Return-Path: X-Original-To: archive-asf-public-internal@cust-asf2.ponee.io Delivered-To: archive-asf-public-internal@cust-asf2.ponee.io Received: from cust-asf.ponee.io (cust-asf.ponee.io [163.172.22.183]) by cust-asf2.ponee.io (Postfix) with ESMTP id 12825200B13 for ; Wed, 15 Jun 2016 09:47:52 +0200 (CEST) Received: by cust-asf.ponee.io (Postfix) id 10E74160A4D; Wed, 15 Jun 2016 07:47:52 +0000 (UTC) Delivered-To: archive-asf-public@cust-asf.ponee.io Received: from mail.apache.org (hermes.apache.org [140.211.11.3]) by cust-asf.ponee.io (Postfix) with SMTP id D790B160A4C for ; Wed, 15 Jun 2016 09:47:50 +0200 (CEST) Received: (qmail 15114 invoked by uid 500); 15 Jun 2016 07:47:50 -0000 Mailing-List: contact user-help@kudu.incubator.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: user@kudu.incubator.apache.org Delivered-To: mailing list user@kudu.incubator.apache.org Received: (qmail 15106 invoked by uid 99); 15 Jun 2016 07:47:49 -0000 Received: from pnap-us-west-generic-nat.apache.org (HELO spamd1-us-west.apache.org) (209.188.14.142) by apache.org (qpsmtpd/0.29) with ESMTP; Wed, 15 Jun 2016 07:47:49 +0000 Received: from localhost (localhost [127.0.0.1]) by spamd1-us-west.apache.org (ASF Mail Server at spamd1-us-west.apache.org) with ESMTP id 97AB8C2232 for ; Wed, 15 Jun 2016 07:47:49 +0000 (UTC) X-Virus-Scanned: Debian amavisd-new at spamd1-us-west.apache.org X-Spam-Flag: NO X-Spam-Score: 1.279 X-Spam-Level: * X-Spam-Status: No, score=1.279 tagged_above=-999 required=6.31 tests=[DKIM_SIGNED=0.1, DKIM_VALID=-0.1, HTML_MESSAGE=2, RCVD_IN_DNSWL_LOW=-0.7, RCVD_IN_MSPIKE_H3=-0.01, RCVD_IN_MSPIKE_WL=-0.01, SPF_PASS=-0.001] autolearn=disabled Authentication-Results: spamd1-us-west.apache.org (amavisd-new); dkim=pass (2048-bit key) header.d=cloudera-com.20150623.gappssmtp.com Received: from mx1-lw-eu.apache.org ([10.40.0.8]) by localhost (spamd1-us-west.apache.org [10.40.0.7]) (amavisd-new, port 10024) with ESMTP id mpcqEeoo2ZUk for ; Wed, 15 Jun 2016 07:47:47 +0000 (UTC) Received: from mail-wm0-f48.google.com (mail-wm0-f48.google.com [74.125.82.48]) by mx1-lw-eu.apache.org (ASF Mail Server at mx1-lw-eu.apache.org) with ESMTPS id E44345F19D for ; Wed, 15 Jun 2016 07:47:46 +0000 (UTC) Received: by mail-wm0-f48.google.com with SMTP id m124so22667623wme.1 for ; Wed, 15 Jun 2016 00:47:46 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudera-com.20150623.gappssmtp.com; s=20150623; h=mime-version:in-reply-to:references:date:message-id:subject:from:to; bh=+9Ap+ItkLQ8zbTnTwsrRnECpCiSPS9w//rotk9iHZfU=; b=lfjEHkQ+Qy/0yMkKtOff980KL8MuYpo2EkCVZxBRf578J8Tx0fZlygZ5Fmt50KY4k3 LrAVzHnoSJ15iOWZ1ZAZ/oQ4YfT5sFHWEra/lStg0lRAFB1Y7Auz9n/Z29WvysRh8zGP 6+B1vPPtb+D3XsUGfpb4dblKVm0nXfskQcwVJoGO+nAdSjzpjnD6mm1uc1pXIqi/qFTi 49aIl0MKJtUowKRRETKf2ftE6vAZu4NEOi1Y52TaNZvdVYMTEgAXuizgA0NPTHq/TUoO 66jy5TfOG3QN/EiTkN1Mna5P6+g0BDTiQVKS7p8JLYzflucUehwOVK5xJhkom5qKQy+I pelQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:mime-version:in-reply-to:references:date :message-id:subject:from:to; bh=+9Ap+ItkLQ8zbTnTwsrRnECpCiSPS9w//rotk9iHZfU=; b=RdHb9o+CKsIrqfCE16X0lz913P5ADhIYJsbwJXMnBzoDZ9L9A1cIe9uTkEm8cn+/Bw yBZuDHaZOviMu33d0/k72vQ759OSiVTi/qLWskOxizMggFeNieKGnHaxs4s+DizprS6G LbELGymQQApQ9Zy3TdZWs6HHx1iBQB+cNrGM0/4w1DqYXtlqkVx1yTQ6fPio0yibSnoB IzloPVj9xu4K/Kkeq5Oju0lKas8xqvPYBPLH0oWcpEB/gkAtPwbat/UUaOPj6Stmg8Z2 or7ksLUT7HerhS8MAZML2Hbq/UOcUTenWkznJh636rycVeJBh2IgK9rKZmYyLqDPfAN3 qquA== X-Gm-Message-State: ALyK8tI9X0ePM69FAmj5qC5xI/oveDOjai2ltvKfu4Z5hlFiVuWJbY5o0CIIKvsfQja3xV4T9OvOTWyQRVquxcZB MIME-Version: 1.0 X-Received: by 10.194.179.199 with SMTP id di7mr9955210wjc.26.1465976866495; Wed, 15 Jun 2016 