Basically, it is
sending FC frames over Ethernet. This localizes the traffic unless you route
based on MAC addresses. So you send 2146 bytes of FC frame plus 18 bytes of
Ethernet overhead as FCoE ?standard? packet. 18 bytes of Ethernet gets stripped
and you have straight FC frame that can go through any FC network. Now you can
have 10G Ethernet pipes into existing FC SANs. Limited market potential as far
as I can see. The key argument is it much easier to implement than iSCSI and
also has less overhead and uses all the benefits of FC. End to End credits are
simulated using PAUSE command on Ethernet and MAC addresses are mapped into
WWNs.
Biggest knock is that
it will not route on the ?global? scale like TCP/IP would.
Eddy
-----
Original Message -----
Sent: Wednesday, April 25, 2007 8:09
AM
Subject: Re: FW: Recent comments
about FCoE and iSCSI
"John Hufferd" <jhufferd@xxxxxxxxxxx> wrote on
25/04/2007 02:45:51:
> Julian,
> To be sure you understand
our position; Brocade is pushing iSCSI as an
> outreach protocol from
the Data Center. We also believe iSCSI is very
> useful for
installations that do not have a Fibre Channel
> infrastructure, and in
that case we will be able to sell them our new
> iSCSI and TOE offload
HBAs.
>
> When I say iSCSI is an outreach protocol, this
is a statement that iSCSI
> is very important to connect "stranded"
servers to the Fibre Channel
> Fabric. That is, we sell
iSCSI-to-FC Gateway devices which will permit
> iSCSI Servers (software
or HBA iSCSI initiators) to connect to the
> Enterprise "Bet Your
Business" FC Storage. This of course also applies
> to Desktops
and Laptop systems, and systems at distance.
>
You make it sound like:
- most of the servers in the world have their
storage on the network - and that is not the case
- FCP is basically better performing than iSCSI -
and that is not true either
- Gatewaying is expensive - and it is perhaps so
but only if you are completely relying on FCP storage (and there are plenty
of good iSCSI vendors of storage)and pushing the price on the servers is not
cheap either - at least not for the server
buyer
> Now with that
positioning, it is important to understand the limitation
> to this
strategy. The primary problem is that iSCSI to FC Bridging
>
(Gatewaying) is relatively expensive (compared to simple FC
>
connections). Though we have some of the best priced Gateways on
the
> market, it is not cost feasible to replace all the server
connectivity
> to FC storage with iSCSI for the hundred to thousands of
servers in the
> Data Center. And since, if there is to be a
consolidated Network
> connection to the servers in the Data Center,
there must be an
> evolutionary replacement of Server Connections to
Storage. That means
> there must be a bridge/Gateway approach.
And as I mentioned before,
> there is just too much cost in the
iSCSI to FC Gateway.
>
> The issue is the server requirement to
have a single connection type to
> handle cluster messaging, general
messaging, and storage. iSCSI is
> clearly an option for the
storage, however, the gateway costs are too
> high for iSCSI to be used
as the "normal" server connect into a FC based
> Fabric. That is true
for the current 1GE; and for 10GE the cost is just
> out of sight.
The reason for this is the requirement for TCP/IP
> termination
and re-initiation with FC at the Gateway.
>
> Now with respect to
FC over Ethernet the important thing to understand
> is that it is not
Ethernet as we have known it up to know. The Ethernet
> we are
talking about is a type of Ethernet that can only be deployed in
> a
constrained environment such as a Data Center. This form of
Ethernet
> is called DCE (Data Center Ethernet) or CEE (Convergence
Enhanced
> Ethernet). This form of Ethernet is a Loss-less type
Ethernet, with
> multi-priority and Flow Control. This is NOT an
Internet or Intranet
> type of Ethernet.
>
> FCoE is all
about using the DCE (CEE) to carry FC frames. The rest of
> the
Host and storage stack remain the same, the functions and features
> of
the switches also remain the same and add the capability to provide
>
Cluster Message Switching which has latency close to InfiniBand
speeds.
>
>
> Because the FC frames are transported to the
switches intact via a DCE
> frame, the Bridging, if you want to call it
that, is virtually non
> existent. Hence you can deliver the FC
frames to FC devices, or send FC
> frames to DCE FCoE devices, just like
one would do if it was all FC. And
> all this is done while performing
Cluster message switching and general
> message trucking to the IP
outfacing network.
>
The rosy
future of the yet to appear DCE/CEE and a layer 2 only world.
First you have some terms confused:
Bridging is the term commonly used for Layer-2
switching and routing is therm used for layer-3 (switching).
Bridging has some advantages (less management) that
have created a movement towards an enterprise wide
LAN. But this has a long way to go and will require significant
equipment and protocol changes.
Even its
proponents do not call for transportless networks, lossless networks
etc.
The second trouble with your argument is
that there are no known large scale networking technologies that
really work at full speed (high speed) and are lossless
(flow-controlled) and errorless like FCoE assumes.
