[Bioclusters] Workstation Selection for Bioinformatics Research

[Stephen Wicks]

--Apple-Mail-12-840296516
Content-Transfer-Encoding: 7bit
Content-Type: text/plain;
	charset=US-ASCII;
	format=flowed

> Hi Dr. Wicks,
>
> can you elaborate on the price tradeoffs? How many more PCs would you 
> get
> for a number of Macs?

Sure. Although it's hard to recommend exact solutions without knowing a 
little more about your needs. Pricing Apples today is easier than it 
was even a year ago. If I were building a room full of generic machines 
(say 20 or so) that will be used for light rendering and ad hoc 
bioinformatics, I might do something like this:

20 eMacs @$699.00 (actual institutional academic pricing will be a bit 
cheaper than this, in my experience...but this pricing is new, so I 
don't know how flexible they would be right now). This gets you 20 17" 
screens and 20 G4 processors. You'll need 20 512GB sticks of RAM@ about 
$50 each. I would recommend a networked fileserver (single processor 
Xserve comes with unlimited client licenses) and shared disk space 
(Xraid) so that anyone could work seamlessly on any machine. I have 
just bought 4 of the dual 1.33 XServes, and although it would be 
improper to quote our final price, I can tell you that it was 
significantly less than you might expect. I also bought 5 Dell 
Poweredge 1650s for the same heterogenuos compute cluster (which 
incidentally, Bioteam has done an excellent job deploying). The Dells 
are underspeced compared to the Apple machines, and although the Apple 
machines were a bit more expensive than the Dells, after accounting for 
any feature/performance lag on the older Dells I believe that the Apple 
machines were the better deal. I look forward to benchmarking to 
substantiate this WRT performance. Anyway, a well-speced file server 
with 1.26 TB of dual fiber channel attached storage from Apple 
Education is under 10 grand.

That gets you 20 quality workstations with terabyte+ NAS for under 
$25,000 USD. Now, some will say that these eMacs are "way too slow" for 
what you want, and maybe they are. I don't know. But in my experience, 
they are fast enough for my modest molecular visualization needs. I use 
raster3d, and molscript and rasmol for publication quality rendering of 
proteins, and I do that on a portable, so I just don't buy it. It may 
take 2 minutes on a powerbook and 36 sec on a 3.06 GHz PIV HT, but that 
is not a big deal, and there is the whole dual boot OS issue and 
client-side licensing. In the end, workflow and efficiency have a 
*much* bigger impact on time-on-task than processor speed. I think the 
only serious potential drawback to the eMac is the 1GB RAM limitation. 
If the machines are for casual multitasking use, then people may find 
value in the extras that come with the Mac, like the iApps. Then there 
are the intangibles like noise. 20 PCs can be pretty loud. The eMac is 
nearly silent.

If the machines are to be used for heavy rendering, I would do it a 
little differently. I would still use cheaper eMacs, but build out some 
compute nodes on a rack. The idea being here that unless you need all 
20 stations to be rendering simultaneously (which you might for a 
course, in which case then you might want to go a different route again 
and do the rendering locally on each machine--in which case you then 
probably want Xeon-class dual-processor workstations with NAS), you 
will probably gain performance with a render farm approach.

Regards,

Stephen

Dr. Stephen Wicks,
Assistant Professor, Biology.
Boston College.
140 Commonwealth Ave.,
Chestnut Hill, MA. 02467.

Phone: 617 552 6851 (office)
Fax: 617 552 2011
Office: Rm 468 Higgins Hall
e-mail: swicks@bc.edu


On Thursday, April 10, 2003, at 02:31 PM, Duzlevski, Ognen wrote:

>  I am not anti-Mac, just curious to know what is an
> actual benefit in even thinking about Macs in a cluster.
>
> Thank you,
> Ognen
>
> On Thu, 10 Apr 2003, Stephen Wicks wrote:
>
>> Hi Chip,
>>
>> Whenever I see someone considering a Linux/Windows dual boot setup for
>> a single job in biology, I recommend that they consider a Macintosh.
>> OSX will give you fully supported access to all of the command line or
>> X11 apps that you will need (including in your case RasMol, Povray,
>> Molscript, raster3d etc.), and still let your users use
>> Office/Photoshop etc. at the same time on the same machine and same 
>> OS.
>> Networking and interoperability these days is a non-issue. The only
>> caveat right now might be that the G4 towers are fast, but not as fast
>> as the top end Intel-based solution.
>>
>> You wanted options...
>>
>> Stephen

--Apple-Mail-12-840296516
Content-Transfer-Encoding: 7bit
Content-Type: text/enriched;
	charset=US-ASCII

<excerpt><fontfamily><param>Hiragino Kaku Gothic Pro</param>Hi Dr.
Wicks,


can you elaborate on the price tradeoffs? How many more PCs would you
get

for a number of Macs?