00:47:46 -0700 (PDT) Received: by 10.28.99.3 with HTTP; Wed, 15 Jun 2016 00:47:45 -0700 (PDT) Received: by 10.28.99.3 with HTTP; Wed, 15 Jun 2016 00:47:45 -0700 (PDT) In-Reply-To: <2E7BBD97-2A48-49F8-AE0C-F7CF6D463EF6@gmail.com> References: <55B8BF95-5704-46CA-A336-64EE4D2B91B2@gmail.com> <0A7D041A-A72D-4151-9476-BCCEC157C5E4@gmail.com> <2E7BBD97-2A48-49F8-AE0C-F7CF6D463EF6@gmail.com> Date: Wed, 15 Jun 2016 09:47:45 +0200 Message-ID: Subject: Re: Performance Question From: Todd Lipcon To: user@kudu.incubator.apache.org Content-Type: multipart/alternative; boundary=047d7b604b10607dc905354c56b1 archived-at: Wed, 15 Jun 2016 07:47:52 -0000 --047d7b604b10607dc905354c56b1 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Hi Benjamin, What workload are you using for benchmarks? Using spark or something more custom? rdd or data frame or SQL, etc? Maybe you can share the schema and some queries Todd Todd On Jun 15, 2016 8:10 AM, "Benjamin Kim" wrote: > Hi Todd, > > Now that Kudu 0.9.0 is out. I have done some tests. Already, I am > impressed. Compared to HBase, read and write performance are better. Writ= e > performance has the greatest improvement (> 4x), while read is > 1.5x. > Albeit, these are only preliminary tests. Do you know of a way to really = do > some conclusive tests? I want to see if I can match your results on my 50 > node cluster. > > Thanks, > Ben > > On May 30, 2016, at 10:33 AM, Todd Lipcon wrote: > > On Sat, May 28, 2016 at 7:12 AM, Benjamin Kim wrote: > >> Todd, >> >> It sounds like Kudu can possibly top or match those numbers put out by >> Aerospike. Do you have any performance statistics published or any >> instructions as to measure them myself as good way to test? In addition, >> this will be a test using Spark, so should I wait for Kudu version 0.9.0 >> where support will be built in? >> > > We don't have a lot of benchmarks published yet, especially on the write > side. I've found that thorough cross-system benchmarks are very difficult > to do fairly and accurately, and often times users end up misguided if th= ey > pay too much attention to them :) So, given a finite number of developers > working on Kudu, I think we've tended to spend more time on the project > itself and less time focusing on "competition". I'm sure there are use > cases where Kudu will beat out Aerospike, and probably use cases where > Aerospike will beat Kudu as well. > > From my perspective, it would be great if you can share some details of > your workload, especially if there are some areas you're finding Kudu > lacking. Maybe we can spot some easy code changes we could make to improv= e > performance, or suggest a tuning variable you could change. > > -Todd > > >> On May 27, 2016, at 9:19 PM, Todd Lipcon wrote: >> >> On Fri, May 27, 2016 at 8:20 PM, Benjamin Kim wrote= : >> >>> Hi Mike, >>> >>> First of all, thanks for the link. It looks like an interesting read. I >>> checked that Aerospike is currently at version 3.8.2.3, and in the arti= cle, >>> they are evaluating version 3.5.4. The main thing that impressed me was >>> their claim that they can beat Cassandra and HBase by 8x for writing an= d >>> 25x for reading. Their big claim to fame is that Aerospike can write 1M >>> records per second with only 50 nodes. I wanted to see if this is real. >>> >> >> 1M records per second on 50 nodes is pretty doable by Kudu as well, >> depending on the size of your records and the insertion order. I've been >> playing with a ~70 node cluster recently and seen 1M+ writes/second >> sustained, and bursting above 4M. These are 1KB rows with 11 columns, an= d >> with pretty old