The TCP/IP has solved this issue for every generation using the proven
end-to-end principle (and is doing so now).
And it is not by chance so and that is why all networking applications
are built above layer-3 and not dropping layer-3 (like FCoE) does.
Although I can understand the DCE arguments as a
management statement I would prefer like any rational engineer, to base my
building blocks on structures that are proven and long lasting. And those are
still the end-to-end TCP/IP that can accommodate even your FCP addicts. The
IPS TWG has developed the iFCP that does exactly what FCoE claims to do an a
better base.
> This means an
evolutionary process is possible to the solution of
> getting a single
Fabric connection for all networks connected to a
> server, further, the
process has very low interconnection cost on the
> Data Center Fabric.
And it maintains all the FC Fabric Services, and all
> the same Storage
Management processes.
>
> By the way, this is primarily a Server
driven value statement, there
> seems to be little value in having FCoE
on the storage controller.
> Therefore FC storage controllers (and
FICON) will be the very last
> things that connect using FCoE and that
evolution will take at least a
> decade or more.
>
It is server cost statement. It costs
nothing to connect a modern server to ethernet it will cost a bundle to
connect to FCoE and it will force users in short lived bad
solutions.
> We see value in
offering switches and Directors that can support DCE
> switching, FC
switching as well as iSCSI interconnect, and the
> "Trunking" of general
messaging to the Outfacing IP network. That said;
> we do not see
FCoE going beyond the constraints of the Data Center.
Data Centers now grow to tens of thousands of nodes.
There is no layer-2 technology for errorless/lossless operation at this scale
and there is no good reason to pursue one. The only possible reason (good
reason) is the bridging infrastructure but that infrastructure has a
completely different rationale than the flowcontrol.
> This issue and message is quite different from the issues and
messages
> we struggled with when we started iSCSI. There is a
consortium of folks
> both working on the DCE (CEE) and the FCoE.
Without the DCE the FCoE
> will not happen.
>
> None of the above cancels out the value of iSCSI in numerous
>
environments.
>
>
iSCSI is
good for all environments. Business consideration (and some politics) keep it
form "exploding" and large storage vendors are completely indifferent to the
network connection they are using.
You and I
have also slightly different views of DCE. I expect DCE (that still has a way
to go) to improve the QoS in the data-center (and for storage too). You expect
it to bring the loss rates down to the levels that FCP assumes (FCP has no
transport layer) and that is probably a pipe dream. Todays transport solution
for loss mitigation are far more cost effective - and that's why iFCP is a
better proposition as a transition technology than FCoE and iSCSI with
gateways is propably better in the long run.
>
> .
> .
> .
> John L Hufferd
>
Sr. Executive Director of Technology
> jhufferd@xxxxxxxxxxx
>
Office Phone: (408) 333-5244; eFAX: (408) 904-4688
> Alt Office Phone:
(408) 997-6136; Cell: (408) 627-9606
>
> -----Original
Message-----
> From: John Hufferd
> Sent: Tuesday, April 24, 2007
12:57 PM
> To: 'Julian_Satran@xxxxxxxxxx'
> Subject: Re:
Recent comments about FCoE and iSCSI
>
> Julian,
> I think
you are wrong on this one. The arguments are quite different
>
then the ones we had in pre iSCSI days. (By the way I missed you
on
> today's Renato meeting/conf call where Brocade took the IBM
technology
> group through FCoE as it is being placed in our plans).
> I will send you more info when I get to my computer. But
you probably
> were sent the Brocade charts. Please review them
and I will follow up
> with more information.
> This does NOT
replace iSCSI it applies only to a DataCenter envuornment
> with
lossless DCE ethernet.
> --------------------------
> John L.
Hufferd
> Sr. Ex. Director of Technology
> Brocade Communications
Systems, Inc.
> Phone: (408) 333-5244
> Mobile: (408)
627-9606
> eMail: jhufferd@xxxxxxxxxxx
> (Sent from my BlackBerry
Wireless)
>
>
> ----- Original Message -----
>
From: Julian Satran <Julian_Satran@xxxxxxxxxx>
> To: ips@xxxxxxxx
<ips@xxxxxxxx>
> Sent: Tue Apr 24 12:10:29 2007
> Subject:
Recent comments about FCoE and iSCSI
>
>
> Dear All,
>
> The trade press is lately full with comments about the
latest and
> greatest reincarnation of Fiber Channel over ethernet.
> It made me try and summarize all the long and hot debates that
preceded
> the advent of iSCSI.
> Although FCoE proponents make
it look like no debate preceded iSCSI that
> was not so - FCoE was
considered even then and was dropped as a dumb
> idea.
>
>
Here is a summary (as afar as I can remember) of the main arguments.
>
They are not bad arguments even in retrospect and technically FCoE
>
doesn't look better than it did then.
>
> Feel free to use this
material in a nay form. I expect this group to
> seriously expand
my arguments and make them public - in personal or
> collective form.