</fontfamily></excerpt><fontfamily><param>Hiragino Kaku Gothic Pro</param>

Sure. Although it's hard to recommend exact solutions without knowing
a little more about your needs. Pricing Apples today is easier than it
was even a year ago. If I were building a room full of generic
machines (say 20 or so) that will be used for light rendering and ad
hoc bioinformatics, I might do something like this:


20 eMacs @$699.00 (actual institutional academic pricing will be a bit
cheaper than this, in my experience...but this pricing is new, so I
don't know how flexible they would be right now). This gets you 20 17"
screens and 20 G4 processors. You'll need 20 512GB sticks of RAM@
about $50 each. I would recommend a networked fileserver (single
processor Xserve comes with unlimited client licenses) and shared disk
space (Xraid) so that anyone could work seamlessly on any machine. I
have just bought 4 of the dual 1.33 XServes, and although it would be
improper to quote our final price, I can tell you that it was
significantly less than you might expect. I also bought 5 Dell
Poweredge 1650s for the same heterogenuos compute cluster (which
incidentally, Bioteam has done an excellent job deploying). The Dells
are underspeced compared to the Apple machines, and although the Apple
machines were a bit more expensive than the Dells, after accounting
for any feature/performance lag on the older Dells I believe that the
Apple machines were the better deal. I look forward to benchmarking to
substantiate this WRT performance. Anyway, a well-speced file server
with 1.26 TB of dual fiber channel attached storage from Apple
Education is under 10 grand.


That gets you 20 quality workstations with terabyte+ NAS for under
$25,000 USD. Now, some will say that these eMacs are "way too slow"
for what you want, and maybe they are. I don't know. But in my
experience, they are fast enough for my modest molecular visualization
needs. I use raster3d, and molscript and rasmol for publication
quality rendering of proteins, and I do that on a portable, so I just
don't buy it. It may take 2 minutes on a powerbook and 36 sec on a
3.06 GHz PIV HT, but that is not a big deal, and there is the whole
dual boot OS issue and client-side licensing. In the end, workflow and
efficiency have a *much* bigger impact on time-on-task than processor
speed. I think the only serious potential drawback to the eMac is the
1GB RAM limitation. If the machines are for casual multitasking use,
then people may find value in the extras that come with the Mac, like
the iApps. Then there are the intangibles like noise. 20 PCs can be
pretty loud. The eMac is nearly silent.


If the machines are to be used for heavy rendering, I would do it a
little differently. I would still use cheaper eMacs, but build out
some compute nodes on a rack. The idea being here that unless you need
all 20 stations to be rendering simultaneously (which you might for a
course, in which case then you might want to go a different route
again and do the rendering locally on each machine--in which case you
then probably want Xeon-class dual-processor workstations with NAS),
you will probably gain performance with a render farm approach.


Regards,


Stephen


</fontfamily><fontfamily><param>Helvetica</param>Dr. Stephen Wicks, 

Assistant Professor, Biology.

Boston College.

140 Commonwealth Ave.,

Chestnut Hill, MA. 02467.


Phone: 617 552 6851 (office)

Fax: 617 552 2011

Office: Rm 468 Higgins Hall

e-mail: swicks@bc.edu

</fontfamily><fontfamily><param>Hiragino Kaku Gothic Pro</param>

</fontfamily>

<fontfamily><param>Hiragino Kaku Gothic Pro</param>On</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>Thursday,</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>April</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>10,</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>2003,</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>at</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>02</fontfamily>:<fontfamily><param>Hiragino Kaku Gothic Pro</param>31</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>PM,</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>Duzlevski,</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>Ognen</fontfamily>
<fontfamily><param>Hiragino Kaku Gothic Pro</param>wrote</fontfamily>:


<excerpt><fontfamily><param>Hiragino Kaku Gothic Pro</param> I am not
anti-Mac, just curious to know what is an

actual benefit in even thinking about Macs in a cluster.


Thank you,

Ognen


On Thu, 10 Apr 2003, Stephen Wicks wrote:


<excerpt>Hi Chip,


Whenever I see someone considering a Linux/Windows dual boot setup for

a single job in biology, I recommend that they consider a Macintosh.

OSX will give you fully supported access to all of the command line or

X11 apps that you will need (including in your case RasMol, Povray,

Molscript, raster3d etc.), and still let your users use

Office/Photoshop etc. at the same time on the same machine and same OS.

Networking and interoperability these days is a non-issue. The only

caveat right now might be that the G4 towers are fast, but not as fast

as the top end Intel-based solution.


You wanted options...


Stephen

</excerpt></fontfamily></excerpt>
--Apple-Mail-12-840296516--