critical back-end server applications (Oracle, ETL, email, CRM) do as well. All these applications need CPU and memory. Why can't they share pools of resources yet still have separation? So in my opinion, the bottom line is that the "Big 3" are driven primarily by leveraging the legacy...but they have customers doing things to their data centers that waste huge amounts of money; create massive complexity; and can't scale without throwing more hardware (and services) at the issue. For example HP sells the following server lines and has been for years: - DL380/D580s for file and print apps - p/c-class blades for blade apps / virtual apps - Itanium servers for OpenVMS apps - Superdomes for heavy HP-UX apps IBM is doing the same thing (x series, Bladecenter, p-Series, z Mainframes). And many clients are OK with fragmenting their server environment because again, they've had this set-up for years. They are OK locking into operating systems like HP-UX or AIX, which tie you to a specific vendor. How can the customer get the best value when their applications are locked on different types of hardware? They have no bargaining power because the proprietary OSs are preventing competitive hardware to be introduced. If you have all different hardware and OSs, think about how that affects disaster recovery and data center management. Not good! The solution:

1. Move towards migrating applications to the most common OSs supported by vendors today (Linux, Windows, Solaris x86 are the most popular). This will give you more leverage with the hardware vendors.
2. Look for single source platform solutions that integrate hardware and software that allow for sharing CPU and memory across different application pools. It's OK to run a VM farm on the same Frame as your mission critical Oracle RAC. You CAN get bare metal performance for your high-end applications on one Frame. You CAN get N+1 high availability, 5 9s reliability, and share CPU and memory without risk.
3. Think like a "client" not just a "customer." Ask for one solution that can solve it all and allows you to make changes as your application set evolves or your business goals change. Do not settle for a piecemeal approach when you don't have to anymore.