Author: zznate
Date: Wed Apr 10 01:16:07 2019
New Revision: 1857225
URL: http://svn.apache.org/viewvc?rev=1857225&view=rev
Log:
CASSANDRA-14765 - Streaming performance post by Sumanth Pasupuleti
Added:
cassandra/site/publish/blog/2019/
cassandra/site/publish/blog/2019/04/
cassandra/site/publish/blog/2019/04/09/
cassandra/site/publish/blog/2019/04/09/benchmarking_streaming.html
cassandra/site/publish/blog/page2/
cassandra/site/publish/blog/page2/index.html
cassandra/site/publish/img/blog-post-benchmarking-streaming/
cassandra/site/publish/img/blog-post-benchmarking-streaming/cassandra_streaming.png
(with props)
cassandra/site/src/img/blog-post-benchmarking-streaming/
cassandra/site/src/img/blog-post-benchmarking-streaming/cassandra_streaming.png
(with props)
Modified:
cassandra/site/publish/blog/index.html
cassandra/site/publish/feed.xml
Added: cassandra/site/publish/blog/2019/04/09/benchmarking_streaming.html
URL:
http://svn.apache.org/viewvc/cassandra/site/publish/blog/2019/04/09/benchmarking_streaming.html?rev=1857225&view=auto
==============================================================================
--- cassandra/site/publish/blog/2019/04/09/benchmarking_streaming.html (added)
+++ cassandra/site/publish/blog/2019/04/09/benchmarking_streaming.html Wed Apr
10 01:16:07 2019
@@ -0,0 +1,232 @@
+<!DOCTYPE html>
+<html>
+
+
+
+
+<head>
+ <meta charset="utf-8">
+ <meta http-equiv="X-UA-Compatible" content="IE=edge">
+ <meta name="viewport" content="width=device-width, initial-scale=1">
+ <meta name="description" content="Streaming is a process where nodes of a
cluster exchange data in the form of SSTables. Streaming can kick in during
many situations such as bootstrap, repair...">
+ <meta name="keywords" content="cassandra, apache, apache cassandra,
distributed storage, key value store, scalability, bigtable, dynamo" />
+ <meta name="robots" content="index,follow" />
+ <meta name="language" content="en" />
+
+ <title>Even Higher Availability with 5x Faster Streaming in Cassandra
4.0</title>
+
+ <link rel="canonical"
href="http://cassandra.apache.org/blog/2019/04/09/benchmarking_streaming.html";>
+
+ <link rel="stylesheet"
href="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.6/css/bootstrap.min.css";
integrity="sha384-1q8mTJOASx8j1Au+a5WDVnPi2lkFfwwEAa8hDDdjZlpLegxhjVME1fgjWPGmkzs7"
crossorigin="anonymous">
+ <link rel="stylesheet" href="./../../../../css/style.css">
+
+
+
+ <link rel="stylesheet"
href="https://use.fontawesome.com/releases/v5.2.0/css/all.css";
integrity="sha384-hWVjflwFxL6sNzntih27bfxkr27PmbbK/iSvJ+a4+0owXq79v+lsFkW54bOGbiDQ"
crossorigin="anonymous">
+
+ <link type="application/atom+xml" rel="alternate"
href="http://cassandra.apache.org/feed.xml"; title="Apache Cassandra Website" />
+</head>
+
+ <body>
+ <!-- breadcrumbs -->
+<div class="topnav">
+ <div class="container breadcrumb-container">
+ <ul class="breadcrumb">
+ <li>
+ <div class="dropdown">
+ <img class="asf-logo" src="./../../../../img/asf_feather.png" />
+ <a data-toggle="dropdown" href="#">Apache Software Foundation <span
class="caret"></span></a>
+ <ul class="dropdown-menu" role="menu" aria-labelledby="dLabel">
+ <li><a href="http://www.apache.org";>Apache Homepage</a></li>
+ <li><a href="http://www.apache.org/licenses/";>License</a></li>
+ <li><a
href="http://www.apache.org/foundation/sponsorship.html";>Sponsorship</a></li>
+ <li><a
href="http://www.apache.org/foundation/thanks.html";>Thanks</a></li>
+ <li><a href="http://www.apache.org/security/";>Security</a></li>
+ </ul>
+ </div>
+ </li>
+
+
+ <li><a href="./../../../../">Apache Cassandra</a></li>
+
+
+
+
