[Bioclusters] Use of FPGA based BLAST accelerators?
tristan
bioclusters@bioinformatics.org
Thu, 12 Dec 2002 11:15:16 -0800
This is a multi-part message in MIME format.
------=_NextPart_000_0012_01C2A1CF.C1018640
Content-Type: text/plain;
charset="us-ascii"
Content-Transfer-Encoding: 7bit
Hello Amar,
Does anyone on the list have any experience with using some of the
commercially available FPGA accelerators for Bioinformatics applications
(e.g. from Timelogic)?
[tristan]
I have been working on a technical team developing computational
accelerators for biological research for some time now. Our research has
led us to use ASIC chips rather than FPGA technology. We have found that
although FPGA certainly has its merits, once you get the code right maximum
acceleration can be achieved by burning it into silicon. Using ASIC
technology we are able to get over a 1000x speed improvement over
traditional computational methods.
Of course, speed is not the only factor to consider when developing
hardware acceleration. The real challenge for accelerators is not to
sacrifice accuracy for speed.
Specifically regarding Hmmer:
We have worked very closely with Sean Eddy on our implementation of Hmmer.
Here's what he says about our collaboration on his website, "We participate
in a close collaborative relationship with the folks at Paracel with the aim
of assuring native HMMER compatibility with the Paracel GeneMatcher
hardware." (http://hmmer.wustl.edu/) Let me know if you would like to speak
with our developers who worked with Sean Eddy. You might also be interested
in our ability to implement Hmmer using SAM models with Viterbi scoring and
that we work very well with Panther models.
Based upon some of the performance claims I've seen, it seems to me that
augmenting a standard Linux cluster with one of these devices should provide
pretty compelling price/performance. Why is this not commonly done?
[tristan]
You are absolutely correct. Using Linux clusters with hardware
acceleration does give the best price performance. By combining our
hardware accelerator with the power of Linux clusters we are able to improve
the speed of our product while at the same time greatly reducing the cost.
By coincidence Paracel offers a product that uses Linux clusters and
hardware acceleration. This product is called GeneMatcher2.J Feel free to
give me a call or drop an email if you have any questions.
Best regards,
Tristan Gill
Applications Scientist
Paracel Inc.
tristan@paracel.com
Office: 626-744-2064
Cell: 626-327-0707
Fax: 626-744-2001
Thanks very much!
Amar
------=_NextPart_000_0012_01C2A1CF.C1018640
Content-Type: text/html;
charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML xmlns=3D"http://www.w3.org/TR/REC-html40" xmlns:o =3D=20
"urn:schemas-microsoft-com:office:office" xmlns:w =3D=20
"urn:schemas-microsoft-com:office:word"><HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Dus-ascii">
<META content=3DWord.Document name=3DProgId>
<META content=3D"MSHTML 5.50.4522.1800" name=3DGENERATOR>
<META content=3D"Microsoft Word 10" name=3DOriginator><LINK=20
href=3D"cid:filelist.xml@01C2A110.5809C860" rel=3DFile-List><!--[if gte =
mso 9]><xml>
<o:OfficeDocumentSettings>
<o:DoNotRelyOnCSS/>
</o:OfficeDocumentSettings>
</xml><![endif]--><!--[if gte mso 9]><xml>
<w:WordDocument>
<w:SpellingState>Clean</w:SpellingState>
<w:GrammarState>Clean</w:GrammarState>
<w:DocumentKind>DocumentEmail</w:DocumentKind>
<w:EnvelopeVis/>
<w:Compatibility>
<w:BreakWrappedTables/>
<w:SnapToGridInCell/>
<w:WrapTextWithPunct/>
<w:UseAsianBreakRules/>
</w:Compatibility>
<w:BrowserLevel>MicrosoftInternetExplorer4</w:BrowserLevel>
</w:WordDocument>
</xml><![endif]-->
<STYLE>@page Section1 {size: 8.5in 11.0in; margin: 1.0in 1.25in 1.0in =
1.25in; mso-header-margin: .5in; mso-footer-margin: .5in; =
mso-paper-source: 0; }
P.MsoNormal {
FONT-SIZE: 12pt; MARGIN: 0in 0in 0pt; FONT-FAMILY: "Times New Roman"; =
mso-style-update: auto; mso-style-parent: ""; mso-pagination: =
widow-orphan; mso-fareast-font-family: "Times New Roman"
}
LI.MsoNormal {
FONT-SIZE: 12pt; MARGIN: 0in 0in 0pt; FONT-FAMILY: "Times New Roman"; =
mso-style-update: auto; mso-style-parent: ""; mso-pagination: =
widow-orphan; mso-fareast-font-family: "Times New Roman"
}
DIV.MsoNormal {
FONT-SIZE: 12pt; MARGIN: 0in 0in 0pt; FONT-FAMILY: "Times New Roman"; =
mso-style-update: auto; mso-style-parent: ""; mso-pagination: =
widow-orphan; mso-fareast-font-family: "Times New Roman"
}
A:link {
