Moderator: Welcome to our Live Expert Q&A on "The Future of Fibre Channel." Joining us today is Skip Jones, chairman of the Fibre Channel Industry Association (FCIA) and director of planning and technology at QLogic. We also have with us FCIA President David Deming. Welcome, gentlemen! Let's get started with our topic...

bvigil134317: Can Ethernet efficiently handle block I/O?

David_Deming : In most of the applications where Fibre Channel has been deployed, the desired bandwidth was not available, even though Gigabit Ethernet was installed. Much of the proposed architecture for storage over an IP network relies on the emergence of 10 Gigabit Ethernet that is being developed concurrently with 10 Gigabit Fibre Channel. Bottom line, I think that you'll need 10 Gig Ethernet to make iSCSI applications really effective. Fortunately, FC is already highly effective at 1 Gbit.

sarrif.nujeerallee482651 : Does the panel have a view on the FC-MI (SANmark II)? Will the impact of iSCSI slow down the development w.r.t current standards progress?

Skip_Jones : Not at all. There has been no sign of a slow down in the Fibre Channel standards arena, quite the opposite. The standards efforts within the Fibre Channel arena are focused on tweaking an already hugely successful market. The focus of the IP storage (IPS) standards community is at the very earliest stage of development. In other words, the FC community is busy polishing the doorknobs in the house while the IPS community is designing the plans to build a house. The question is "Won't the accelerating market adoption of Fibre Channel thwart the efforts of the IPS community?"

rkugele880884: Will Fibre Channel switches be replaced by chips? Do you even need a switch in a SoIP environment?

Skip_Jones : These are two good questions. Will switch functionality be required in building a SAN with any plumbing? The answer is sure. The type of plumbing does not impact the need for switch functionality. Over time switch functionality may be distributed in different ways. The question of integration is always a constant in any technology. Where it makes sense (cost, real estate, power, and sometimes performance) integration happens.

kgattine398680: There have been many complaints about lack of FC interoperability. Will IP solve this problem?

David_Deming : In order to get any type of interoperability it takes shipping products and years of testing. FC products have made huge strides in interoperability and are pretty close to "plug & play". Ethernet-based storage devices are still proprietary implementations, and are not interoperable since there are no standards yet to comply to. In the end, even so iSCSI is using IP as a transport, iSCSI products will inevitably have to go through the same interoperability issues that have now been resolved for FC.

vijayv238138: What are the drawbacks of Fibre Channel compared to iSCSI?

David_Deming : Maybe the question should be what are the drawbacks of iSCSI.

carol.ann.wallace238534 : What is the expected timeframe for products to hit the market for 2Gb FC?

David_Deming : There are numerous vendors that offer 2Gb FC products today. In fact you can obtain numerous 2Gb solutions from some OEM suppliers. Today you can purchase every FC infrastructure component that can operate at 2Gb/s.

kconnor47554: Which technology offers superior performance: FC or IP?

Skip_Jones : The proposed solutions for storage over Ethernet rely on a form of SCSI running on top of IP, running on top of Ethernet. All have more layers and are more complex than Fibre Channel. We think that at the same wire speed, FC will always have superior performance to Ethernet.

vijayv238138: Is it possible for Fibre Channel and iSCSI to coexist?

Skip_Jones : Absolutely. I see iSCSI adoption as more of a NAS play than SAN play. iSCSI is where FC was in late 1997/early 1998. It will demand brand new behavior and designs that must also mesh with a legacy market that requires a different set of cost, performance, features, and functions. iSCSI will have to invest in new equipment regardless of how they architect a SAN. IP SCSI is a whole new animal, using old tools to fix a new problem while still not affecting the old problem. History dictates that it is even more difficult for a new I/O technology to mature if it is also taxed with maintaining legacy applications cost and performance points. It becomes quite a juggling act. The notion that, somehow, a $20 NIC card is going to solve the problems of today's SANs that are already being solved with Fibre Channel probably stems from wishful thinking, exaggerated promotion, and a bit of ignorance in the field of I/O ASIC and firmware technology.

sarrif.nujeerallee482651 : What is the panel's feeling w.r.t the introduction of iSCSI within an existing Fibre Channel environment...

David_Deming : I imagine that numerous iSCSI implementations will provide routers that will take a Fibre Channel SAN and then use iSCSI (which in reality is the IP infrastructure) to extend SAN across long haul distances. Keep in mind that there are numerous vendors that already extend FC SANs across WAN distances.

vijayv238138: Can you give a brief introduction to Fibre Channel?

David_Deming : There is not enough room in this forum to address this. You should visit the FCIA web site for possible providers of FC education solutions.

mmovalli336961: How will IP SCSI handle the performance issues of TC/IP, specifically flow-control handling, such that there is sufficient performance for storage.

