Forrester Research says Fibre Channel Network Connectivity Costs 3-7 Times More Than iSCSI

Last month (February 7, 2008) Forrester Research published a research report titled “Cost Comparison Of iSCSI Versus Fibre Channel SAN Components” by Andrew Reichman. They provide a detailed and practical overview explaining some often overlooked implementation factors related to installing a SAN. The research focuses on 4 primary factors:

 1. Server side connectivity issues
 2. Switch costs
 3. Array side connectivity
 4. Cabling

Key differences in server side connectivity are that an iSCSI SAN can utilize a variety of approaches beginning with utilization of onboard NICs within a server.  In some cases, server load impact due to iSCSI traffic can require an upgrade to a TOE or a fully offloaded HBA (though our internal testing shows that modern NICs run just fine in typical SME workload scenarios without imposing excessive overhead).  Other incremental costs for Fibre channel come by virtue of having to pay for multi-pathing, both software and incremental hardware, whereas Ethernet/iSCSI multi-pathing comes for free in Windows, BSD and Linux.

Switch costs are higher for Fibre channel on a per port basis due to the fact that they do not enjoy the same economy of scale that we see in enterprise Ethernet switches. A high end Ethernet switch is typically one fifth of the cost per port ($300) compared to a director class FC switch ($1,500). For smaller SANs, the difference is a 50% reduction ($600 per port for a non-director class FC switch).

On the array side, since common server components are used, the cost per Ethernet attach port is also lower. Most iSCSI SANs have two Ethernet ports standard per controller, but can be inexpensively upgraded to 4 or 6 ports (see Wasabi VMX series for example). With 10GbE coming into play at under $1k per port and rapidly dropping, the differential is preserved.

Cabling costs are also often not considered. In the report, Reichman compares 3 meter optical cabling to 6 meter CAT 6 Ethernet and arrives at an average cost of $12 per connection for iSCSI and $50-150 per connection for Fibre Channel.

Putting it all together, Fibre Channel costs 2.5 to 7 times the price of iSCSI connectivity (ignoring the costs of the array and disks). In dollars – compare $400,000 for FC networking to $100,000 for a similar 100-server iSCSI networking.

Beyond the hard cost difference above (which are compelling stand-alone) there are soft costs that further reduce TCO resulting from increased staff efficiency since there is only one network topology to manage. Additional benefits come from having a broader pool of qualified IT professionals capable of managing storage, improved switch utilization due to commonality, more parts sparing, improved SAN security and enhanced virtualization capabilities.

I like the Recommendations section that starts with “Beware of politics and religion” – the main source of objections to iSCSI. Large enterprise Fibre Channel shops want to preserve the status quo and not cede their control and budget to the network IT staff… Managers should be aware of this and look into the cost differences. The report is cheap (less than a single FC port @ $279) and can be purchased here.

A new article in Processor (April 08) cites this research in a very positive way, but attempts to throw some FC FUD at it – good examples of politics and religion mobilized.

Users Flocking to iSCSI SANs

Deni Connor just published an article this week in Byte and Switch that cites a number of reasons why IT managers are deploying iSCSI SANs. 

Key reasons she reports are:

==> They are easy to configure and install,
==> They are inexpensive compared to “costly Fibre Channel” SANs, and
==> They do not require a specialized skill set to manage.

We have known for years that “iSCSI is coming” for these reasons. Ethernet had to win once the performance playing field was leveled. So, finally iSCSI is here to stay – Gartner projects growth to $3B in revenue by 2011.

So what’s the downside? There really isn’t one, other than perhaps some last gasp ridicule from the “big iron” dinosaurs. With 10Gb Ethernet coming into the mainstream this year and next, the reasons for adopting are even more compelling. The old argument that iSCSI has limited performance compared to Fibre Channel is just not grounded in fact.

IDC Says that iSCSI Market Grew 70%

The iSCSI market grew 70% in 2007 according to IDC. This is another solid validation of the technology and it’s growing reach. Key reasons cited by Liz Connor were simplicity, cost and broader application support.

Some other numbers cited: overall growth in fourth quarter for disk systems shipped was up 7.6% to $7.5 billion. Total petabytes shipped was 1,645 petabytes. Doing some quick (arbitrary and only partially scientific) math, assuming average drive size of 500GB, an “average” array size of 12 drives; that means more than 250,000 arrays shipped in the quarter. I may be way off, but it’s an interesting and big number.

Even in a market downturn with fears of imminent recession, demand for data storage keeps moving. As budgets become tighter more people will consider iSCSI SANs to replace their DAS as they seek to consolidate and simplify their storage infrastructure.

