Storage arrays in 5 years.

I don't know exactly how it will change things, but virtualization has to affect what storage arrays will look like in 5 years. The virtual has to be tied to the physical. Managing the virtual will bring new challenges to managing the physical.

The documentary series about what EMC is doing internally to create a private cloud has been pretty interesting on what a data center will look like in the future. The index for that series is here.

I've been looking at a couple of interesting storage related technologies in the market that I thing will change things.

Would be interesting if anyone has any experience of using any of  the following technologies, or any opinions ?

Solid State will definately change things, open up more possibilities and create alot more flexibility in how things are configured/set up.  I did notice an interesting technology:
1)Solid state in the form of  PCI Express solid-state drives that sit on the system bus of a server.

This one I'm not so sure of, but  thought was interesting because as they work quite differently compared to traditonal storage systems, and are newish...

whether they are better suited to a specific market or type of data, a specific tier of data in the real world...I'm not sure:

2)Clustered storage systems. From what I've read, they use very large amounts of cache, a number of these systems use RAIN, as opposed to RAID (how the data is stored and protected), tend to use SATA disks as opposed to Fibre Channel Drives, but do boast very high levels of perfrmance and protection.

Whatever the future holds, I am sure you will be able to manage it from your iPhone application! 

ck

• types of connections for hosts

• In high density environments I believe the trend is FCoE at the moment. For regular racks with 42 hosts in it, I think separated FC and eth will be there for a while.

• types of connections for storage

• FCoE is the latest trend, but that's only when eth is needed, like in NAS solutions with CIFS, NFS or iSCSI. In FC only environments perhaps the FCoE connection will be there, but only FC is used ?? Perhaps that in 5 years everything will be running on really fast eth like 100Gb or so ?

• types of storage

• I believe SSD or any flash based storage will be used in 50% of all storage solutions and 100% in standalone hosts or even your home pc. I think every home pc will have SSD in 5 years. For long term storage and low requirements in performance SATA2/3/4 will be around for a while. FC will be the new tier 2, where SSD is tier 1. I think in time even SSD will be replaced by faster types of storage: RAM based.

  SAS ? I have no clue. I've heard HDS guys say that SAS is the new FC and that FC will disappear completely. So what makes FC so much better than SAS ? The command queue ? Perhaps SAS2 or whatever the next gen will be called will be equal to the current FC ?

• RAID levels used

• No change there. RAID1, RAID5 and RAID6

• changes in caching

• Simply more cache since RAM prices are always dropping. I'd say 128GB or 256GB in a Clariion or 4TB in a Symm are not unthinkable.

• and more (virtualization, bussiness continuity, disaster recovery .....)

• RAID1 where disk1 is in data center 1 and disk 2 is in data center 2 ? But then agani, latency will be a problem and 5 years is not enough to overcome that part. I don't know....

Hi Joe,

- possibly drive vendors offering an encryption service within their f/w, with key managament offered by the array.... maybe ?

- DC voltages required to sustain the logic levels driven down to reduce power consumption... maybe... (well, I suppose that this does not necessarily apply only to RAID arrays) ?

I think that in 5year perspective EFD would be much cheaper and more affordable for all enterprise- and mid-level storages, so no need to use HDDs (probably at all!) -> less power consumption & no need to use HDD form factor -> smaller size of storage system and new interfaces for modules etc...

BR, Evgeny

Interesting question. Taking current trends like "green" and "cloud" or even EMC's own FAST in to the equation should give us some sort of insight.

With the current development we are seeing in the SATA area, we will find that the rebuild times take too long. SUN already introduced triple parity RAID in their ZFS which is a good thing with the 2TB+ SATA drives we are seeing. Since we need speed to put all this data somewhere and to be able to actually do something with it, I think we will see some developments in the FCoE corner, and I am not certain, but Infiniband might still play it's part when it comes to the interconnects between the servers and their storage.

We are seeing smaller form factors for flash based media, but at the same time have the problem that with increased 'ease of provisioning', the demands for capacity will increase and we will continue to add more weight and power consumption to a single rack, simply because everything got smaller (except the size of the disks ).

I expect to see a change in the way we provision. SLA's won't just be defined inside a single box or product, but will allow to traverse the various storage tiers/platforms. Simply put, we will see FAST V3 where we will define an SLA on our data/capacity and things will be moved not just between disks in one box, but between the various entry/mid/high-end arrays according to the customer requirements. Simply put, the entire cloud development will require this from us since we will add meta-data that tells us what piece of data has what SLA. Cloud will also be more deeply integrated in to the products (the start of this trend change can already be seen in the V-Max).

One important change will also be in the way we handle our storage. Mangement needs to be easier if we want to cope with the larger amounts of raw disk capacity.

All in all I can only say that a lot is going to change. It's going to be interesting if you ask me.

I think will change the virtualization, Performance, Pricing etc.

Shahed

Asst.Manager,Customer Service

SCA IT Ltd

EMC Authorized Partner