Re: Itanium Madison blasts Sun, IBM in encryption specs!
From: Rick Jones (foo_at_bar.baz.invalid)
Date: 05/05/03
- Next message: Paul Sture: "Re: The first spam ever came from DEC, apparently"
- Previous message: briggs_at_encompasserve.org: "Re: DCL coding (was: Re: DCL routine required.)"
- In reply to: Andrew Harrison SUNUK Consultancy: "Re: Itanium Madison blasts Sun, IBM in encryption specs!"
- Next in thread: Bob Ceculski: "Re: Itanium Madison blasts Sun, IBM in encryption specs!"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Date: Mon, 05 May 2003 17:11:39 GMT
Andrew Harrison SUNUK Consultancy <Andrew_No.Harrison_No@nospamn.sun.com> wrote:
> In the circumsatances pointing out that the V210 is a less
> expandible version of the V240 is a very minor misdemeanour.
I guess that depends on how much smaller a SPECweb99_SSL result it
achieves :) Still, I'm not going to lose any sleep over it.
>> Now that does not include a credit for the 2GB of RAM that needs to
>> be junked from the base config, assume there is one and give it
>> that credit ($1590) and that seems to suggest $12,580, not 8K
>> dollars.
> Sorry forgot about the 8 GB, so the system config costs 10,680
> or 21360 for 2.
So we are still apart by about 20%.
> The standard 4 GB system ships with 4 1GB DIMMS, we just add another
> 2 x 2 x 1GB DIMMS.
What is the public URL to the standard 4GB system config you are
quoting?
The Price and Buy link for the "large" V240 config from which I
started is a 2 GB config with 4x512 MB DIMMs. It listed no 4x1GB DIMM
config.
> Hum, my suspicion is that nobody has bothered because it is a
> horizontally scaled workload.
One thing I have learned (though perhaps have not always practised) is
to spell assume "ass-u-me."
I had a hand in the specification of the SPECweb99_SSL benchmark and
some familiarity with it. There is one particular aspect of the
dynamic API workload that requires a bit of work to keep the state
consistent. If you are curious, you can lookup the "post log" stuff
from the workload. There are internal consistency requirements for the
post log that require a clustered result to be more than just a bunch
of systems joined with wet string. Still there is nothing in the
workload that absolutely precludes clustering it if a vendor is so
motivated and able.
> And if it isn't then it isn't terribly indicative of real world
> because I have a number of customers using farms of Web servers with
> cryptos which do horizontally scale.
Yet someone in Sun marketing seems to think SPECweb99_SSL is at least
a worthwhile benchmark to publish, indicative or not, or we would not
have those tepid V240 and Netra20/280R SPECweb99_SSL results to
discuss here :)
The next generation of SPEC web benchmark is under development. If
you would like to contribute your thoughts, and do not already have
the contacts, I can send email to Sun's representatives to SPEC for
the benchmark suggesting they get in touch with you.
> Humm, how many customers do you have using Zeus, most will
> be using Apache, SunOne or the IBM WebSphere web server.
If the Netcraft surveys are any indication, many more folks are using
Zeus than Sun ONE. The April 2003 survey at:
http://news.netcraft.com/archives/webserver_survey.html
suggests that there many more sites running Zeus than Sun ONE.
Netcraft does not break-out an historical curve for Zeus. The current
historical curve for Sun ONE (nee Netscape) is there and shows its
considerable long-term decline.
> Nice to do benchmarks with but it hardly figures in the Web server
> stakes.
Seems to figure rather more than Sun ONE :) Both are dwarfed by Apache
and Microsoft.
> Sun does supply Apache supported with Solaris
As does HP with HP-UX. (I do not know what comes with VMS)
> and the Crypto cards also accelerate Apache as well as does the SNCA
If the SNCA is worth its salt, you aught be able to put just about any
web server behind it and see decent results since the SNCA would be
doing virtually all the work.
You would think that using crypto cards rated for thousands of
simultaneous SSL transactions per second would yield really high
SPECweb99_SSL performance. Yet that does not seem to be the case from
the results published thusfar. It shows there are still other things
that the host processor needs to do and a weak host processor cannot
hide behind a crypto accelerator the same way a web server can be
hidden behind an in-kernel accelerator.
> But since you seem to want nit pick a complete costing on the
> HP will need to add VxFS to the HP config price as well as the
> Zeus licenses.
There is a partitioning of VxFS in HP-UX. The base VxFS is included
with the OS. Certain "advanced" features are addon. None of the
advanced features appear to have been used and so the add-on Online
VxFS stuff would not be required. Also, while the OS used VxFS
filesystems (they are the default), the file_set and web server logs
happened to be on HFS filesystems. The only filesystems that see any
activity during a SPECweb99* run are those for the file set and the
web server logs.
The V240 results do not say anything about filesystem types.
Presumeably they were whatever are default for Solaris 9 4/03. The
SPEC reporting rules only require disclosure when something isn't
default.
rick jones
--
denial, anger, bargaining, depression, acceptance, rebirth...
where do you want to be today?
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to raj in cup.hp.com but NOT BOTH...
- Next message: Paul Sture: "Re: The first spam ever came from DEC, apparently"
- Previous message: briggs_at_encompasserve.org: "Re: DCL coding (was: Re: DCL routine required.)"
- In reply to: Andrew Harrison SUNUK Consultancy: "Re: Itanium Madison blasts Sun, IBM in encryption specs!"
- Next in thread: Bob Ceculski: "Re: Itanium Madison blasts Sun, IBM in encryption specs!"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|