HDD-only nodes. I think newer flash-based nodes could do >> better. >> >> >>> >>> To answer your questions, we have a DMP with user profiles with many >>> attributes. We create segmentation information off of these attributes = to >>> classify them. Then, we can target advertising appropriately for our sa= les >>> department. Much of the data processing is for applying models on all o= r if >>> not most of every profile=E2=80=99s attributes to find similarities (ne= arest >>> neighbor/clustering) over a large number of rows when batch processing = or a >>> small subset of rows for quick online scoring. So, our use case is a >>> typical advanced analytics scenario. We have tried HBase, but it doesn= =E2=80=99t >>> work well for these types of analytics. >>> >>> I read, that Aerospike in the release notes, they did do many >>> improvements for batch and scan operations. >>> >>> I wonder what your thoughts are for using Kudu for this. >>> >> >> Sounds like a good Kudu use case to me. I've heard great things about >> Aerospike for the low latency random access portion, but I've also heard >> that it's _very_ expensive, and not particularly suited to the columnar >> scan workload. Lastly, I think the Apache license of Kudu is much more >> appealing than the AGPL3 used by Aerospike. But, that's not really a dir= ect >> answer to the performance question :) >> >> >>> >>> Thanks, >>> Ben >>> >>> >>> On May 27, 2016, at 6:21 PM, Mike Percy wrote: >>> >>> Have you considered whether you have a scan heavy or a random access >>> heavy workload? Have you considered whether you always access / update = a >>> whole row vs only a partial row? Kudu is a column store so has some >>> awesome performance characteristics when you are doing a lot of scannin= g of >>> just a couple of columns. >>> >>> I don't know the answer to your question but if your concern is >>> performance then I would be interested in seeing comparisons from a per= f >>> perspective on certain workloads. >>> >>> Finally, a year ago Aerospike did quite poorly in a Jepsen test: >>> https://aphyr.com/posts/324-jepsen-aerospike >>> >>> I wonder if they have addressed any of those issues. >>> >>> Mike >>> >>> On Friday, May 27, 2016, Benjamin Kim wrote: >>> >>>> I am just curious. How will Kudu compare with Aerospike ( >>>> http://www.aerospike.com)? I went to a Spark Roadshow and found out >>>> about this piece of software. It appears to fit our use case perfectly >>>> since we are an ad-tech company trying to leverage our user profiles d= ata. >>>> Plus, it already has a Spark connector and has a SQL-like client. The >>>> tables can be accessed using Spark SQL DataFrames and, also, made into= SQL >>>> tables for direct use with Spark SQL ODBC/JDBC Thriftserver. I see fro= m the >>>> work done here http://gerrit.cloudera.org:8080/#/c/2992/ that the >>>> Spark integration is well underway and, from the looks of it lately, a= lmost >>>> complete. I would prefer to use Kudu since we are already a Cloudera s= hop, >>>> and Kudu is easy to deploy and configure using Cloudera Manager. I als= o >>>> hope that some of Aerospike=E2=80=99s speed optimization techniques ca= n make it >>>> into Kudu in the future, if they have not been already thought of or >>>> included. >>>> >>>> Just some thoughts=E2=80=A6 >>>> >>>> Cheers, >>>> Ben >>> >>> >>> >>> -- >>> -- >>> Mike Percy >>> Software Engineer, Cloudera >>> >>> >>> >>> >> >> >> -- >> Todd Lipcon >> Software Engineer, Cloudera >> >> >> > > > -- > Todd Lipcon > Software Engineer, Cloudera > > > --047d7b604b10607dc905354c56b1 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable

Hi Benjamin,

What workload are you using for benchmarks? Using spark or s= omething more custom? rdd or data frame or SQL, etc? Maybe you can share th= e schema and some queries

Todd

Todd

On Jun 15, 2016 8:10 AM, "Benjamin Kim"= ; <bbuild11@gmail.com> wrot= e:
Hi Todd,

Now that Kudu 0.9.0 is= out. I have done some tests. Already, I am impressed. Compared to HBase, r= ead and write performance are better. Write performance has the greatest im= provement (> 4x), while read is > 1.5x. Albeit, these are only prelim= inary tests. Do you know of a way to really do some conclusive tests? I wan= t to see if I can match your results on my 50 node cluster.

<= div>Thanks,
Ben

On May 30, 2016, at 10:33 AM, Todd = Lipcon <todd@clou= dera.com> wrote:

On Sat, May 28, 2016 at 7:12 AM, Benjamin Kim=C2=A0<bbuild= 11@gmail.com>=C2=A0wrote:
=
Todd,

It sounds like= Kudu can possibly top or match those numbers put out by Aerospike. Do you = have any performance statistics published or any instructions as to measure= them myself as good way to test? In addition, this will be a test using Sp= ark, so should I wait for Kudu version 0.9.0 where support will be built in= ?

We don't have a lot of be= nchmarks published yet, especially on the write side. I've found that t= horough cross-system benchmarks are very difficult to do fairly and accurat= ely, and often times users end up misguided if they pay too much attention = to them :) So, given a finite number of developers working on Kudu, I think= we've tended to spend more time on the project itself and less time fo= cusing on "competition". I'm sure there are use cases where K= udu will beat out Aerospike, and probably use cases where Aerospike will be= at Kudu as well.

From my perspective, it would be = great if you can share some details of your workload, especially if there a= re some areas you're finding Kudu lacking. Maybe we can spot some easy = code changes we could make to improve performance, or suggest a tuning vari= able you could change.

-Todd

<= blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-l= eft-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;pa= dding-left:1ex">

On May 27, 2016, at 9:19 PM, Todd Lipcon &l= t;todd@cloudera.com<= /a>> wrote:

On Fri, May 27, 2016 at 8:20 PM, Benjamin Kim=C2=A0<bbuild11@gmail.com>=C2=A0wrote:=
Hi Mike,

<= /div>
First of all, thanks for the link. It looks like an interesting r= ead. I checked that Aerospike is currently at version 3.8.2.3, and in the a= rticle, they are evaluating version 3.5.4. The main thing that impressed me= was their claim that they can beat Cassandra and HBase by 8x for writing a= nd 25x for reading. Their big claim to fame is that Aerospike can write 1M = records per second with only 50 nodes. I wanted to see if this is real.