Blue Men, Purple Cow and Kingdom Come

I went to a Blue Man Group concert the other night with my wife, not knowing what to expect and I had a blast. From the opening routine, that was somewhere between a Coldplay concert and the ape scene from 2001: A Space Odyssey, to the closing streamers falling from the rafters, I was mesmerized. They guys were very entertaining and the band was terrific. It is clear to me that these guys have their “remarkable” product down pat and deliver value that I will talk about to my friends. I think that Seth Godin would certainly agree that they understand the concept behind “Purple Cow” and are passionate about delivering.

At Wasabi, we grapple with the concept of being “remarkable” a lot these days. I truly believe that we have to bring out our strengths in a way that connects with our customers and prospective customers so that they want to talk about us. I personally don’t think that trade shows and magazine ads are helpful these days beyond getting your name out as a starter. We are working hard on a next generation product that “is” the core of our marketing of the future. In developing that product, I’m just as interested in learning what music our prospective customers like as I am in how big there storage requirement is. Why? Because I think it will help us frame our message and influence how we implement our UI and how we prioritize features. We need to be able to speak their language and frame our message in a way that they understand and relate to. Otherwise, we won’t get noticed.

It’s about being original and being a rock star in whatever you do. I was listening to the Sirius radio on my drive up to NYC yesterday and heard what I thought was Led Zeppelin playing a song that I did not remember ever hearing. The singer, the band, the melody – everything sounded “almost” right. Of course, my wife laughed and said “that’s Kingdom Come” a Zeppelin copycat band. Amazing – same sound, immediately the Zeppelin brand came to mind. Who knows or cares what happened to Kingdom Come (Kingdom Gone?). Originality, rock star, brand – integrating our worldview is critical in our jobs now. Creating a unique identity that people remember is crucial.

Bottom line – storage can be a dull topic. It’s something that is not often brought up at a dinner party or social gathering. But, the companies and people building it or using it don’t have to be. People who are having fun create better products and care more about their customers.

Leap Year!

I won't get to post another entry on February 29th until 2012!  

The next 4 years in the world and in the storage business should be interesting.

Gaming Security - Watching for Cheats!

Last week I attended the World Game Protection Conference held at the Paris hotel in Las Vegas. My motivation was that our biggest market segment today is in IP based video surveillance, primarily in casinos. What a great excuse to take a trip to Sin City…

In any case, this was a fascinating glimpse into the gaming industry and things that cause them to be paranoid. This is a very sophisticated bunch (at least in terms of the back room surveillance operations).

I don’t think most people realize how much money moves around in a casino and the considerable steps taken to keep it once the players give it to them. Conversely, I had no idea what a substantial problem they have in tracking cheaters. You would think they have this all solved and down to a science, but it was enlightening to me to hear how much cheating goes on and how creative the cheaters can be, from individual “advantage players” (that’s the name for card counters) to folks that are slipping in marked cards (amazing how easy and how sophisticated this can be now – think rf chip/xray glasses stuff) to get an advantage in one of the varieties of Poker.

The most interesting presentation I listened to was by Ed Thorp – the MIT Mathematics professor who wrote the original book “Beat the Dealer” back in the 60’s. You can see background here and order the book for Amazon here. This guy was amazing. Soft spoken, articulate, clearly brilliant and funny (in a wry way). He made a ton of money for himself and his backers after mathematically modeling Black Jack and figuring out how to win. Eventually he got bored with this and moved on the stock market where he can make more money with his applied theory / models.

From a storage industry perspective, this is a great market for IP based tiered storage. First tier for real-time floor and back room surveillance – interesting requirement is to quickly replay a winning bet in real-time (to make sure it was legit) for the floor manager before paying off – fast SAS drives or SSD’s. Next tier is what I would consider the “hold file” or cache that gets flushed out after 30 days – high capacity SATA. Finally, having the ability to quickly grab a segment of video, capture it in a way that you can guarantee authenticity and trace-ability down the road for evidence purposes and ensure it is well protected (hmm… objects?).

The game operators are dying for better cameras – everyone knows to look up at the little dome in the ceiling. The technology is there to embed cameras in the tables and even the slots – more video, faster video, better resolution = more storage. I love it! Please help them.

Wasabi's VMX iSCSI SAN products

Today, Wasabi announced it's new storage appliances. This is the starting point for us and we have a number of additions in the queue throughout 2008 and into 2009.

If you did not guess, the VM is short for VMware and the X is short for XEN. We originally designed and optimized these appliances for use in a virtualized server infrastructure. Thus, we tip our hats to the the commercial leader and the open source leader in the space.

Read the press release here: Wasabi VMX Announcement


Look for thin provisioning and snapshots to be added this summer..