+ <li>Even Higher Availability with 5x Faster Streaming in Cassandra
4.0</li>
+
+
+
+
+
+
+ </ul>
+ </div>
+
+ <!-- navbar -->
+ <nav class="navbar navbar-default navbar-static-top" role="navigation">
+ <div class="container">
+ <div class="navbar-header">
+ <button type="button" class="navbar-toggle collapsed"
data-toggle="collapse" data-target="#cassandra-menu" aria-expanded="false">
+ <span class="sr-only">Toggle navigation</span>
+ <span class="icon-bar"></span>
+ <span class="icon-bar"></span>
+ <span class="icon-bar"></span>
+ </button>
+ <a class="navbar-brand" href="./../../../../"><img
src="./../../../../img/cassandra_logo.png" alt="Apache Cassandra logo" /></a>
+ </div><!-- /.navbar-header -->
+
+ <div id="cassandra-menu" class="collapse navbar-collapse">
+ <ul class="nav navbar-nav navbar-right">
+ <li><a href="./../../../../">Home</a></li>
+ <li><a href="./../../../../download/">Download</a></li>
+ <li><a href="./../../../../doc/">Documentation</a></li>
+ <li><a href="./../../../../community/">Community</a></li>
+ <li>
+ <a href="./../../../../blog">Blog</a>
+ </li>
+ </ul>
+ </div><!-- /#cassandra-menu -->
+
+
+ </div>
+ </nav><!-- /.navbar -->
+</div><!-- /.topnav -->
+
+ <div class="content">
+ <div class="container">
+ <h2>Even Higher Availability with 5x Faster Streaming in Cassandra 4.0</h2>
+ <p>Posted on April 09, 2019 by The Apache Cassandra Community</p>
+ <h5><a href="/blog">« Back to the Apache Cassandra Blog</a></h5>
+ <hr />
+ <p>Streaming is a process where nodes of a cluster exchange data in the form
of SSTables. Streaming can kick in during many situations such as bootstrap,
repair, rebuild, range movement, cluster expansion, etc. In this post, we
discuss the massive performance improvements made to the streaming process in
Apache Cassandra 4.0.</p>
+
+<h2 id="high-availability">High Availability</h2>
+<p>As we know Cassandra is a Highly Available, Eventually Consistent database.
The way it maintains its legendary availability is by storing redundant copies
of data in nodes known as replicas, usually running on commodity hardware.
During normal operations, these replicas may end up having hardware issues
causing them to fail. As a result, we need to replace them with new nodes on
fresh hardware.</p>
+
+<p>As part of this replacement operation, the new Cassandra node streams data
from the neighboring nodes that hold copies of the data belonging to this new
nodeâs token range. Depending on the amount of data stored, this process can
require substantial network bandwidth, taking some time to complete. The longer
these types of operations take, the more we are exposing ourselves to loss of
availability. Depending on your replication factor and consistency
requirements, if another node fails during this replacement operation, ability
will be impacted.</p>
+
+<h2 id="increasing-availability">Increasing Availability</h2>
+<p>To minimize the failure window, we want to make these operations as fast as
possible. The faster the new node completes streaming its data, the faster it
can serve traffic, increasing the availability of the cluster. Towards this
goal, Cassandra 4.0 saw the addition of <a
href="https://en.wikipedia.org/wiki/Zero-copy";>Zero Copy</a> streaming. For
more details on Cassandraâs zero copy implementation, see this <a
href="../../../2018/08/06/faster_streaming_in_cassandra.html">blog post</a> and
<a
href="https://issues.apache.org/jira/browse/CASSANDRA-14556";>CASSANDRA-14556</a>
for more information.</p>
+
+<h2 id="talking-numbers">Talking Numbers</h2>
+<p>To quantify the results of these improvements, we, at Netflix, measured the