Skip_Jones : Good question. That is a big challenge and there are as many opinions about how to solve it as there are people trying to solve it. I suppose it will take a few years to see.

hpike30125: Aren't vendors like Adaptec going to move a lot of the overhead into chips thus making iSCSI as fast as FC?

David_Deming : Any vendor that chooses to implement iSCSI will probably have to implement the TCP/IP protocol stack in hardware. Will it be as fast as FC? That question will remain to be seen once products are available to test in a real production environment.

mmovalli336961: Does that imply that there may be a lot of wasted bandwidth without redesign IP?

Skip_Jones : Yes.

david.mccormick968216 : Assuming iSCSI was readily available today with all its promised features and functionality and I was implementing a SAN, why would I choose FC over iSCSI?

David_Deming : On the surface it may seem like it will be easier to implement IP SANs. The IT community is well versed in IP technology while Fibre Channel knowledge is less common. This alone however is not enough to convince me to recommend that companies wait for IP based storage.

jdoyle2193203: Will the eventual majority acceptance of one protocol over another help or hurt the speed at which technology benefits?

Skip_Jones : Fibre Channel already has about every conceivable protocol mapped. For any transport technology, certainly the more protocols that a manufacturer has to support the more resources and/or time required.

me68326: How do you think InfiniBand will affect FC?

David_Deming : In the near future it will have little affect. In fact, IB will make it easier to cluster processors and make the storage easier to access through IB to FC routers. Remember that IB was originally designed for processor interconnects and the replacement to the PCI bus. Therefore, in order to dramatically impact FC, IB will have to gain market acceptance and have numerous vendors/products available to impact FC in any manner.

potter513323: Has FC superiority for remote mirroring of large data storage subsystems?

Skip_Jones : Absolutely. Mirroring is one of the earliest applications that Fibre Channel solved. Today there are numerous products that pipe Fibre Channel over existing MAN and WAN infrastructures for any degree of "remoteness."

gnader91394: What is the primary difference between iSCSI and IP SCSI?

David_Deming : Maybe the question might be, will iSCSI become the dominant choice for transporting storage over IP.

bjjones89929: What type of growth do you see for Fiber Channel within the next two years. (Both Bandwidth and Speed.)

Skip_Jones : 2Gig now, 10Gig 2002/3, (4 Gig FC disks in the box 2002/3), and 100Gig by 2006. Bandwidth for any given speed will continue to improve by virtue of increased functionality handled by silicon and the developing new standards in Fibre Channel that are again "tweaking" its already proven performance for storage networks.

lu_robert668964 : Is there a lot of IP over FC implementations being used?

David_Deming : Again, maybe the question should be, are there many FC over IP implementations? There are at least three companies that extend FC SANs over IP.

ehurley614122: Is iSCSI ready for deployment?

Skip_Jones : One hurdle that all new IO technologies must clear before getting to the track is to prove a clear and decisive, compelling advantage over an incumbent technology within that incumbent's already owned market space. More times than not, this approach for introducing new IO technologies do much to confuse and hype the industry, succeeding in gathering much airtime. However, in the end, history shows that the ones that actually make it out of the egg and nest do so by addressing some nuance or adjacent area of need sparked or already addressed by the incumbent. Understanding how "trickle-down" technology typically flows, SANs can be easily seen to expand down into areas that have different requirements that may be a good fit for Ethernet. We have already seen this development with FC over IP being deployed today. Once seen as a threat to Fibre Channel, it is clear now that these approaches are very synergistic and complementary.

thereinsels937354: Does the adoption of iSCSI have any affect on the hard disk drive interface, whether it be ATA, SCSI, or FC?

David_Deming : This is a great question. If you are a disk drive manufacturer and you take into consideration the fact that you are building close to five different versions of SCSI drives and a couple of versions of ATA drives. And, if you are into FC at least one version of an FC drive, it will require a huge market to retool for iSCSI. The other concern might come from the enclosure manufacturers and whether they want to retool their backplanes. There were some detractors that said SCSI would die with the advent of FC. This, however, has only fueled the SCSI disk market because major OEMs want affordable industry standard components and they already know how to implement SCSI storage.

potter513323: What does "plumbing" mean in your earlier answer about building a SAN?

David_Deming : Plumbing - the interconnecting infrastructure. This is the hardware and protocol that is used to transmit information between users and storage.

mcohodas467623: Since IT Professionals are generally more familiar with IP than FC, doesn't that mean that iSCSI will grow faster than FC?

Skip_Jones : There seems to be an implicit assumption that IP network professionals are equally versed in storage architectures. That's typically not the case. In large Enterprises, storage is managed by IT professionals who are versed in storage and storage protocols. FC is a natural extension to their knowledge of SCSI, and deploying Fibre Channel based SANs is therefore not that big of a challenge as the IP community would like you to think.