Blogs können ganz schön lang sein - zumindest in den "kulturlosen" USA. Und sie haben auch etwas zu sagen. Wie dieser Eintrag vom 1. März hier:
"Storage Decisions in a Down Economy
Forbes sent out an email today warning that a recession is coming and stating that the market always tends to lead the economy as a whole by 90 to 120 days. So, let’s gird our loins and face the fact that a recession might be in the offing. It will cause dislocations (as all of us know who remember the last one) and it will lead to some tough decisions regarding IT infrastructure.
Companies will need to decide on a real method for constraining storage growth and for reducing management costs. That much is certain. With name brand storage and FC fabrics conspiring to make storage equipment buys between 30 and 70 cents of every dollar being spent on IT hardware, it is doubtless that storage will become one of those laser-painted priority targets that the front office guided missiles will lock on to immediately. You thought we had “do more with less” a couple of years ago, stand by for the real deal.
We could build a decision tree to describe what business IT bosses will be noodling.
Do I buy brand x storage exclusively because (1) a name brand vendor is better known in the front office and more likely to get the nod on a PO and (2) because well heeled tenured companies are more likely to weather the economic storm than newbies?
OR
Do I think about lower cost alternatives to name brand and emphasize above all else management tools to enable me to do just-in-time storage, leveraging off of the falling cost dynamics of the underlying commodity disk?
Approach 1 has its merits to be sure. Homogeneous architecture can be managed quite effectively (in most cases) using the management tools that come with the array. Going all vendor X gives me fewer vendors to manage and hopefully fewer interoperability difficulties to surmount.
The problems with approach one are also worth considering. When you go homogeneous, you lose bargaining power. You have no alternatives to the vendor’s gear you are buying, because it would cost too much to rip and replace what you have. You are locked in.
Second problem is with the assumption that the management tools supplied by the vendor will be what you need to manage growing capacity with the same staff (or probably a few less hands). All management tools are not equal. Some don’t work across all of the gear that you deploy in your single vendor fabric (which is usually a kluge of the products of several vendors operating in an “ecosystem”). Many hardware vendor management tools, as discussed in this blog, do not give you full visibility into the very gear that they sold you.
Ask around, you will hear what I hear from customers who have to pay 100 percent of the software costs for value add software on the array or switch and then have to dip back into the till to buy many more software products from third party vendors because the tools provided don’t work as advertised or aren’t best of breed or make you blind to the inner workings of the platforms where they are installed.
Third, interoperability difficulties, as mentioned above, do NOT go away because you elected to do the one stop shop. Finisar had a great slide in the presentation they made to me today. They refer to FABRIC BLINDNESS — (SPOILER ALERT for a forthcoming column in ESJ.com) “which referred to a condition that occurs far too often with Fibre Channel “SANs” when fabric complexity, switch, storage and host bus adapter heterogeneity, software- or hardware-enabled virtualization, and technology change over time combine together to make it impossible for a storage administrator to understand the cause of any “SAN” issue he is facing.”
You can’t avoid interoperability problems. Like the data deluge itself, they are inevitable.
There is more that can be said (and probably will be in the comments on this post), but let’s turn to option two.
What if you elect to cobble together infrastructure on your own? What if you build out a heterogeneous environment – preferably one based on a real network but always one that is appropriate to the applications you need to support and to the business processes that they, in turn, support — and prioritize above all else, management?
Risks include interoperability problems to be sure. Can these be managed proactively through a combination of real time oversight technology and platform agnostic SRM?
There is also the issue of staff sizes — not only is staff expensive, but if trend watchers are right, qualified bodies will be in increasingly shorter supply. Will fielding a really qualified storage management staff cost more than going homogeneous? How could we possibly know?
I, for one, would really like to see a qualified report contrasting the costs of mainframe infrastructure and the costs of “open systems” infrastructure, but I have never seen this research (maybe I will dig into it myself). We know we threw the management baby out with the bathwater when we left Big Blue’s nest (coincidentally, this began to happen during the last recession, when IT managers got fired — finally — for buying IBM at a premium price). Looking back, does this tell us anything about the differences between building storage to purpose rather than buying one size fits most?
Another issue: Can a purpose built model be sustained over time? If open standards (for storage) are meaningless — as most folks seem to think that they are — what is the protection afforded to storage investments — whether purpose built or one size fits most?
These are tough questions for which there are no easy answers. Corporate culture will doubtless define the outcome.
Conservative companies will try to take the safe route — either buying one stop shop from name brands or outsourcing to another set of three-letter acronyms (EDS, CSC, etc.) This strategy will have a well defined price tag hanging off of it. Conservative organizations like that. Predictability is everything.
Only the rule about outsourcing (which in my book is very similar to insourcing from a homogeneous infrastructure vendor) is that most deals sour within two years. Expectations raised too high by the aggressive vendor are unmet. SLAs set too high by the nervous consumer are unmeetable. Agreements are dissolved almost with the rapidity of a Britney Spears marriage.
Which leaves the alternative. We make intelligent and informed decisions based on an understanding of process and application and using about the only open standard that everyone takes for granted: the IP network.
Will it solve all the problems? Of course not. But it is the only long term strategy that I have seen in my 20+ years in IT.
For now, IP storage might not be the best performer for high performance transaction processing systems. (Remote Device Memory Access (RDMA) might fix that gap, by the way.) So, a niche market will probably remain for FC. However, the preponderance of applications are well served by IP storage. What we need to focus on is making infrastructure manageable and placing as many provisioning and protection services into the network as possible where they can be made readily available to all of our spinning rust.
That’s my take and I am always interested in yours."
[source: Jon William Toigo, drunkendata, 1-3-07]