>
> And do not forget - it is a technical dispute - although we
all must
> have some doubts about the way it is pursued.
>
> Regards,
> Julo
>
>
---------------------------------------------------------------------
>
> What a piece of nostalgia :-)
>
> Around 1997 when a
team at IBM Research (Haifa and Almaden) started
> looking at connecting
storage to servers using the "regular network"
> (the ubiquitous LAN) we
considered many alternatives (another team even
> had a look at ATM -
still a computer network candidate at the time). I
> won't get you over
all of our rationale (and we went over some of them
> again at the end
of 1999 with a team from CISCO before we convened the
> first IETF BOF
in 2000 at Adelaide that resulted in iSCSI and all the
> rest) but some
of the reasons we choose to drop Fiber Channel over raw
> Ethernet where
multiple:
>
>
> * Fiber Channel Protocol (SCSI
over Fiber Channel Link) is
> "mildly" effective because:
>
> * it implements endpoints in a dedicated engine
(Offload)
> * it has no transport layer (recovery
is done at the
> application layer under the assumption that the error
rate will be very
> low)
> * the network is
limited in physical span and logical span
> (number of switches)
> * flow-control/congestion control is achieved
with a
> mechanism adequate for a limited span network (credits). The
packet loss
> rate is almost nil and that allows FCP to avoid using a
transport
> (end-to-end) layer
>
> *
FCP she switches are simple (addresses are local and the
> memory
requirements cam be limited through the credit mechanism)
>
* However FCP endpoints are inherently costlier than
>
simple NICs - the cost argument (initiators are more expensive)
>
* The credit mechanisms is highly unstable for
large
> networks (check switch vendors planning docs for the network
diameter
> limits) - the scaling argument
> *
The assumption of low losses due to errors might
> radically change when
moving from 1 to 10 Gb/s - the scaling argument
> *
Ethernet has no credit mechanism and any mechanism with
> a similar
effect increases the end point cost. Building a transport
> layer in the
protocol stack has always been the preferred choice of the
> networking
community - the community argument
> * The
"performance penalty" of a complete protocol stack
> has always been
overstated (and overrated). Advances in protocol stack
> implementation
and finer tuning of the congestion control mechanisms
> make
conventional TCP/IP performing well even at 10 Gb/s and over.
> Moreover
the multicore processors that become dominant on the computing
> scene
have enough compute cycles available to make any "offloading"
> possible
as a mere code restructuring exercise (see the stack reports
> from
Intel, IBM etc.)
> * Building on a complete stack
makes available a wealth of
> operational and management mechanisms
built over the years by the
> networking community (routing,
provisioning, security, service location
> etc.) - the community
argument
> * Higher level storage access over an IP
network is widely
> available and having both block and file served over
the same connection
> with the same support and management structure is
compelling - the
> community argument
> *
Highly efficient networks are easy to build over IP with
> optimal
(shortest path) routing while Layer 2 networks use bridging and
> are
limited by the logical tree structure that bridges must follow. The
>
effort to combine routers and bridges (rbridges) is promising to
change
> that but it will take some time to finalize (and we don't know
exactly
> how it will operate). Untill then the scale of Layer 2 network
is going
> to seriously limited - the scaling argument
>
>
>
> As a side argument - a performance
comparison made in
> 1998 showed SCSI over TCP (a predecessor of the
later iSCSI) to perform
> better than FCP at 1Gbs for block sizes
typical for OLTP (4-8KB). That
> was what convinced us to take the path
that lead to iSCSI - and we used
> plain vanilla x86 servers with
plain-vanilla NICs and Linux (with
> similar measurements conducted on
Windows).
> The networking and storage community
acknowledged those
> arguments and developed iSCSI and the companion
protocols for service
> discovery, boot etc.
>
> The community also acknowledged the need to
support existing
> infrastructure and extend it in a reasonable fashion
and developed 2
> protocols iFCP (to support hosts with FCP drivers and
IP connections to
> connect to storage by a simple conversion from FCP
to TCP packets) FCPIP
> to extend the reach of FCP through IP (connects
FCP islands through TCP
> links). Both have been
>
implemented and their foundation is solid.
>
> The current attempt of developing a "new-age" FCP
over an
> Ethernet link is going against most of the arguments that have
given us
> iSCSI etc.
>
> It
ignores the networking layering practice, build an
> application
protocol directly above a link and thus limits scaling,
> mandates
elements at the link layer and application layer that make
>
applications more expensive and leaves aside the whole "ecosystem"
that
> accompanies TCP/IP (and not Ethernet).
>
> In some related effort (and at a point also when
developing
> iSCSI) we considered also moving away from SCSI (like some
"no
> standardized" but popular in some circles software did - e.g.,
NBP) but
> decided against. SCSI is a mature and well understood
access
> architecture for block storage and is implemented by many
device
> vendors. Moving away from it would not have been justified at
the time.
>
_______________________________________________
Ips mailing
list
Ips@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ips
|
