So, personally, I think this is a big deal.
I’m on record saying – hey, I think NAS + iSCSI (and post-facto, I’ll add Object storage as I’ve become wiser, though still learning) will “win” long term (note that “win” doesn’t mean exclusively).
BTW – if you want to see the dialog/wager, you can read it here. One way or another, someone’s getting a nice bottle of wine! But.. Over the last couple years, I’ve learned a lot…
Interesting stats on FC storage target FC vs. Ethernet storage targets per year. Thank goodness EMC is an IP storage leader, that graph isn’t looking good for FC in 2009, and I bet 2010 didn’t fare better :-)
Our growth in the midrange, backup and other spaces is almost all in Ethernet-connected storage (see the Q4 FY’10 results – which were very good here).
I’ve come, over time, to understand Chuck’s perspective more clearly – and that expanded understanding is one of the reasons I think the the native FCoE software stack form intel is a big deal.
At the time of the debate, I was coming from the commercial (mid-sized customer) side of EMC where I came in via acquisition of a company that had a long history with iSCSI targets, and software initiators (so I certainly had a bias).
In that commercial market, customers have little to no existing FC infrastructure, iSCSI is a natural shared block storage choice (always coupled with NAS).
Conversely, over the last few year, I’ve meet customers with thousands of existing FC ports they are happy with, and there, it’s a different story altogether – they have considered iSCSI completely, and rejected it almost absolutely. As is often the case, perspective matters a lot.
As an aside, that Chuck/Chad public debate is a bit of an interesting archeological artifact. You can see CLEAR evidence of Chuck dragging me kicking and screaming to blog.
While I still think you’re wrong on this Chuck (iSCSI will win over FCoE by port/target count, revenue, various measures) – I respect that you’re right on another note – for a large existing FC shop, there’s little of interest in iSCSI, and they need to have a good evolutionary path to a converged 10GbE Ethernet network.
BUT – that’s why today’s announcement is a big deal. FCoE’s strength is that it enables convergence for that huge install base of FC customers. There’s a consideration when you’re buying a new host about whether you want a converged adapter. Where this becomes a no-brainer is “what if iSCSI/FCoE/NAS/LAN” is just part of the NIC (particularly if it’s LOM)? And what if the acquisition price is way lower than having separate Storage/Network infrastructure?
So – whammo, today Intel formally announces what the X520 software FCoE initiator. It looks like this:
The x520 is one of the most robust, more popular 10GbE adapters in it’s own right. You can also get it in the Intel® 82599 10 Gigabit Controller family, which may be embedded on your server MB.
EMC and Intel are VERY close partners. Since everything we ship as a product runs on Intel (we are fully invested into x86 as the winning wager over custom ASICs and other hardware), and we ourselves use their tech for much of our 10GbE stuff – it’s no surprise that these Intel software initiators are fully supported by EMC and on the E-Lab Matrix with our VNX (and previous generation) Native FCoE (and frankly all FC) targets DAY ONE.
You want iSCSI? Great. You want FCoE? Great. You want NAS? Great. You want object storage? You want it all over 10GbE as part of your converged network strategy? Great – I’m glad that EMC and Intel can help you.
Long and short – this is a big milestone into making the “what protocol do I use” a “100% moot point” at this point – the debate is over, Ethernet has won, and the only question is are you running FCoE, iSCSI, NAS, Object, or any combination thereof.
If THIS is the debate you’re having (“no it’s FC! No, it’s Ethernet!”), or someone is trying to start that debate with you – with some caveats, I’d suggest you’re talking to the wrong people.
BTW – I’m checking in to see when this will appear on the VMware HCL. Stay tuned for that update.
Chad,
While Intel is the market leader for Ethernet (and 10GbE), for FC customers that are looking at converged networking, Intel is not the incumbent. Both Emulex and QLogic have LOMs that can provide the full suite of protocols - iSCSI, FCoE, NAS, LAN - too. Pricing for Intel isn't even better than many of the options out there. I did a full write-up at http://wikibon.org/blog/hp-and-intel-help-open-the-fcoe-market/ - it's my take that this is a big deal, but it's likely going to take Intel a couple of years to prove themselves to the storage (SAN/FC) teams and Emulex and QLogic will be working closely with customers and partners (EMC still uses both in plenty of embedded solutions) to not get steamrolled by the juggernaut.
As for VMware, it is more than a simple qualification, there would need to be code in the hypervisor, so if you'd like to share how any future releases would impact this, would be great to hear.
Thanks,
@stu
Posted by: Stuart Miniman | January 28, 2011 at 10:07 AM
Chad -- I forgot all about that dialog over three years ago.
It looks like that -- yes -- iSCSI and NAS protocols will probably be greater than FCoE revenues by the end of 2012, but that will largely be the result of unmigrated FC. Looking back, it has been tough sledding for FCoE: stalled standards, missing functionality early on, and not exactly a boatload of clamoring customers as a result.
I should start picking out that bottle of wine for you ...
-- Chuck
Posted by: Chuck Hollis | January 28, 2011 at 11:02 AM
I agree that iSCSI will take over fiber channel and that FCoE is a great technology. The Intel X520 is a great entry level product. I would venture to say that most Storage Administrators would spend the extra couple hundred dollars to get a Qlogic or Emulex CNA though.
Posted by: pslager | January 28, 2011 at 01:39 PM
Come by the Intel both at Cisco Live in London this week or the VMware Partner Exchane in Orlando next week to see OpenFCoE working in VMware. I will be demonstrating it in a technology demo at both events. I am showing the X520 adapter running both iSCSI and FCoE concurrently.
Brian Johnson
Intel Corp
Posted by: Brian Johnson | January 31, 2011 at 01:18 AM
@pslager
Yes, but it would be a MUCH more difficult story to sell the $1,200-1,5000 cards if the 2-4 on-motherboard ports were already CNA's. This is where I see a huge potential win for Intel, they take back marketshare from Broadcomm and from the CNA vendors. Of course it's going to have to be a lot less than $400 per port before HP et al change over from Broadcomm chips that cost them probably way less than $1 per port.
Posted by: Andrew Fidel | January 31, 2011 at 11:31 AM
I have also heard about another way of pdf recover saved form information
Posted by: qwertyliber | February 15, 2011 at 10:53 AM
Ethernet's been blowing them away for a while now.
Posted by: Church Contribution software | February 15, 2011 at 04:14 PM