AF:
NF:0
PS:10
SRH:1
SFN:
DSR:
MID:<20080501134714.0b9215fc@ripper.onstor.net>
CFG:
PT:0
S:andy.sharp@onstor.com
RQ:
SSV:onstor-exch02.onstor.net
NSV:
SSH:
R:<mtracy@css.glasshouse.com>,<dl-cstech@onstor.com>
MAID:1
X-Sylpheed-Privacy-System:
X-Sylpheed-Sign:0
SCF:#mh/Mailbox/sent
RMID:#imap/andys@onstor.net@onstor-exch02.onstor.net/INBOX	0	026601c8abc7$a3627740$654d7e0a@glasshousetech.com
X-Sylpheed-End-Special-Headers: 1
Date: Thu, 1 May 2008 13:48:52 -0700
From: Andrew Sharp <andy.sharp@onstor.com>
To: "Michael Tracy" <mtracy@css.glasshouse.com>
Cc: "DL-CStech" <dl-cstech@onstor.com>
Subject: Re: [Trans Union] 00007968: "Performance Concerns - nfs" - New
 comment to Case
Message-ID: <20080501134852.6021e330@ripper.onstor.net>
In-Reply-To: <026601c8abc7$a3627740$654d7e0a@glasshousetech.com>
References: <017801c8abb6$7db153b0$654d7e0a@glasshousetech.com>
	<026601c8abc7$a3627740$654d7e0a@glasshousetech.com>
Organization: Onstor
X-Mailer: Sylpheed-Claws 2.6.0 (GTK+ 2.8.20; x86_64-pc-linux-gnu)
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

Eric's answers are probably better than mine, but I'll take a shot at
some of these anyway, but don't take my answers to be official policy or
executive decisions or absolute fact :^)

As for the throughput, our boxes are quite capable of maxing out the
hardware, ie., sustaining 440MB/sec if you aggregate all 4 ports, I've
been told.  Always remember, however, that the data source, the storage
will represent some sort of limit as well, so that needs to be
understood in terms of how it is limiting the customer's throughput.  I
would expect that for sequential read of a large file by two clients on
the described setup, we would have no problem running at the max which
would be somewhere around 220+MB/s, but again, only if the array can
serve up the data fast enough.

I would say the calculation that 118.86M (MiB/s I assume) translates to
980 Mb/s is backwards: it translates to more like 1069.74 Mb/s, because
to transfer a whole byte across ethernet includes 1-2 stop/ECC bits per
byte down at the hardware layers.  Not that it really matters.

No we can't run their software on our box.  We don't run a general
purpose operating system so the software wouldn't work.

Since the Bobcat has 5 CPUs besides the management CPU, understanding
the CPU and memory usage data for someone who doesn't have intimate
experience and knowledge of our hardware and software architecture is
going to be pretty doubtful.

It's my understanding that IOZONE is essentially a random read/write
test, so the storage is going to be the bottle neck here.  Although
perhaps there are different ways to run it.  If the customer would like
to supply the parameters under which they ran the tests....


On Thu, 1 May 2008 16:12:10 -0400 "Michael Tracy"
<mtracy@css.glasshouse.com> wrote:

> OK, round 2
> customer has gone the next level and now asks
> 
> --------------------------------------------------------------------------------
> 
> I have some questions:
> 
> - what is the actual network bandwidth that the Onstor can support,
> and how can we ensure that we are reaching that speed?
> - how do we check network flow control to see if it is enabled or
> disabled
> - do you have any sort of tools that will provide access to the CPU
> and memory usage via SNMP or the root console?
> 
> Next, we have some people in our team that worked at Intel helping
> develop NIC drivers and such.  They are asking to run some tests that
> require some binaries to be run on the Onstor, specifically, "nttcp"
> (http://sd.wareonearth.com/~phil/net/ttcp/) and "netperf"
> (http://www.freebsd.org/projects/netperf/)
> 
> Both will require the source to be compiled so that we can run them at
> the root console on the Onstor.  
> 
> 
> --------------------------------------------------------------------------------
> 
> I don't think we support what they want to do.  Any feedback?
> 
> Michael
>   ----- Original Message ----- 
>   From: Michael Tracy 
>   To: DL-CStech 
>   Sent: Thursday, May 01, 2008 2:09 PM
>   Subject: Fw: [Trans Union] 00007968: "Performance Concerns - nfs" -
> New comment to Case
> 
> 
>   Hi all 
>   Customer asking those fun performance questions
>   Have to say, what he describes does sound slow.
> 
>   Any suggestions? 
> 
>   Michael
> 
>   ----- Original Message ----- 
>   From: "Chris Nigh" <chris.nigh@Qsent.com>
>   To: "Call Center TSE" <support@cc.onstor.com>;
> <clnigh@transunion.com> Sent: Thursday, May 01, 2008 2:01 PM
>   Subject: RE: [Trans Union] 00007968: "Performance Concerns - nfs" -
> New comment to Case
> 
> 
>   I have run some preliminary tests against the Onstor/3Par and here
> is what I have noticed.
> 
>   - we have a single FC connection to the Onstor.  Should be 2Gb
>   - our clients are plugged into a Cisco 6506 core switch via GIG
> ports. Verified Full duplex, 1000Mb
>   - the Onstor is plugged into the Cisco 6506 core switch via DUAL
>   etherchanneled gig ports.
>   - both Onstor and clients are on same network, no router involved.
> 
>   When I hit the Onstor with "iozone" from 14 concurrent clients,
> when I watch the "vsvr stats agg" output, I never see anything higher
> than "118.86M".  This is roughly 950Mb network speed.  
> 
>   QUESTION.  This seems low.  Since we have dual Gig Ethernet ports,
> and a 2g FB port, shouldn't we see higher than that?
> 
>   Chris Nigh, CIS Production Team
>   (W) (503) 372-7062
>   (C) (503) 781-6377 
