Once again, Dilbert voices the truth in the IT world. I get Cloud Computing in a personal context and maybe even a SMB context. But what does Cloud Computing means in a corporate context? What’s really a “Private Cloud” for an enterprise? Just the buzzword du jour invented by the marketing companies and propagated (or propagandized) by the clueless media groping for eyeballs. It is like the 90’s or 00’s again with utility computing or grid computing – just more buzzwords.
Since I found my long lost Security Wiki on Wikidot.com, I plan to write some basics of enterprise computing on enterpriseitfaq.wikidot.com. That would contain enterprise computing hardware platforms, hypervisors, OSes and such. I hope to find and discover ways to integrate between the Wiki and this blog but I am not really there yet.
What makes a server “enterprise” class? Is it big and expensive that makes a server “enterprise” class? Is it the “cool & hip” factor? Is it…?
I would content it is RAS (Reliability, Availability, Serviceability) that makes a server “enterprise” class. IBM is famous for making enterprise class servers (their mainframes and AIX frames). However, I am now including the HP Bladecenter.
I have attended a week long HP Bladecenter class and came away impressed. Years ago, I used HP’s p-class blades and was left unimpressed, as with IBM’s Bladesystem. These are 1st generation blades where they are just vertically placed slimmed down 1U servers. However, with 2nd generation HP blades – c-class blades, the enclosure itself takes on a dramatic metamorphosis. It is now a “system” instead of just a collection of small servers. Virtual connect modules on a c-class Bladecenter allows for high degree of RAS. Properly configured, any blade server in the enterprise (up to 16,000 blade servers) can take on a designed MAC and WWN, thus making rezoning, rehosting and “special” MAC a thing of the past. Once the initial configurations are given to the SAN and Network team, they can be out of the picture for any future add/move/change processes. The OS and supporting applications can be “moved” to another blade server if a scheduled or unscheduled hardware event happened within a few minutes!
What was even more impressive is the “scriptability” of their Bladecenter. A little primitive, Cisco IOS like, but nonetheless, command scripts (and can be ssh to the “onboard administrator module” and virtual connect modules and scripted from Powershell or csh/ksh/bash). It is extremely exciting to be able to open a PuTTY session remotely (thousands of miles away), create some SAN fabrics, network links and network trunking, spin-up a blade, assign the MAC and WWN, RDP or snap an OS image to it and have it up and running. All scripting, all remotely!
The next adventure will be to have VMWare vSphere 4 on a set of blades and also Cisco Nexus 1000V. Testing out the capability of vMotion between enclosures and supporting multiple trust-level networks (separated networks, firewalled zones and VLANs).
I am highly impressed with HP’s Bladecenter and I think it is really an enterprise-class system at this point.
I have been involved in the implementation of a “converged network” PoC (proof of concept) in the lab for the past 2 months. One may ask, “What’s a converged network”?
Up until recently, datacenters are running with both ethernet network and fibre channel network to each server. That’s normally 4 ethernet connections (2 for etherchannel client network, 1 for administration network and 1 for backup) and 2 FC SAN connection – grand total of 6 ports, 2 to 3 ethernet switches and 2 FC SAN switches. Quite a bit of infrastructure that’s needed to support this modern-day datacenter. Cisco’s converged network is aimed to reduce that – 2 Nexus switches and 2 CNA (converged network adapter) to each server instead. The FC SAN connection can fan out at the distribution switch level or at the access switch level. In all, this is a great reduction in the number of switches, adapters and cables. The connection speed at each port is 10Gb; fast enough for FC SAN (4Gb) and a few ethernets!
Currently, our IBM AIX and HP Proliant servers (with appropriate drivers and OS releases) support this. CNA for HP Bladecenter is coming, but not here yet. We will be testing that later when CNA are GA (general availability).
Converged network represents a slightly modified paradigm in how things are in the datacenters of corporate America –not enough to really change things up, but enough to get corporate fiefdoms in a reactionary mode. Questions such as, “which group is responsible for the administration of the switch in terms of access and configuration”, “will there be ‘leakage’ of data from one network to another”, “will the savings be enough to offset the additional administrative overhead”, etc. needed to be answered and fiefdom battles needed to be fought, won, losers vanquished, new procedures and policies written and winners crowned. The final question, really, is, “will the benefit outweighs the hassles?” For a new datacenter or other radical scenarios, the answer may be “yes”. But for the majority, sadly, the answer may be a resounding “no”. Not because of the technology or anything else, it is because of the corporate behemoth is an entity that eschews change.
Greetings!
I am an IT professional that’s been working in enterprise-type computing for nearly 20 years. Yes, that’s a very long time. Although most of my years are spent with using Microsoft type solutions, that is not my only focus. I am versed in most Microsoft server solutions (Windows Servers, Exchange, SQL, System Center, Forefront, Sharepoint, BizTalk, Dynamics CRM and Dynamics NAV), Cisco, IBM AIX, Unix and Linux (Redhat and Ubuntu). Because of business integration needs, I also write programs (for ERP migrations, AD/LDAP migrations/integration, etc).
The company where I am working right now provides some technologies that are “cutting edge” and the working experience of that will be what I am sharing here.
I hope this prove to be interesting and informative. Enjoy!
