Quote:
Originally Posted by
port43
At my place of employment the push is all virtualization and moving things to a cloud environment.
Of course it is. Now, have a close look at what "cloud environment" stands for and you will notice that it is some pretty old ideas coined into a new buzzword. This happens all the time in IT: "object orientation" was, in fact, what FORTRAN/77 programmers called "function with multiple entry points" and did 20 years before it was all the new hype; "virtualization" is what IBM mainframes did since the seventieths (there is simply no real difference between an LPAR and a CICS partition); etc., etc.. The same is true for all the other buzzwords we encounter daily. In fact most of them are
old wine in new skins, as the proverb goes - and most of the rest is a bad idea anyway. For instance: ITIL. What is good on ITIL is what every competent admin (programmer, ...) always did anyway, simply because it was the right thing to do. Everything else is bureaucratic nonsense invented by people who never came anywhere near real work at all.
From there you might see that learning and understanding the basic technologies and the problems they are supposed to solve is instrumental in keeping up with the technologic advance.
Quote:
Originally Posted by
port43
So between that and the allocation of SAN to support it, those two groups are hopping.
It is
not about doing - it is all about
understanding. It doesn't matter if you know this tool or that tool, but it is about the fact that you know how zones are set up, what they are and what they are for. It is about being able to come up with a zoning schema for a big data center, which will remain, when applied, consistent even when lots of virtual disks are created and propagated over several fabrics.
Similar for networks: would it be within your ability to create a plan for a corporate-level network with many heterogenous subnets, backup networks, a DMZ, extra-net capabilities, etc., etc.? Only then you have understood networks and the problems their management poses. It is not about being able to configure this or that VLAN on a XYZ-switch. Creating such a plan would need an instinctive understanding how VLANs work, how you assess bandwidth necessities, throughput capacities - such things.
In my current project we manage about 500 virtualized systems, running on several IBM p780 (and some dozens of small systems, p570-p740), we manage the storage (about 4 PB of online capacity, EMC VMAXes mostly, several Brocade FC switches, storage virtualization with VPLEXes), a job scheduling system (BMC Control-M). Boot disks are iSCSI, data disks are NPIV. All this with three people. The question is not "how do i point and click a new LUN for my system" - this is easy. The question is:
how do i consistently create storage entities (aka "LUNs") for the IBM mainframes, the Linux world, our own AIX world and even after having created some 2000 such entities still know what is where.
It takes a lot more than some learned knowledge - it takes the ability and willingness to organize the work, consistently document it (the really essential things, not some managerial blabla), set up and describe consistent procedures, the discipline to strictly adhere to the self-set standards and so on and so on.
This sets apart "big data" from the home-grown eclectic. If one sets up his own system and does it one way and then a second system and does it some other way, so what? If i would install/maintain/whatever every system individually i would very fastly create a mess our limited group of three admins could never sort out again.
This is why i hold my job, not because i am such a wizard with AIX - in fact i am pretty good, but not outstanding. But i can understand, develop and adhere to procedures one needs to do work at such a scale without creating a big mess.
i hope this helps.
bakunin