COLOR: blue; TEXT-DECORATION: underline; text-underline: single
}
SPAN.MsoHyperlink {
COLOR: blue; TEXT-DECORATION: underline; text-underline: single
}
A:visited {
COLOR: purple; TEXT-DECORATION: underline; text-underline: single
}
SPAN.MsoHyperlinkFollowed {
COLOR: purple; TEXT-DECORATION: underline; text-underline: single
}
SPAN.EmailStyle17 {
COLOR: windowtext; FONT-FAMILY: Arial; mso-style-type: =
personal-compose; mso-style-noshow: yes; mso-ansi-font-size: 10.0pt; =
mso-bidi-font-size: 10.0pt; mso-ascii-font-family: Arial; =
mso-hansi-font-family: Arial; mso-bidi-font-family: Arial
}
SPAN.SpellE {
mso-style-name: ""; mso-spl-e: yes
}
DIV.Section1 {
page: Section1
}
</STYLE>
<!--[if gte mso 10]>
<style>
/* Style Definitions */=20
table.MsoNormalTable
{mso-style-name:"Table Normal";
mso-tstyle-rowband-size:0;
mso-tstyle-colband-size:0;
mso-style-noshow:yes;
mso-style-parent:"";
mso-padding-alt:0in 5.4pt 0in 5.4pt;
mso-para-margin:0in;
mso-para-margin-bottom:.0001pt;
mso-pagination:widow-orphan;
font-size:10.0pt;
font-family:"Times New Roman";}
</style>
<![endif]--></HEAD>
<BODY lang=3DEN-US style=3D"tab-interval: .5in" vLink=3Dpurple =
link=3Dblue>
<DIV><SPAN class=3D435171019-12122002><FONT face=3DArial color=3D#0000ff =
size=3D2>Hello=20
Amar,</FONT></SPAN><FONT face=3DTahoma size=3D2><BR></FONT></DIV>
<BLOCKQUOTE dir=3Dltr=20
style=3D"PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #0000ff 2px =
solid; MARGIN-RIGHT: 0px">
<DIV class=3DSection1>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt"><SPAN class=3D435171019-12122002><FONT =
face=3DArial=20
color=3D#0000ff size=3D2> </FONT></SPAN>Does anyone on the list =
have any=20
experience with using some of the commercially available FPGA =
accelerators for=20
Bioinformatics applications (e.g. from <SPAN=20
class=3DSpellE>Timelogic</SPAN>)?<BR></SPAN></FONT></P>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt"><FONT face=3DArial><FONT =
color=3D#0000ff><FONT=20
size=3D2><SPAN=20
=
class=3D435171019-12122002>[tristan] </SPAN><o:p></o:p></FONT>=
</FONT></FONT></SPAN></FONT></P>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt"><o:p><SPAN class=3D435171019-12122002><FONT =
face=3DArial=20
color=3D#0000ff size=3D2><FONT face=3D"Times New Roman" =
color=3D#000000=20
size=3D3> </FONT></P>
<P class=3DMsoNormal>I have been working on a technical team =
developing=20
computational accelerators for biological research for some time =
now.<SPAN=20
style=3D"mso-spacerun: yes"> </SPAN>Our research has led us to =
use ASIC=20
chips rather than FPGA technology.<SPAN style=3D"mso-spacerun: =
yes"> =20
</SPAN>We have found that although FPGA certainly has its merits, once =
you get=20
the code right maximum acceleration can be achieved by burning it into =
silicon.<SPAN style=3D"mso-spacerun: yes"> </SPAN>Using ASIC =
technology we=20
are able to get over a 1000x speed improvement over traditional =
computational=20
methods. </P>
<P class=3DMsoNormal> <o:p></o:p></P>
<P class=3DMsoNormal><SPAN style=3D"mso-spacerun: yes"> </SPAN>Of =
course,=20
speed is not the only factor to consider when developing hardware=20
acceleration.<SPAN style=3D"mso-spacerun: yes"> </SPAN>The real =
challenge=20
for accelerators is not to sacrifice accuracy for speed.<SPAN=20
style=3D"mso-spacerun: yes"> </SPAN></P>
<P class=3DMsoNormal> <o:p></o:p></P>
<P class=3DMsoNormal>Specifically regarding Hmmer:</P>
<P class=3DMsoNormal> <o:p></o:p></P>
<P class=3DMsoNormal><SPAN style=3D"mso-bidi-font-size: 10.0pt">We =
have worked=20
very closely with Sean Eddy on our implementation of Hmmer.<SPAN=20
style=3D"mso-spacerun: yes"> </SPAN>Here's what he says about =
our=20
collaboration on his website, "We participate in a close collaborative =
relationship with the folks at Paracel with the aim of assuring native =
HMMER=20
compatibility with the Paracel GeneMatcher hardware."=20
(http://hmmer.wustl.edu/)<SPAN style=3D"mso-spacerun: yes"> =
</SPAN>Let me=20
know if you would like to speak with our developers who worked with =
Sean=20
Eddy.<SPAN style=3D"mso-spacerun: yes"> </SPAN>You might also be =
interested in our ability to implement Hmmer using SAM models with =
Viterbi=20
scoring and that we work very well with Panther =
models.<o:p></o:p></SPAN></P>
<P =
class=3DMsoNormal> </FONT></SPAN> </o:p></SPAN></FONT><FO=
NT=20
face=3DArial color=3D#0000ff><SPAN style=3D"FONT-SIZE: =
12pt"><o:p><SPAN=20
class=3D435171019-12122002> </SPAN></o:p></SPAN></FONT></P>
<P class=3DMsoNormal><FONT face=3DArial color=3D#0000ff size=3D2><SPAN =
style=3D"FONT-SIZE: 12pt"><o:p></o:p></SPAN></FONT> </P>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt">Based upon some of the performance claims =
I’ve seen,=20
it seems to me that augmenting a standard Linux cluster with one of =
these=20
devices should provide pretty compelling price/performance.<SPAN=20
style=3D"mso-spacerun: yes"> </SPAN>Why is this not commonly=20
done?<BR><FONT face=3DArial><FONT color=3D#0000ff><FONT size=3D2><SPAN =
=
class=3D435171019-12122002></SPAN></FONT></FONT></FONT></SPAN></FONT></P>=
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt"><FONT face=3DArial><FONT =
color=3D#0000ff><FONT=20
size=3D2><SPAN=20
=
class=3D435171019-12122002>[tristan]</SPAN></FONT></FONT></FONT></SPAN></=
FONT></P>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt"><FONT face=3DArial><FONT =
color=3D#0000ff><FONT=20
size=3D2><SPAN=20
=
class=3D435171019-12122002></SPAN></FONT></FONT></FONT></SPAN></FONT>&nbs=
p;</P>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt"><FONT face=3DArial><FONT =
color=3D#0000ff><FONT=20
size=3D2><SPAN class=3D435171019-12122002><FONT face=3D"Times New =
Roman"=20
color=3D#000000 size=3D3> </FONT></P>
<P class=3DMsoNormal>You are absolutely correct.<SPAN=20
style=3D"mso-spacerun: yes"> </SPAN>Using Linux clusters with =
hardware=20
acceleration does give the best price performance.<SPAN=20
style=3D"mso-spacerun: yes"> </SPAN>By combining our hardware =
accelerator=20
with the power of Linux clusters we are able to improve the speed of =
our=20
product while at the same time greatly reduc<SPAN=20
class=3D435171019-12122002><FONT face=3DArial =
size=3D2>ing </FONT></SPAN> the=20
cost.<SPAN style=3D"mso-spacerun: yes"> </SPAN></P>
<P class=3DMsoNormal> <o:p></o:p></P>
<P class=3DMsoNormal>By coincidence Paracel offers a product that uses =
Linux=20
clusters and hardware acceleration.<SPAN style=3D"mso-spacerun: =
yes"> =20
</SPAN>This product is called GeneMatcher2.<SPAN=20
style=3D"FONT-FAMILY: Wingdings; mso-ascii-font-family: 'Times New =
Roman'; mso-hansi-font-family: 'Times New Roman'; mso-char-type: symbol; =
mso-symbol-font-family: Wingdings"><SPAN=20
style=3D"mso-char-type: symbol; mso-symbol-font-family: =
Wingdings">J</SPAN></SPAN><SPAN=20
style=3D"mso-spacerun: yes"> </SPAN>Feel free to give me a =
call or=20
drop an email if you have any questions.</P>
<P class=3DMsoNormal> <o:p></o:p></P>
<P class=3DMsoNormal>Best regards,</P>
<P class=3DMsoNormal> <o:p></o:p></P>
<P class=3DMsoNormal style=3D"mso-layout-grid-align: none"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Tristan =
Gill<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"mso-layout-grid-align: none"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Applications=20
Scientist<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"mso-layout-grid-align: none"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Paracel =
Inc.<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"mso-layout-grid-align: none"><U><SPAN=20
style=3D"FONT-SIZE: 10pt; COLOR: blue; FONT-FAMILY: =
Arial">tristan@paracel.com</SPAN></U><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial"><o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"mso-layout-grid-align: none"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Office:<SPAN=20
style=3D"mso-spacerun: yes"> =
</SPAN>626-744-2064<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"mso-layout-grid-align: none"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Cell:<SPAN=20
style=3D"mso-spacerun: yes"> =20
</SPAN>626-327-0707<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"mso-layout-grid-align: none"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Fax:<SPAN=20
style=3D"mso-spacerun: yes"> =20
</SPAN>626-744-2001<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"mso-layout-grid-align: none"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"> <o:p></o:p></SPAN></P>
<P=20
=
class=3DMsoNormal> </SPAN><o:p></o:p></FONT></FONT></FONT></SP=
AN></FONT></P>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt"><o:p> </o:p></SPAN></FONT></P>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt">Thanks very =
much!<o:p></o:p></SPAN></FONT></P>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: 12pt"><o:p> </o:p></SPAN></FONT></P>
<P class=3DMsoNormal><FONT face=3D"Times New Roman" size=3D3><SPAN=20
style=3D"FONT-SIZE: =
12pt">Amar<o:p></o:p></SPAN></FONT></P></DIV></BLOCKQUOTE></BODY></HTML>
------=_NextPart_000_0012_01C2A1CF.C1018640--