Ethernet Wins -- Again

Doing what I do for EMC, I always get asked certain questions along the lines of "do you think that technology X or technology Y will win?".

That's a bit of a trick question, because no technology rarely "wins", but -- that being said -- there are pronounced shifts to preferred approaches in our industry.  The current shift towards x86 CPU architectures is a good example -- sure, there are alternatives, but you're seeing a lot less use of them going forward.

And that's just what's happening in storage networking ...

Shifts Happen

In the storage networking world, it's a pronounced shift to ethernet-based approaches.  Whether it's iSCSI, NAS, FCoE (or some hybrid of all three), it's pretty clear that ethernet will be the preferred approach for connecting servers to storage devices.

As with any industry shift, EMC has a role in both accelerating the transition and helping customers move from their existing world to the new one.

We did this in the late 1990s when FC and SANs first emerged.  FC replaced trusty SCSI cables, and there was an entirely new topic of "storage area networking" to discuss.  Not only did EMC supply products (remember how EMC used to own McData?) we invested heavily in qualification (EMC e-Lab), consulting services (SAN design and migration, for example) as well as mission-critical customer support.

And now it's time for another transition ..

Moving Our Customers Forward 

As you go through the press release, you'll see some important elements:

• a full range of professional services (assessment, planning, design and implementation), delivered both through EMC and our partners.
   
• an enhanced set of tools to help with analysis, sizing, design and migration (again, available directly through EMC or our partners)
   
• Training and certification of EMC technical consulting professionals on not only the products and tools, but the methodologies and processes.
   
• EMC resale of both Brocade and Cisco products in this space.
   

What This Means

If you think this means that technologies such as FC are "dead", you'd be wrong. 

FC has been exceptionally successful in the industry, and will be actively enhanced and used widely for many years to come.  Customers don't usually like ripping and replacing stuff that's working, and -- in every sense of the word -- FC is working, and working well.

What it *does* mean is that we expect most of the new builds going forward to show a strong preference for ethernet-based technologies.

As customers build private clouds (or whatever you want to call fully-virtualized environments with different operational and consumption models), they're showing an overwhelming preference to do so with the technology of the future: our old friend ethernet.

And I think you're seeing the same sort of comprehensive approach from EMC as we've done in the past -- creating a full portfolio of products, tools, solutions and skills that help our customers move forward in a way that makes sense for their specific environments.

Plenty of customers will want to keep on exactly what they're doing today, and continue to extend the investment they've made in FC.  We've got them covered.

Some will want to build all-new environments with the latest technologies, and won't be too worried about the legacy.  We've got them covered as well.

But I think the majority will start slowly and incrementally start to move more of their environment to ethernet approaches and run in both worlds for the near future.  That's how the real world usually works.

And I'm very happy that we've got these people completely covered.

Comments

[nate]

I wouldn't call FCoE ethernet, it's really not. Unless you can plug it into a regular ethernet switch it's not ethernet in my opinion.

ranted about it last year
http://www.techopsguys.com/2009/08/17/fcoe-hype/

Not that I have a problem with FC, but I do have a problem with folks out there trying to blur the lines between FCoE and regular old "E" as if they were somehow the same thing.

Some vendors out there are pitching ATA over ethernet as well, never used it but at least they're using real ethernet.

http://en.wikipedia.org/wiki/ATA_over_Ethernet

[Joe Svankanski]

Hi Nate,
I'm not sure that converging regular and disk I/O traffic on the one NIC is such a great idea. I prefer to segregate FCoE traffic and IP traffic separate NICs, switches, common parts inventory.

FYI: AoE uses MAC addresses and is not routeable, unlike iSCSI which is routeable. I've only used AoE with 1G-E so can't comment how it works with 10G-E.

[Stuart Savill]

Hi Chuck,

Interesting article - and a topic that I have been tracking for some time. The whole discussion around protocols, pieces of wire has been an interesting one...

Wrote a bit of a blog article on the topic here: http://www.stuiesav.com/2010/05/confusedobject-storage-bycast-bygone.html

and guess i came to the same sort of conclusions - ethernet will ultimitely (and is already win).

10-Gig-E is becoming commodity and with it performance and all the good stuff that allows us to make choices once more.

It will be interesting to see how much DCE / DCB (What ever you want to call it) and FCoE will become the defacto Vs people starting to use file systems again (i.e. NFS / CIFS etc) over more traditional 10-GiG-E...

Its also very interesting to watch the convergence of technologies starting to happen in the end storage device (merged technologies around Clariion / Celerra / Atmos etc being case and point)... Things are getting really interesting...

Thanks for the article!

Cheers

Stuart

Twitter @stuiesav

[Brian L. Hummel@essay]

Nice article. Very meaningful. BTW thanks for sharing us information about protocols. I've been surfing for this too long.