Bladewatch.com

I was having a chat with a friend the other day about how things have changed only recently, it’s interesting to see how things progress.

He was telling me about 18 months ago the default standard at their organization was built around two main server types, there was the ‘standard’ and the ‘performance’ specification server.  In summary and not being too vendor/organizational specific:

Standard: (might be a web or batch server, even a domain controller)

• 2 processors
• Up to say 8GB RAM (newer servers might support more)
• At least four hard drives (two for Windows, two for Data)
• Two network ports

Performance: (might be for a batch/database or ESX host)

• 4 processors
• Up to 16GB RAM (newer servers might support more)
• Up to four hard drives (most data would be SAN attached)
• Many PCI slots to host network or fibre cards

More recently however, there has been a change a downsize if you wish. There’s been a change in the standard specification box from 3u (a DL380 let’s say) to a 1u rack mount or blade equivalent. The 1u server can support several of the small form factor drives, it will have typically two network ports (in case you need the second one for backups/clustering), they will have two multi core processors so can cope with typical workloads.

With this in mind, we might change the standard to a 1u box:

• 2 multi core processors
• 8GB RAM
• Local small form factor drives

There are of course exceptions to the rule (as we’d always expect), it’s just interesting to watch that just a few years ago colleagues were talking of the end of the 1u server, “not enough storage”, they’d say, “I need more than 2 pci slots” or more memory support – use a virtual machine.

The vendors seem to be catching up with this and making the 1u, not only more powerful, but scalable. But then are we not changing the nature of the server? Do I not just need a network attached memory and cpu device?

• As we virtualize, does the underlying tin need to worry about specifications?
• Do I not just need a box that I plug into a thing that makes it work?
• Do I not just move the virtual machines around the network as performance/business requirements require it?
• In essence does the number of virtual machines per server matter? Particularly in a world of commodity, where I can simply rack another 1u at low cost than provisioning high cost high value servers? It’ll depend on your strategy, platform and energy policies/constraints.

We’ll see what matters to the end user is that the newer 1u servers we’re seeing (and have seen) coming on stream are bringing more options in terms of performance and connectivity, whilst at the same time improving their management which has to be a good thing for the end user community.

Are the higher specification servers and blades being undermined? No, I think we’re just seeing a movement more to extremes as we see with the car market, the middle of the road market (products) need to justify how they add value, find their ‘niche’ and operate on that basis, not worry about the other markets. Concentrating on your core business what you do best and building on that is the way forward.

Let’s not forget, the 1u server might be fine for Janice’s real estate business where she needs an email server, somewhere to store some files and an internal web server. For Mike though, who needs to provide a remote office with a virtualization platform, to host their file server, a few application servers for Citrix, the 3u will live on. We have different servers, different vendors and platforms (x86/PowerPC/Sparc) for different markets and needs, you as an end user need to establish which ones work for you and your business.

No related posts.

Related posts brought to you by Yet Another Related Posts Plugin.