NetBSD-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: iSCSI: NetBSD-9 target, Linux initiator



On Fri, 21 Feb 2020 at 19:29, Chavdar Ivanov <ci4ic4%gmail.com@localhost> wrote:
>
> On Fri, 21 Feb 2020 at 18:46, BERTRAND Joel <joel.bertrand%systella.fr@localhost> wrote:
> >
> > Michael van Elst a écrit :
> > > joel.bertrand%systella.fr@localhost (BERTRAND Joel) writes:
> > >
> > >> If I understand... Error is triggered by test
> > >> /usr/src/netbsd-9/src/external/bsd/iscsi/lib/../dist/src/lib/disk.c line
> > >> 1405 when initiator sends data greater than 1MB.
> > >
> > >> Why this limitation ?
> > >
> > > Apparently to avoid allocating an arbitrary sized buffer.
> >
> >         Yes, but why not allocate for example a 4MB buffer ?
> >
> > > In my experience the iscsi target (and the userland iscsi initiator)
> > > are pretty limited and somewhat broken. I suggest you use the
> > > iscsi target from pkgsrc (net/istgt). Much better, faster and more
> > > compatible.
>
> Years ago, perhaps around NetBSD v.5 or 6, I tried the built-in iSCSI
> target and found the same - it was, shall we say, rather flaky. I then
> started using net/istgt, it worked for me as well.
>
> Lately, since I enabled it on one of my systems running -current of
> the day with 2-3 updates a week, I've found the built-in target to be
> reasonably ok, at least with initiators running on various Windows
> versions; I haven't had any reason to try it with other initiators so
> far. As I mentioned before, my extents are either /dev/rdkxx from a
> GPT disk, or a ZFS zvols. I do get
> src/external/bsd/iscsi/lib/../dist/src/lib/disk.c:1364: ***ERROR***
> UNKNOWN OPCODE 0xa1 (or 0xdf or 0x9e), but they don't seem to affect
> the operation. Performancewise, things are not so good - i have a
> FreeNAS 11.3 on the same gigabit switch, from the same Windows 10
> initiator I get 2-4 times better figures (measured by
> CrystalDiskMark), even if the FreeNAS server is only a guest on a
> XCP-NG host, whereas the NetBSD machine is physicall; both with
> gigabit interfaces.
>
> I should try again net/istgt just to check the performance figures.
>
>
> >
> >         Same configuration files or not ?
>
> Definitely not...


And here we are with the numbers... Michael is of course right about
the performance.
NetBSD-current, amd64 from yesterday, built-in iscsi-target:
......
-----------------------------------------------------------------------
CrystalDiskMark 6.0.2 x64 (C) 2007-2018 hiyohiyo
                          Crystal Dew World : https://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

   Sequential Read (Q= 32,T= 1) :     9.148 MB/s
  Sequential Write (Q= 32,T= 1) :     3.197 MB/s
  Random Read 4KiB (Q=  8,T= 8) :     0.708 MB/s [    172.9 IOPS]
 Random Write 4KiB (Q=  8,T= 8) :     0.241 MB/s [     58.8 IOPS]
  Random Read 4KiB (Q= 32,T= 1) :     0.683 MB/s [    166.7 IOPS]
 Random Write 4KiB (Q= 32,T= 1) :     0.307 MB/s [     75.0 IOPS]
  Random Read 4KiB (Q=  1,T= 1) :     0.700 MB/s [    170.9 IOPS]
 Random Write 4KiB (Q=  1,T= 1) :     0.284 MB/s [     69.3 IOPS]

  Test : 50 MiB [G: 67.3% (20.2/30.0 GiB)] (x1)  [Interval=5 sec]
  Date : 2020/02/21 20:23:43
    OS : Windows 10  [10.0 Build 18363] (x64)
  .....

Same system, same target, using net/istgt:
.......
-----------------------------------------------------------------------
CrystalDiskMark 6.0.2 x64 (C) 2007-2018 hiyohiyo
                          Crystal Dew World : https://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

   Sequential Read (Q= 32,T= 1) :    16.854 MB/s
  Sequential Write (Q= 32,T= 1) :    10.903 MB/s
  Random Read 4KiB (Q=  8,T= 8) :    12.360 MB/s [   3017.6 IOPS]
 Random Write 4KiB (Q=  8,T= 8) :    12.594 MB/s [   3074.7 IOPS]
  Random Read 4KiB (Q= 32,T= 1) :    12.185 MB/s [   2974.9 IOPS]
 Random Write 4KiB (Q= 32,T= 1) :    12.421 MB/s [   3032.5 IOPS]
  Random Read 4KiB (Q=  1,T= 1) :     0.703 MB/s [    171.6 IOPS]
 Random Write 4KiB (Q=  1,T= 1) :     0.571 MB/s [    139.4 IOPS]

  Test : 50 MiB [G: 67.3% (20.2/30.0 GiB)] (x1)  [Interval=5 sec]
  Date : 2020/02/21 20:23:21
    OS : Windows 10  [10.0 Build 18363] (x64)
  .....
FreeNAS 11.3 target:
....
-----------------------------------------------------------------------
CrystalDiskMark 6.0.2 x64 (C) 2007-2018 hiyohiyo
                          Crystal Dew World : https://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

   Sequential Read (Q= 32,T= 1) :    18.639 MB/s
  Sequential Write (Q= 32,T= 1) :    20.656 MB/s
  Random Read 4KiB (Q=  8,T= 8) :     9.985 MB/s [   2437.7 IOPS]
 Random Write 4KiB (Q=  8,T= 8) :    13.202 MB/s [   3223.1 IOPS]
  Random Read 4KiB (Q= 32,T= 1) :    10.803 MB/s [   2637.5 IOPS]
 Random Write 4KiB (Q= 32,T= 1) :    13.384 MB/s [   3267.6 IOPS]
  Random Read 4KiB (Q=  1,T= 1) :     0.727 MB/s [    177.5 IOPS]
 Random Write 4KiB (Q=  1,T= 1) :     0.604 MB/s [    147.5 IOPS]

  Test : 50 MiB [J: 72.8% (218.4/300.0 GiB)] (x1)  [Interval=5 sec]
  Date : 2020/02/21 20:24:05
    OS : Windows 10  [10.0 Build 18363] (x64)
  .....

Obviously one should switch to istgt . . .

>
> >
> >         Best regards,
> >
> >         JKB
>
>
>
> --
> ----



-- 
----


Home | Main Index | Thread Index | Old Index