QuicksearchCodenews SearchDisclaimerThe individual owning this blog works for Oracle in Germany. The opinions expressed here are his own, are not necessarily reviewed in advance by anyone but the individual author, and neither Oracle nor any other party necessarily agrees with them.
NavigationCategories
|
![]() Do-it-yourself X4500 ?Wednesday, July 19. 2006Comments
Display comments as
(Linear | Threaded)
4500 looks lice, but I think you also have to consider what happens when you have the disks hung off couple of PCIe cards, and whether, actually, many customers who want the capacity will also care so much about the peak performance.
Real question is if home-made 4-CPU box with 2-core 885 Opterons and 4 GbE and same 48 drives connected to 3-4 PCI-X controllers would not be both significantly (like 2x) cheaper AND faster. Power consumption would be higher only by about 20-25%
Okay, you can´t compare a 2 CPU to an 4 CPU-BOX, but this is only a side note.
But: 1. 20-25% can annihilate any cost advantage of a 3 year operation.. 2. I don´t think it would be faster at I/O as general purpose boards have not the connectivity to do such stuff. You have to use an mainboard with 4 independent PCIX-busses and you have to be sure, that the onboard cpu and cache poses no bottleneck. Do you know a board with 4 independent busses ?
Don't get me wrong, I think very highly of Thumper from a technical point of view. Price is another matter.
1. Extra power consumption + cooling is about 400-500W, which results in extra 10-12 KWh/day - cost about $2 or less; so in 3 years it's about $2,000 extra or so, far smaller than $15-25,000 price difference we're talking about 2. 2 PCI-X busses should be able to provide about 1.5GB/s (2 PCI-X 100 controllers on each bus connected to 12 drives each - 2x2x12 ). Are you sure Thumper can actually provide more than that? and where this data will be going if there are only 2 GbE links?
We were able to get out 2.5 GBytes per second in peak. of of this system. So 1.5 Gigabytes per Second would not sufficient. And this was done with 60 MByte/s-Disks. Now imagine perpendicular recording discs, that have a even higher bandwidth. With 100 MByte/s or 120 MByte/s per disk the problem gets even worse. The Thumper has the headroom for such disks. A PCIX solution with would is already maxed ... with the slow discs.
What kind of program/command were you running that was reading (or writing) data FROM DISKS at 2.5GB/s? Are you sure this data was NOT cached? 2.5GB/s from cache seems plausible, from disks - almost too good to be true. But again, I've never seen Thumper "in real life" and you apparently did.
Several mkfiles in parallel. To be honest, it were 2.1 to 2.2 average.
Oh yeah, i have one at the office. Would be nice at home for video and mp3
Tyan's got a motherboard that would gets you close as far as I can tell.
ftp://ftp.tyan.com/manuals/m_s3892_110.pdf Sorry that file is huge, but it has the architecture picture in it. If you squint hard at the grainy image it looks like you get 2 x16 PCIe channels and 2 PCI-X 133 channels. Although that probably saturates the HyperTransport link to the processors. Of course, your 2.2GB/s sounds like HT's saturation point anyway.
Disregard my HT saturation comment on the X4500, obviously I didn't squint at the architecture pic you conveniently included.
Also, the fact that Sun has uplinked some of the PCI-X I/O to CPU1 is sweet. Unlike the Tyan that routes everything through CPU0. Now if I can just convince the wife to tolerate the noise and keep driving the junker for another 4-6 years maybe I can get some of that classy Sun kit. Anyway, that's the closest I've seen 'commodity' get, and obviously Sun is out ahead quite a ways.
SuperMicro has a very nice 4-way board with 2 PCI-X busses. It's using the same AMD8132 chipset as the Thumper. Also the same Intel NICs. This would make a nice 'lil' Thumper if you put it in a 24-bay chassis. It won't have some of the Thumper's capabilities, but at
|
Links in this articleThe LKSF bookThe book with the consolidated Less known Solaris Tutorials is available for download here
Web 2.0Contact
Networking xing.com My photos SyndicationTagged articlesAMD Apple avs Bahn Blogging Blogosphere braindump Business Travel CeBIT cec cec2006 CMT del.icio.us deutsch dtrace fliegen Fundsache General Hamburg IBM i hate sundays Intel iscsi jumpstart Links Linux lksf Mindfuck Movies Music Musik Niagara Opensolaris Opteron Photographie policy of ... Politik Security Solaris storage Sun suncec2007 sunw t1 The IT Business Ultrasparc ultrasparc t1 Wirtschaft Work ZFS
CommentsThu, 09.09.2010 13:04
Okay, I must have overseen tha
t
Thu, 09.09.2010 12:59
1. Gerne:
zb. für ne SAN Migr
ation. Ich weiss das Sun das G
efühl hat, dass man sowas nich
t braucht.
Das ist ähn [...]
Thu, 09.09.2010 12:00
Believe it or not ... there ar
e company obeying the licenses
. So that's a very practical c
hange ....
Thu, 09.09.2010 11:54
So practically, there is zero
difference:
90 day evaluati
on period which wouldn't expir
e anyway, vs. a "perpetu [...]
Thu, 09.09.2010 11:49
There is no timelimit ...
Buttons![]() This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Germany License
![]() ![]() ![]() Blog Administration |
The biggest knock against the X4500 I’ve heard is that it is too expensive. From a storage perspective it is actually absurdly cheap compared to the 5-10x charged by the name-brand storage vendors - and for lower data integrity than the X4500/ZFS...
Tracked: Jul 20, 07:50
Nun, dritter Tag ... die letzten Breakout-Sessions ... beginnen wir mal den Tag mit "Building multipetabyte architectures with Thumper". Nicht wirklich viel los, die meisten Leute sind wohl irgendwie noch in einem Zustand der frühmorgendlichen Pseudostas
Tracked: Oct 04, 18:17