eWeek - Virtualization in general
orbist 060000HPM5 Visits (5357)
Over on eWeek there is a long (24 min) podcast interview hosted by Mike Vizard, and the director of product marketing for IBM System Storage, Charlie Andrews. Charlie explains amongst other things why the rise of virtualization is going to lead to tighter coupling between server and storage management.
From a 'techies' perspective its also interesting to relate this to what I see from talking with customers and what I see as being possible with the new breed of storage products. More importantly why a product like SVC is key to the easy adoption and infrastructure capability to make the best use of yesterdays, todays and tomorrows storage innovations.
Here we are really talking about the Open Systems environments. System Z folks have been taking a lot of these things for granted for a long time. Not everyone can afford, or sees the need for a mainframe system, but infrastructures grow, and the Open Systems market has a need to provide just the same consolidation, virtualization and integration benefits.
IBM Systems provide the ability to virtualizse System P, IBM i, Storage and System X using a common set of block or file based storage virtualization solutions. As Charlie discusses, capacity utilisation is one of the key drivers in this area. Mike asks some key loaded questions about performance and consolidation, and while its true that storage virtualization provides users with the ability to consolidate, there is always a trade off between capacity utilisation (consolidation) and performance. Afterall, todays HDD based disks still only provide a fairly fixed 200-300 IOPs per spindle. So while you can easily consolidate and get to more than 99% capacity utilisation, performance characteristics do need to be considered. However, thats where up coming technologies like SSDs, Racetrack or any other 'storage acceleration' physical technologies play directly into a virtual storage environment.
The cloud is in the storage. Much as I hate using such 'buzzwords' - you can think of the cloud as being a "single resource that provides everything to everyone" but ultimately there has to be different cost and different performance characteristics from the underlying systems. The key is that users don't want to manage it all. They just want to point, click and provision, and let the system manage it from there on.
Migration, hot-detection (not just archive ability like Compellent) but real time data allocation, that is agnostic to the actual physical mediums. Forget the protocol (FC, FCAL, SAS, SATA, iSCSI, FCiP, FCoE, DCE...) its irrelevant, what users want is data mobolity, server mobility.
Ultimately I think Charlie and Mike hit the nail on the head. The Management, and ease of management is key. Today we have to decide if its RAID-5/6 or RAID-1, tier 0 or tier 3, FC, SAS, SATA etc etc - thats not what you want, you just want some bit buckets that provide the performance you need, and is reliable. I see the RAID type discussion going away, all storage will be commoditized as "protected", and as SSDs reduce in price, the active, vs archive, vs retention will turn into SSD, HDD, Tape.
Management, simplistic management, common provisioning (with subsystem intelligence) that can maintain data access characteristics, performance requirements and minimal intervention, thats where we are headed.
A system that can evolve to match todays needs, while taking into account yesterdays characteristics, and pulling those metrics into a longer term window means you can have minimal SSD, some capacity of HDD and vast retention (Tape - WORM - MAID).
While today we think in silos, servers, storage, infrastructure, the future is much more integrated, a "VM" will provision the "server", "storage", and how much of the "infrastructure" - it can make sense to think of a single vendor solution to provide all three, yet, one the of major values in storage virtualization (SVC especially) is that it is open and lets you purchase the physical bit buckets from the vendor of your choice, (this if you like has caused the most "adoption issues" from lumbering monoliths) yet the small startups are happy to embrace such a concept, it may seem quite unique to find a large vender like IBM embrace such an open ended game changer. I see the same happening in the server markets... and the openness of your vendor(s) is likely to dictate which of them survive as we see the next round of consolidation (at a much higher level).