1M records per second on 50 nodes = is pretty doable by Kudu as well, depending on the size of your records and= the insertion order. I've been playing with a ~70 node cluster recentl= y and seen 1M+ writes/second sustained, and bursting above 4M. These are 1K= B rows with 11 columns, and with pretty old HDD-only nodes. I think newer f= lash-based nodes could do better.
=C2=A0
=

To answer your ques= tions, we have a DMP with user profiles with many attributes. We create seg= mentation information off of these attributes to classify them. Then, we ca= n target advertising appropriately for our sales department. Much of the da= ta processing is for applying models on all or if not most of every profile= =E2=80=99s attributes to find similarities (nearest neighbor/clustering) ov= er a large number of rows when batch processing or a small subset of rows f= or quick online scoring. So, our use case is a typical advanced analytics s= cenario. We have tried HBase, but it doesn=E2=80=99t work well for these ty= pes of analytics.

I read, that Aerospike in the re= lease notes, they did do many improvements for batch and scan operations.

I wonder what your thoughts are for using Kudu for = this.

Sounds like a good Kudu u= se case to me. I've heard great things about Aerospike for the low late= ncy random access portion, but I've also heard that it's _very_ exp= ensive, and not particularly suited to the columnar scan workload. Lastly, = I think the Apache license of Kudu is much more appealing than the AGPL3 us= ed by Aerospike. But, that's not really a direct answer to the performa= nce question :)
=C2=A0

Thanks,
Ben


On May 27, 2016, at 6:21 PM,= Mike Percy <mp= ercy@cloudera.com> wrote:

Have you considered whether = you have a scan heavy or a random access heavy workload? Have you considere= d whether you always access / update a whole row vs only a partial row? Kud= u is a column store so has some awesome=C2=A0performance characteristics wh= en you are doing a lot of scanning of just a couple of=C2=A0columns.
I don't know the answer to your question but if your conce= rn is performance then I would be interested=C2=A0in seeing comparisons fro= m a perf perspective on certain workloads.

Finally, a year ago=C2=A0Aerospike did quite poorly in a Jepsen test:= =C2=A0https://aphyr.com/posts/324-jepsen-aerospike

=
I wonder if they have addressed any of those issues.

Mike

On Friday, May 27, 2016, Benjamin Kim <bbuild11@gmail.com> wrote= :
I am just curious. How will Kudu compare with Aerospi= ke (http://www.aero= spike.com)? I went to a Spark Roadshow and found out about this piece o= f software. It appears to fit our use case perfectly since we are an ad-tec= h company trying to leverage our user profiles data. Plus, it already has a= Spark connector and has a SQL-like client. The tables can be accessed usin= g Spark SQL DataFrames and, also, made into SQL tables for direct use with = Spark SQL ODBC/JDBC Thriftserver. I see from the work done here=C2=A0= http://gerrit.cloudera.org:8080/#/c/2992/=C2=A0that th= e Spark integration is well underway and, from the looks of it lately, almo= st complete. I would prefer to use Kudu since we are already a Cloudera sho= p, and Kudu is easy to deploy and configure using Cloudera Manager. I also = hope that some of Aerospike=E2=80=99s speed optimization techniques can mak= e it into Kudu in the future, if they have not been already thought of or i= ncluded.

Just some thoughts=E2=80=A6

Cheers,
Ben


--=C2=A0
--
Mike Percy
Software Engineer, Cloudera






--=C2=A0<= /span>
Todd Lipcon
Software E= ngineer, Cloudera




--=C2=A0
Todd Lipcon
S= oftware Engineer, Cloudera

--047d7b604b10607dc905354c56b1--