4 posts • joined Thursday 20th December 2012 01:59 GMT
AH. Nothing is for free. Lets take a simple case of Exchange. You can purchase a storage solution that can handle 20 thousand Mailboxes (published Microsoft ESRP number) across 2 HA sites for 0.33c per user per month. Microsoft Exchange in the cloud costs between $4 and $8 per user per month. That leaves a lot of profit margin in there for Microsoft if you buy their software on a perpetual license or in the cloud.
It would take a lot of software and server hardware and services/operational costs to make up the difference.
Here is a blog on the subject:
Bottom line. They have to give you free stuff to make it anywhere near as effective from a performance standpoint as an in house solution and as Lusty comments it is not real storage anyway. It is a cache and you can get a hybrid storage solution that can do that internally for all of your workloads very cost effectively.
The only way is hybrid!
Yes. The only way is hybrid. The problem with EMCs implementation in VNX is that is is still multiple OSs, a mix of RAID groups and RAID groups underyling disk pools and all the SSD resources are for caching are mirrored so are in effect less that 50% utilized.
If you read the Goldman Sachs case study from EMC they are in fact recommending 12.5% of SSD to deliver 5% of usable capacity. I wrote about it in my Blog if you want to investigate the numbers http://blog.starboardstorage.com/blog/bid/209395/How-to-choose-SSD-How-much-flash-is-enough
The fact is that other implementations today are delivering flash at 4-8X the efficiency of EMC by seperating read and write caches, virtualizing across more workloads, having superior and pervsive data reduction techniques and removing the complexity of traditional RAID Group constructs from the SSD/Flash layer. The solutions are out there is you look. This is not a vision from EMC it is a roadmap forced on them by the complexity (and age) of their portfolio.
All Flash. No Cache
For the majority of customers a hybrid approach will be more mainsteam. There was a great article by Dave Floyer just published on Wikibon showing this. : http://wikibon.org/wiki/v/Hybrid_Storage_Poised_to_Disrupt_Traditional_Disk_Arrays
If you want to eliminate SILOs a unified hybrid approach will encompass more workloads.
There is definately a place for all-flash arrays but hybrid will capture the mainstream.
Lots of great comments here. NetApp is a 20 year old architecture that it is hard to innovate on top of but so are most of the other big storage vendors. NetApp did do a lot right and one thing that is still key for them is that they are unified. I agree that block behind file is not the route to go but consolidation to eliminate silos and stranded storage is effective when done well. There is a new wind blowing through storage and Starboard Storage who I work for is a good example of a vendor riding the wave. We have a Block based system that also does NAS in a single Storage OS that is our IP. Enterprise feature set, solid state caching, dynamic storage pool that is truly virtualized for all workloads.
Storage simply costs too much and consumes too high a percentage of the IT Budget. With new architectures you can get the performance and cost of capacity you need affordably and with virtualization you get both performance and capacity on demand..
I also agree with comments on deduplication vs. compression. Brute force hardware (larger disks effectively managed) and compression, cover more applications with less headaches.
- World's OLDEST human DNA found in leg bone – but that's not the only boning going on...
- Facebook offshores HUGE WAD OF CASH to Caymans - via Ireland
- Microsoft teams up with Feds, Europol in ZeroAccess botnet zombie hunt
- Three offers free US roaming, confirms stealth 4G rollout
- Justin Bieber BEGGED for a $200k RIM JOB – and got REJECTED