On Tuesday, December 31, 2019 at 7:49:49 AM UTC-8, Bobby wrote: > > Hi all, > > I have come across this research paper (attached) called "*Design and > implementation of IP-based iSCSI Offoad Engine on an FPGA*" and the > authors have mentioned they have used open source software based > *Open-iSCSI* for their research. At the moment there are 2 questions > based on this paper. > > *Question 1:* > On page 3 and under section 2.4 ( *Performance Analysis of Open-iSCSI*), > they have started the paragraph with following lines: > > "*We analyzed iSCSI traffic with Wireshark, the open source network > packet analyzer. We measured traffic between a software initiator and a > target by using a set of microbenchmarks. The microbenchmarks transmitted > arbitrary number of data in both directions* " > > The question is...what are these microbenchmarks. There is no reference to > these microbenchmarks in this paper. Any idea, what are these > microbenchmarks? >
I have no idea. They didn't consult me when doing this paper. :) > > *Question 2:* > Similarly, on the same page 3 and under section 2.3 (Related Work), they > have written "*The most common software implementations in the research > community are open source Open-iSCSI and UNH-iSCSI projects*". > > After my research on UNH-iSCSI, I have found some work where some > researchers have proposed a hardware accelerator for data transfer iSCSI > functions. They analyzed UNH-iSCSI source code and presented a general > methodology that transforms the software C code into the hardware HDL > (FPGA) implementation. Hence their hardware accelerator is designed with > direct C-to-HDL translation of specific sub-modules of UNH-iSCSI software. > > The question: Is there any similar work like this for Open-iSCSI where > specific sub-modules of Open-iSCSI are translated to a hardware language > like Verilog/VHDL on hardware (FPGA)? If not, can you please give a hint > what would possibly a starting point in case of Open-iSCSI? Because the > attached paper does not mention the specific functions of Open-iSCSI code > that could be translated to HDL. > No, none that I know of. There are really two major chunks of open-iscsi: user-land and kernel driver(s). The user-land is only used for error handling, setting up connections, tearing them down, and other administrative tasks (like directing discovery). The kernel code is where all the IO goes on. There are several adapters available for Linux that move the iSCSI stack into hardware. See the qedi driver, for example. These effectively act as the "transport" for open-iscsi, when available. I'd be interested in comparing throughput using these available adapters to the FPGA in the paper -- if I had infinite time. :) > > Thanks ! > > -- You received this message because you are subscribed to the Google Groups "open-iscsi" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/open-iscsi/e6557758-3d4e-4f71-8374-c308e2f09835%40googlegroups.com.
