Blade servers will continue to be all the rage for some time to come, but there seems to be a growing backlash against the idea that blades are the best solution all the time. In a nutshell, people are starting to wonder, is there a right way and a wrong way to use blades?
As evidence, we give you this InfoStor Magazine piece from Dave Vellante, co-founder of The Wikibon Project, which helps foster advanced technology adoption. Vellante spells out the reasons blades are best suited for things like Web serving, e-mail and parallel analytical applications, but not so much for transaction-based heavy workloads with high read-write ratios. There's also the issue of storage to consider. Since storage has been stripped from most blades, will the cost of additional SAN infrastructure outweigh the savings you get with blades?
This blog from Dell server manager Glenn Keels also raises a lot of questions regarding blades' supposed ubiquitousness. Blades are designed to be modular, meaning they don't have the flexibility to adapt to unique situations and can thus add to data center complexity. You also run into power and cost issues when you push blade densities to the max.
The idea that general-purpose blades are good for whatever ails you may also be coming to an end within the vendor community. Witness IBM's new Information Server blade, which integrates the Information Server platform with a BladeCenter box to specifically target data integration and business intelligence issues that arise in the wake of, say, company consolidations and mergers and acquisitions. The package provides tools like the IBM Systems Director and Tivoli Workload Scheduler to simplify management and improve grid and virtualization capabilities.
Another specialized blade is the Satori Server 12000 by start-up Dataupia. Pitched as a data warehousing tool, it's based on a dual-core Opteron blade design featuring 2 TB of Seagate storage. Although it lacks its own relational database system -- relying on Oracle, DB2 or SQL Server instead -- it does lay the foundation to bring massively parallel processing (MPP) to symmetrical multiprocessing (SMP) environments. The system provides near-instant data warehousing capability that can be easily scaled up by adding more blades.
There's every reason to expect to see more specialty blade designs in the future, considering the market for general-purpose systems is quickly reaching the saturation point. In hindsight, it shouldn't be any surprise that those who claimed blades would take over the data center were wrong. Useful as they are, enterprises today are simply too diverse to be completely remade by any one, single technology.