[dm-devel] [Multipath] Round-robin performance limit

Wed Oct 5 03:07:06 UTC 2011

On Tue, 2011-10-04 at 16:19 -0400, Adam Chasen wrote:
> Unfortunately even with playing around with various settings, queues,
> and other techniques, I was never able to exceed the bandwidth of more
> than one of the Ethernet links when accessing a single multipathed
> LUN.
> 
> When communicating with two different multipathed LUNs, which present
> as two different multipath devices, I can saturate two links, but it
> is still a one to one ratio of multipath devices to link saturation.
> 
> After further research on multipathing, it appears people are using md
> raid to achieve multipathed devices. My initial testing of using raid0
> md-raid device produces the behavior I expect of multipathed devices.
> I can easily saturate both links during read operations.
> 
> I feel using md-raid is a less elegant solution than using
> dm-multipath, but it will have to suffice until someone can provide me
> some additional guidance.
> 
> Thanks,
> Adam
We recently changed from the RAID0 approach to multipath multibus.
RAID0 did seem to give more even performance over a variety of IO
patterns but it had a critical flaw.  We could not use the snapshot
capabilities of the SAN because we could never be certain of
snapshotting the RAID0 disks in a transactionally consistent state.  If
I have four disk in a RAID0 array and snapshot them all, how can I be
assured that I have not done something like written two of three stripes
and no parity.  This was our singular reason for discarding RAID0 over
iSCSI for multipath multibus - John

> 
> On Mon, Oct 3, 2011 at 11:08 PM, Adam Chasen <adam at chasen.name> wrote:
> > Malahal,
> > After your mentioning bio vs request based I attempted to determine if
> > my kernel contains the request based mpath. It seems in 2.6.31 all
> > mpath was switched to request based. I have a kernel 2.6.31+ (actually
> > .35 and .38), so I believe I have requrest-based mpath.
> >
> > All,
> > There also appears to be a new multipath configuration option
> > documented in the RHEL 6 beta documentation:
> > rr_min_io_rq    Specifies the number of I/O requests to route to a path
> > before switching to the next path in the current path group, using
> > request-based device-mapper-multipath. This setting should be used on
> > systems running current kernels. On systems running kernels older than
> > 2.6.31, use rr_min_io. The default value is 1.
> >
> > http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/6-Beta/html/DM_Multipath/config_file_multipath.html
> >
> > I have not tested using this setting vs rr_min_io yet or even if my
> > system supports the configuration directive.
> >
> > If I trust some of the claims of several VMware ESX iscsi multipath
> > setups, it is possible (possibly using different software) to gain a
> > multiplicative throughput by adding additional Ethernet links. This
> > makes me hopeful that we can do this with open-iscsi and dm-mulitpath
> > as well.
> >
> > It could be something obvious I am missing, but it appears a lot of
> > people experience this same issue.
> >
> > Thanks,
> > Adam
> >
> > On Tue, May 3, 2011 at 6:12 AM, John A. Sullivan III
> > <jsullivan at opensourcedevel.com> wrote:
> >> On Mon, 2011-05-02 at 22:04 -0700, Malahal Naineni wrote:
> >>> John A. Sullivan III [jsullivan at opensourcedevel.com] wrote:
> >>> > I'm also very curious about your findings on rr_min_io.  I cannot find
> >>> > my benchmarks but we tested various settings heavily.  I do not recall
> >>> > if we saw more even scaling with 10 or 100.  I remember being surprised
> >>> > that performance with it set to 1 was poor.  I would have thought that,
> >>> > in a bonded environment, changing paths per iSCSI command would give
> >>> > optimal performance.  Can anyone explain why it does not?
> >>>
> >>> rr_min_io of 1 will give poor performance if your multipath kernel
> >>> module doesn't support request based multipath. In those BIO based
> >>> multipath, multipath receives 4KB requests. Such requests can't be
> >>> coalesced if they are sent on different paths.
> >> <snip>
> >> Ah, that makes perfect sense and why 3 seems to be the magic number in
> >> Linux (4000 / 1460 (or whatever IP payload is)).  Does that change with
> >> Jumbo frames? In fact, how would that be optimized in Linux?
> >>
> >> 9KB seems to be a reasonable common jumbo frame value for various
> >> vendors and that should contain two pages but, I would guess, Linux
> >> can't utilize it as each block must be independently acknowledged. Is
> >> that correct? Thus a frame size of a little over 4KB would be optimal
> >> for Linux?
> >>
> >> Would that mean that rr_min_io of 1 would become optimal? However, if
> >> each block needs to be acknowledged before the next is sent, I would
> >> think we are still latency bound, i.e., even if I can send four requests
> >> down four separate paths, I cannot send the second until the first has
> >> been acknowledged and since I can easily place four packets on the same
> >> path within the latency period of four packets, multibus gives me
> >> absolutely no performance advantage for a single iSCSI stream and only
> >> proves useful as I start multiplexing multiple iSCSI streams.
> >>
> >> Is that analysis correct? If so, what constitutes a separate iSCSI
> >> stream? Are two separate file requests from the same file systems to the
> >> same iSCSI device considered two iSCSI streams and thus can be
> >> multiplexed and benefit from multipath or are they considered all part
> >> of the same iSCSI stream? If they are considered one, do they become two
> >> if they reside on different partitions and thus different file systems?
> >> If not, then do we only see multibus performance gains between a single
> >> file system host and a single iSCSI host when we use virtualization each
> >> with their own iSCSI connection (as opposed to using iSCSI connections
> >> in the underlying host and exposing them to the virtual machines as
> >> local storage)?
> >>
> >> I hope I'm not hijacking this thread and realize I've asked some
> >> convoluted questions but optimizing multibus through bonded links for
> >> single large hosts is still a bit of a mystery to me.  Thanks - John
> >>
> >> --
> >> dm-devel mailing list
> >> dm-devel at redhat.com
> >> https://www.redhat.com/mailman/listinfo/dm-devel
> >>
> >
> 
> --
> dm-devel mailing list
> dm-devel at redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel