<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta charset="utf-8">
<meta name="generator" content="pdf2htmlEX">
<meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1">
<link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css">
<link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css">
<link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/6253c03b74bc5c01050c640e/raw.css">
<script src="https://static.pudn.com/base/js/compatibility.min.js"></script>
<script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script>
<script>
try{
pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({});
}catch(e){}
</script>
<title></title>
</head>
<body>
<div id="sidebar" style="display: none">
<div id="outline">
</div>
</div>
<div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/6253c03b74bc5c01050c640e/bg1.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0">DO<span class="_ _0"></span>T<span class="_ _1"></span>:<span class="_"> </span><span class="ff2 fc1 sc1 ls1">一个开发处理大数据软件的分析模型</span></div><div class="t m0 x2 h3 y2 ff2 fs1 fc2 sc2 ls2 ws0">张晓东</div><div class="t m0 x3 h4 y3 ff2 fs2 fc0 sc3 ls1 ws0">美国俄亥俄州立大学</div><div class="t m0 x4 h5 y4 ff1 fs3 fc2 sc0 ls0 ws0">T<span class="_ _2"></span>eam<span class="_"> </span>Members</div><div class="t m0 x5 h6 y5 ff3 fs4 fc1 sc0 ls0 ws0">Y<span class="_ _3"></span>in Huai</div><div class="t m0 x6 h7 y6 ff3 fs5 fc1 sc0 ls0 ws0">1</div><div class="t m0 x7 h8 y5 ff4 fs4 fc1 sc0 ls0 ws0">, Rubao Lee</div><div class="t m0 x8 h9 y6 ff4 fs5 fc1 sc0 ls0 ws0">1</div><div class="t m0 x9 h8 y5 ff4 fs4 fc1 sc0 ls3 ws0">, </div><div class="t m0 xa h8 y7 ff4 fs4 fc1 sc0 ls0 ws0">Simon Zhang</div><div class="t m0 xb h9 y8 ff4 fs5 fc1 sc0 ls0 ws0">2</div><div class="t m0 xc h8 y7 ff4 fs4 fc1 sc0 ls0 ws0">, Cath<span class="_ _0"></span>y H. Xia</div><div class="t m0 xd h9 y8 ff4 fs5 fc1 sc0 ls0 ws0">1</div><div class="t m0 xe ha y9 ff1 fs5 fc1 sc0 ls0 ws0">1</div><div class="t m0 xf hb ya ff1 fs4 fc1 sc0 ls0 ws0">Dept. of<span class="_ _4"> </span> Computer <span class="_ _5"></span>Sci. & Eng<span class="_ _1"></span>.<span class="_ _5"></span>, <span class="fc0">The Ohio <span class="_ _5"></span>State Uni<span class="_ _0"></span>versity</span></div><div class="t m0 x10 ha yb ff1 fs5 fc1 sc0 ls0 ws0">2</div><div class="t m0 x11 hb yc ff1 fs4 fc1 sc0 ls0 ws0">Depar<span class="_ _5"></span>tment of<span class="_ _4"> </span> Computer <span class="_ _5"></span>Science, <span class="_ _5"></span><span class="fc0">Cor<span class="_ _5"></span>nell Univ<span class="_ _0"></span>er<span class="_ _5"></span>sity</span></div></div><div class="pi" data-data='{"ctm":[1.333333,0.000000,0.000000,1.333333,0.000000,0.000000]}'></div></div>
</body>
</html>
<div id="pf2" class="pf w0 h0" data-page-no="2"><div class="pc pc2 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/6253c03b74bc5c01050c640e/bg2.jpg"><div class="t m0 x12 hc yd ff1 fs6 fc1 sc0 ls0 ws0">T<span class="_ _5"></span>he Ev<span class="_ _1"></span>olution of<span class="_ _6"> </span> Computing</div><div class="t m0 x13 hd ye ff5 fs7 fc1 sc0 ls0 ws0"><span class="_ _7"></span><span class="ff1">Centraliz<span class="_ _3"></span>ed computing era</span></div><div class="t m0 x14 he yf ff6 fs8 fc1 sc0 ls0 ws0">•<span class="_ _8"> </span><span class="ff7">T<span class="_ _5"></span>he <span class="ff1 fc0">V<span class="_ _9"></span>on Neumann model <span class="ff7 fc1">(1945): A baseline mod<span class="_ _5"></span>el <span class="_ _0"></span>to <span class="_ _5"></span>guide </span></span></span></div><div class="t m0 x15 hf y10 ff7 fs8 fc1 sc0 ls0 ws0">centralized compute<span class="_ _3"></span>r arc<span class="_ _0"></span>hitecture <span class="_ _5"></span>design</div><div class="t m0 x13 hd y11 ff5 fs7 fc1 sc0 ls0 ws0"><span class="_ _7"></span><span class="ff1">P<span class="_ _2"></span>aral<span class="_ _5"></span>lel <span class="_ _0"></span>computing <span class="_ _5"></span>era</span></div><div class="t m0 x14 he y12 ff6 fs8 fc1 sc0 ls0 ws0">•<span class="_ _8"> </span><span class="ff7">T<span class="_ _5"></span>he B<span class="_ _3"></span>ulk Sync<span class="_ _0"></span>hron<span class="_ _5"></span>ous <span class="_ _0"></span>P<span class="_ _0"></span>arallel (<span class="_ _5"></span><span class="ff1 fc0">BSP</span>) model <span class="_ _3"></span>(<span class="fc3">Leslie V<span class="_ _a"></span>aliant<span class="fc1 ls4">, </span></span></span></div><div class="t m0 x15 hf y13 ff8 fs8 fc1 sc0 ls0 ws0">1990): A “scale<span class="ff7">-</span>up” model to guide parallel arc<span class="_ _0"></span>hitecture </div><div class="t m0 x15 hf y14 ff7 fs8 fc1 sc0 ls0 ws0">impro<span class="_ _3"></span>v<span class="_ _0"></span>ement and <span class="_ _3"></span>softw<span class="_ _0"></span>are optimization<span class="_ _5"></span> <span class="_ _3"></span>softw<span class="_ _0"></span>are for HPC</div><div class="t m0 x13 h10 y15 ff5 fs9 fc1 sc0 ls0 ws0"><span class="_ _7"></span><span class="ff9">“The data center as a computer” <span class="_ _5"></span>era</span></div><div class="t m0 x14 he y16 ff6 fs8 fc1 sc0 ls0 ws0">•<span class="_ _8"> </span><span class="ff7">Only Softw<span class="_ _0"></span>are framew<span class="_ _0"></span>orks for big <span class="_ _3"></span>data analytics <span class="_ _5"></span>a<span class="_ _0"></span>v<span class="_ _0"></span>ailab<span class="_ _5"></span>le</span></div><div class="t m0 x1 h11 y17 ffa fs2 fc1 sc0 ls0 ws0">–<span class="_ _6"> </span><span class="ff7">MapR<span class="_ _0"></span>educe<span class="_ _3"></span><span class="ls5">, <span class="ls0">Hadoop<span class="_ _1"></span>, Dr<span class="_ _b"></span>yad and sev<span class="_ _0"></span>eral others</span></span></span></div><div class="t m0 x14 h12 y18 ff6 fsa fc0 sc0 ls0 ws0">•<span class="_ _8"> </span><span class="ff1">No model<span class="_ _3"></span>s y<span class="_ _0"></span>et</span></div><div class="t m0 x16 h5 y19 ff7 fs3 fc4 sc0 ls0 ws0">3</div></div><div class="pi" data-data='{"ctm":[1.333333,0.000000,0.000000,1.333333,0.000000,0.000000]}'></div></div>
<div id="pf3" class="pf w0 h0" data-page-no="3"><div class="pc pc3 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/6253c03b74bc5c01050c640e/bg3.jpg"><div class="t m0 x17 h13 y1a ff1 fs7 fc0 sc0 ls0 ws0">V<span class="_ _9"></span>on Neumann Model:<span class="_ _5"></span> <span class="fc1">Computer Architectur<span class="_ _5"></span>e Design</span></div><div class="t m0 x6 h11 y1b ff7 fs2 fc1 sc0 ls0 ws0">Before <span class="fc0">V<span class="_ _a"></span>on Neumann<span class="_ _2"></span><span class="ff8 fc1">’<span class="_ _a"></span>s com<span class="_ _3"></span>puter </span></span></div><div class="t m0 x18 h11 y1c ff7 fs2 fc1 sc0 ls0 ws0">project, sev<span class="_ _2"></span>eral <span class="_ _5"></span>operational </div><div class="t m0 x18 h14 y1d ff7 fsb fc1 sc0 ls0 ws0">computers w<span class="_ _0"></span>ere built: </div><div class="t m0 x19 h15 y1e ff6 fsc fc1 sc0 ls0 ws0">•<span class="_ _c"> </span><span class="ff7">1936, <span class="ff8">Zuse’<span class="_ _a"></span>s<span class="_ _d"> </span><span class="ff7 fc0 ls6">Z1<span class="_"> </span><span class="fc1 ls7">(1</span></span></span></span></div><div class="t m0 x1a h16 y1f ff7 fsd fc1 sc0 ls8 ws0">st</div><div class="t m0 x1b h17 y1e ff7 fsc fc1 sc0 ls0 ws0">binar<span class="_ _5"></span>y ditital</div><div class="t m0 x1c h17 y20 ff7 fsc fc1 sc0 ls0 ws0">computer) in Ger<span class="_ _5"></span>many</div><div class="t m0 x19 h15 y21 ff6 fsc fc1 sc0 ls0 ws0">•<span class="_ _c"> </span><span class="ff7">1937, Atanasoff<span class="_ _e"> </span><span class="ff8">and Ber<span class="_ _5"></span>r<span class="_ _5"></span>y’<span class="_ _1"></span>s <span class="_ _3"></span><span class="ff7 fc0">ABC<span class="_ _d"> </span><span class="fc1 ls7">(1</span></span></span></span></div><div class="t m0 x1d h16 y22 ff7 fsd fc1 sc0 ls8 ws0">st</div><div class="t m0 x1c h17 y23 ff7 fsc fc1 sc0 ls0 ws0">electronic digital computer)<span class="_ _5"></span> <span class="_ _0"></span>in Iow<span class="_ _0"></span>a </div><div class="t m0 x1c h17 y24 ff7 fsc fc1 sc0 ls0 ws0">State Uni<span class="_ _0"></span>versity</div><div class="t m0 x19 h15 y25 ff6 fsc fc1 sc0 ls0 ws0">•<span class="_ _c"> </span><span class="ff7 ls9">1943,<span class="_"> </span><span class="fc0 ls0">ENIA<span class="_ _0"></span>C <span class="fc1">based <span class="_ _3"></span>on ABC in UP<span class="_ _0"></span>enn</span></span></span></div><div class="t m0 x6 h11 y26 ff7 fs2 fc1 sc0 ls0 ws0">T<span class="_ _5"></span>he <span class="_ _3"></span>most impor<span class="_ _5"></span>tant milestone V<span class="_ _a"></span>on </div><div class="t m0 x18 h11 y27 ff8 fs2 fc1 sc0 ls0 ws0">Neumann left is his paper<span class="_ _5"></span>: “<span class="ff7 fc5">First </span></div><div class="t m0 x18 h11 y28 ff7 fs2 fc5 sc0 ls0 ws0">Draft of<span class="_ _f"> </span> a R<span class="_ _0"></span>epor<span class="_ _5"></span>t to the ED<span class="_ _0"></span>V<span class="_ _9"></span>A<span class="_ _2"></span>C<span class="ff8 fc1">”, </span></div><div class="t m0 x18 h11 y29 ff7 fs2 fc1 sc0 ls0 ws0">1945. (a c<span class="_ _3"></span>onsulting repor<span class="_ _5"></span>t to U<span class="_ _5"></span>S </div><div class="t m0 x18 h14 y2a ff7 fsb fc1 sc0 ls0 ws0">Ar<span class="_ _b"></span>m<span class="_ _0"></span>y) before his IAS Computer </div><div class="t m0 x18 h11 y2b ff7 fs2 fc1 sc0 ls0 ws0">Project</div><div class="t m0 x13 h18 y2c ff4 fse fc0 sc0 ls0 ws0">V<span class="_ _2"></span>on Ne<span class="_ _5"></span>umann model </div><div class="t m0 x13 h19 y2d ff6 fse fc1 sc0 ls0 ws0">•<span class="_ _10"> </span><span class="ff4 fsc fc5">a memory<span class="_"> </span><span class="fc1">con<span class="_ _3"></span>t<span class="_ _3"></span>aining both da<span class="_ _0"></span>ta and </span></span></div><div class="t m0 x13 h1a y2e ff4 fsf fc1 sc0 ls0 ws0">ins<span class="_ _3"></span>tructions </div><div class="t m0 x13 h19 y2f ff6 fs4 fc1 sc0 ls0 ws0">•<span class="_ _11"> </span><span class="ff4 fsc fc5">a computing <span class="_ _3"></span>unit <span class="fc1">f<span class="_ _0"></span>or both arithmetic </span></span></div><div class="t m0 x13 h19 y30 ff4 fsc fc1 sc0 ls0 ws0">and logic<span class="_ _3"></span>al oper<span class="_ _0"></span>ations </div><div class="t m0 x13 h19 y31 ff6 fs4 fc1 sc0 ls0 ws0">•<span class="_ _11"> </span><span class="ff4 fsc fc5">a contr<span class="_ _0"></span>ol unit <span class="fc1">t<span class="_ _3"></span>o in<span class="_ _3"></span>t<span class="_ _3"></span>erpr<span class="_ _3"></span>e<span class="_ _3"></span>t </span></span></div><div class="t m0 x13 h19 y32 ff4 fsc fc1 sc0 ls0 ws0">ins<span class="_ _3"></span>tructions and mak<span class="_ _2"></span>e actions</div></div><div class="pi" data-data='{"ctm":[1.333333,0.000000,0.000000,1.333333,0.000000,0.000000]}'></div></div>
<div id="pf4" class="pf w0 h0" data-page-no="4"><div class="pc pc4 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/6253c03b74bc5c01050c640e/bg4.jpg"><div class="t m0 x1e h1b y33 ff1 fs10 fc0 sc0 ls0 ws0">BSP<span class="_"> </span><span class="fc1">is a Scale<span class="_ _3"></span>-Up Model f<span class="_ _5"></span>or <span class="_ _0"></span>HPC </span></div><div class="t m0 x1f h1c y34 ff5 fsc fc1 sc0 ls0 ws0"><span class="_ _12"> </span><span class="ff1">A parall<span class="_ _3"></span>el architecture model</span></div><div class="t m0 x20 h1d y35 ff6 fs11 fc0 sc0 ls0 ws0">•<span class="_ _13"> </span><span class="ff1">K<span class="_ _5"></span>ey paramet<span class="_ _3"></span>ers: <span class="_ _5"></span><span class="ff7 fc1">p<span class="_ _2"></span>, node sp<span class="_ _5"></span>eed, <span class="_ _3"></span>messag<span class="_ _5"></span>e </span></span></div><div class="t m0 x21 h1e y36 ff7 fs11 fc1 sc0 ls0 ws0">speed, sync<span class="_ _3"></span>h latency </div><div class="t m0 x20 h1f y37 ff6 fs12 fc0 sc0 ls0 ws0">•<span class="_ _13"> </span><span class="ff1">Lo<span class="_ _3"></span>w ratio of<span class="_ _f"> </span> comput<span class="_ _3"></span>ing/message <span class="_ _5"></span><span class="ff7 fc1">is key </span></span></div><div class="t m0 x1f h1c y38 ff5 fsc fc1 sc0 ls0 ws0"><span class="_ _12"> </span><span class="ff1">A pro<span class="_ _0"></span>g<span class="_ _5"></span>ramming model</span></div><div class="t m0 x20 h1d y39 ff6 fs11 fc0 sc0 ls0 ws0">•<span class="_ _13"> </span><span class="ff1">R<span class="_ _3"></span>educing messa<span class="_ _5"></span>ge/synch latency <span class="ff7 fc1">is key: </span></span></div><div class="t m0 x22 h20 y3a ffa fs13 fc1 sc0 ls0 ws0">–<span class="_ _14"> </span><span class="ff7">Ov<span class="_ _3"></span>erlapping <span class="_ _0"></span>computing and <span class="_ _3"></span>comm<span class="_ _0"></span>unication</span></div><div class="t m0 x22 h20 y3b ffa fs13 fc1 sc0 ls0 ws0">–<span class="_ _14"> </span><span class="ff7">Exploiting <span class="_ _0"></span>locality</span></div><div class="t m0 x22 h20 y3c ffa fs13 fc1 sc0 ls0 ws0">–<span class="_ _14"> </span><span class="ff7">Load balance <span class="_ _0"></span>to minimize sync<span class="_ _0"></span>h latency</span></div><div class="t m0 x1f h1c y3d ff5 fsc fc1 sc0 ls0 ws0"><span class="_ _12"> </span><span class="ff1">A cost <span class="_ _3"></span>model </span></div><div class="t m0 x20 h1d y3e ff6 fs11 fc1 sc0 ls0 ws0">•<span class="_ _13"> </span><span class="ff7">Combining <span class="_ _5"></span>both hardw<span class="_ _3"></span>are/software </span></div><div class="t m0 x21 h1e y3f ff7 fs11 fc1 sc0 ls0 ws0">parameters<span class="_ _0"></span>, w<span class="_ _3"></span>e can predict <span class="_ _5"></span><span class="ff1 fc0">ex<span class="_ _0"></span>ecution time</span></div><div class="t m0 x1f h1c y40 ff5 fsc fc1 sc0 ls0 ws0"><span class="_ _12"> </span><span class="ff1">BSP doe<span class="_ _3"></span>s not <span class="_ _3"></span>suppor<span class="_ _5"></span>t</span></div><div class="t m0 x20 h1d y41 ff6 fs11 fc3 sc0 ls0 ws0">•<span class="_ _13"> </span><span class="ff1">Data-intensi<span class="_ _3"></span>v<span class="_ _0"></span>e <span class="_ _5"></span>applications, <span class="_ _5"></span>big data </span></div><div class="t m0 x20 h1f y42 ff6 fs12 fc3 sc0 ls0 ws0">•<span class="_ _13"> </span><span class="ff1">hardw<span class="_ _3"></span>are independ<span class="_ _3"></span>ent <span class="_ _5"></span>perfor<span class="_ _b"></span>mance</span></div><div class="t m0 x20 h1d y43 ff6 fs11 fc3 sc0 ls0 ws0">•<span class="_ _13"> </span><span class="ff1">sustained scalabi<span class="_ _3"></span>lity <span class="_ _5"></span>and high throughput <span class="_ _5"></span> </span></div><div class="t m0 x16 h5 y19 ff7 fs3 fc4 sc0 ls0 ws0">5</div><div class="t m0 x23 h21 y44 ff9 fs14 fc1 sc0 ls0 ws0">…</div><div class="t m0 x23 h22 y45 ff9 fs15 fc1 sc0 ls0 ws0">…</div><div class="t m0 x24 h23 y46 ff1 fse fc1 sc0 ls0 ws0">Bar<span class="_ _5"></span>rier</div><div class="t m1 x25 h24 y47 ff1 fs16 fc1 sc0 ls0 ws0">P</div><div class="t m1 x26 h25 y48 ff1 fs17 fc1 sc0 ls0 ws0">0</div><div class="t m1 x27 h24 y47 ff1 fs16 fc1 sc0 ls0 ws0">P</div><div class="t m1 x28 h25 y48 ff1 fs17 fc1 sc0 ls0 ws0">1</div><div class="t m1 x29 h24 y47 ff1 fs16 fc1 sc0 ls0 ws0">P</div><div class="t m1 x2a h25 y48 ff1 fs17 fc1 sc0 ls0 ws0">2</div><div class="t m1 x2b h24 y49 ff1 fs16 fc1 sc0 ls0 ws0">P</div><div class="t m1 x2c h25 y4a ff1 fs17 fc1 sc0 ls0 ws0">n</div><div class="t m2 x2d h23 y4b ff1 fse fc1 sc0 ls0 ws0">Super<span class="_ _5"></span>step<span class="_"> </span>i</div><div class="t m0 x23 h22 y4c ff9 fs15 fc1 sc0 ls0 ws0">…</div></div><div class="pi" data-data='{"ctm":[1.333333,0.000000,0.000000,1.333333,0.000000,0.000000]}'></div></div>
<div id="pf5" class="pf w0 h0" data-page-no="5"><div class="pc pc5 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/6253c03b74bc5c01050c640e/bg5.jpg"><div class="t m0 x2e h13 y1a ff1 fs7 fc0 sc0 ls0 ws0">Scale-out<span class="_"> </span><span class="fc1">is the F<span class="_ _0"></span>oundation <span class="_ _5"></span>for Big Data Ana<span class="_ _5"></span>lytics </span></div><div class="t m0 x13 h26 y4d ff5 fs8 fc1 sc0 ls0 ws0"><span class="_ _15"> </span><span class="ff1">Scale-out = sustained through<span class="_ _5"></span>put <span class="_ _3"></span>g<span class="_ _5"></span>rowth <span class="_ _3"></span>as # nodes g<span class="_ _5"></span>rows</span></div><div class="t m0 x14 h27 y4e ff6 fs2 fc1 sc0 ls0 ws0">•<span class="_ _16"> </span><span class="ff7">MR processed <span class="ff1 fc0">1 PB </span>data in 6h 2m on <span class="ff1 fc0">4000 nodes </span>in 11/2008</span></div><div class="t m0 x14 h28 y4f ff6 fsb fc1 sc0 ls0 ws0">•<span class="_ _16"> </span><span class="ff7">MR processe<span class="_ _3"></span>d <span class="ff1 fc0">10 PB </span>in 6 h 27 m on <span class="ff1 fc0">8000 no<span class="_ _3"></span>des<span class="_"> </span><span class="ff8 fc1">9/2011 (VLDB’11)</span></span></span></div><div class="t m0 x14 h27 y50 ff6 fs2 fc1 sc0 ls0 ws0">•<span class="_ _16"> </span><span class="ff7">T<span class="_ _5"></span>he <span class="_ _3"></span><span class="ff1 fc0">data is not mo<span class="_ _3"></span>v<span class="_ _0"></span>a<span class="_ _0"></span>ble<span class="_"> </span><span class="ff7 fc1">after it is placed in a system</span></span></span></div><div class="t m0 x13 h26 y51 ff5 fs8 fc1 sc0 ls0 ws0"><span class="_ _10"> </span><span class="ff1">Existing <span class="_ _3"></span>big data softw<span class="_ _0"></span>ar<span class="_ _5"></span>e <span class="_ _3"></span>is in a scale-out mode by</span></div><div class="t m0 x14 h27 y52 ff6 fs2 fc1 sc0 ls0 ws0">•<span class="_ _16"> </span><span class="ff7">F<span class="_ _2"></span>ocusing on scalability and fault-tolerance in larg<span class="_ _5"></span>e syst<span class="_ _3"></span>ems</span></div><div class="t m0 x14 h27 y53 ff6 fs2 fc1 sc0 ls0 ws0">•<span class="_ _16"> </span><span class="ff7">Pro<span class="_ _3"></span>vide a easy<span class="_ _3"></span>-<span class="lsa">to</span>-prog<span class="_ _b"></span>ramming en<span class="_ _0"></span>vir<span class="_ _5"></span>onme<span class="_ _3"></span>nt </span></div><div class="t m0 x13 h26 y54 ff5 fs8 fc1 sc0 ls0 ws0"><span class="_ _10"> </span><span class="ff1">Ef<span class="_ _5"></span>fecti<span class="_ _0"></span>v<span class="_ _3"></span>ely responds urge<span class="_ _5"></span>nt big data challenges</span></div><div class="t m0 x14 h27 y55 ff6 fs2 fc1 sc0 ls0 ws0">•<span class="_ _16"> </span><span class="ff7">P<span class="_ _0"></span>arallel <span class="_ _3"></span>databases with limited scalability cannot handle big data</span></div><div class="t m0 x14 h27 y56 ff6 fs2 fc1 sc0 ls0 ws0">•<span class="_ _16"> </span><span class="ff7">Big data<span class="_ _3"></span> demands a larg<span class="_ _5"></span>e scope of<span class="_ _f"> </span> data analytics </span></div><div class="t m0 x14 h28 y57 ff6 fsb fc1 sc0 ls0 ws0">•<span class="_ _16"> </span><span class="ff7">T<span class="_ _1"></span>raditional database business mo<span class="_ _3"></span>dels are not affordable</span></div><div class="t m0 x14 h27 y58 ff6 fs2 fc1 sc0 ls0 ws0">•<span class="_ _16"> </span><span class="ff7">Scale-out model is required for big data<span class="_ _3"></span> analytics </span></div><div class="t m0 x16 h5 y19 ff7 fs3 fc4 sc0 ls0 ws0">6</div></div><div class="pi" data-data='{"ctm":[1.333333,0.000000,0.000000,1.333333,0.000000,0.000000]}'></div></div>