performance impact of streaming in 4.0 vs 3.0, using our open source <a
href="https://github.com/Netflix/ndbench";>NDBench</a> benchmarking tool with
the CassJavaDriverGeneric plugin. Though we knew there would be improvements,
we were still amazed with the overall results of a <strong>five fold
increase</strong> in streaming performance. The test setup and operations are
all detailed below.</p>
+
+<h3 id="test-setup">Test Setup</h3>
+<p>In our test setup, we used the following configurations:</p>
+<ul>
+ <li>6-node clusters on i3.xl, i3.2xl, i3.4xl and i3.8xl EC2 instances, each
on 3.0 and trunk (sha dd7ec5a2d6736b26d3c5f137388f2d0028df7a03).</li>
+ <li>Table schema</li>
+</ul>
+<div><pre>
+CREATE TABLE testing.test (
+ key text,
+ column1 int,
+ value text,
+ PRIMARY KEY (key, column1)
+) WITH CLUSTERING ORDER BY (column1 ASC)
+ AND bloom_filter_fp_chance = 0.01
+ AND caching = {'keys': 'ALL', 'rows_per_partition': 'NONE'}
+ AND comment = ''
+ AND compaction = {'class':
'org.apache.cassandra.db.compaction.LeveledCompactionStrategy'}
+ AND compression = {'enabled': 'false'}
+ AND crc_check_chance = 1.0
+ AND dclocal_read_repair_chance = 0.1
+ AND default_time_to_live = 0
+ AND gc_grace_seconds = 864000
+ AND max_index_interval = 2048
+ AND memtable_flush_period_in_ms = 0
+ AND min_index_interval = 128
+ AND read_repair_chance = 0.0
+ AND speculative_retry = '99PERCENTILE';
+</pre></div>
+
+<ul>
+ <li>Data size per node: 500GB</li>
+ <li>No. of tokens per node: 1 (no vnodes)</li>
+</ul>
+
+<p>To trigger the streaming process we used the following steps in each of the
clusters:</p>
+<ul>
+ <li>terminated a node</li>
+ <li>add a new node as a replacement</li>
+ <li>measure the time taken to complete streaming data by the new node
replacing the terminated node</li>
+</ul>
+
+<p>For each cluster and version, we repeated this exercise multiple times to
collect several samples.</p>
+
+<p>Below is the distribution of streaming times we found across the clusters
+<img src="/img/blog-post-benchmarking-streaming/cassandra_streaming.png"
alt="Benchmark results" title="Benchmark results" /></p>
+
+<h3 id="interpreting-the-results">Interpreting the Results</h3>
+<p>Based on the graph above, there are many conclusions one can draw from it.
Some of them are</p>
+<ul>
+ <li>3.0 streaming times are inconsistent and show high degree of variability
(fat distributions across multiple samples)</li>
+ <li>3.0 streaming is highly affected by the instance type and generally
looks generally CPU bound</li>
+ <li>Zero Copy streaming is approximately 5x faster</li>
+ <li>Zero Copy streaming time shows little variability in its performance
(thin distributions across multiple samples)</li>
+ <li>Zero Copy streaming performance is not CPU bound and remains consistent
across instance types</li>
+</ul>
+
+<p>It is clear from the performance test results that Zero Copy Streaming has
a huge performance benefit over the current streaming infrastructure in
Cassandra. But what does it mean in the real world? The following key points
are the main take aways.</p>
+
+<p><strong>MTTR (Mean Time to Recovery):</strong> MTTR is a KPI (Key
Performance Indicator) that is used to measure how quickly a system recovers
from a failure. Zero Copy Streaming has a very direct impact here with a
<strong>five fold improvement</strong> on performance.</p>
+
+<p><strong>Costs:</strong> Zero Copy Streaming is ~5x faster. This translates
directly into cost for some organizations primarily as a result of reducing the
need to maintain spare server or cloud capacity. In other situations where
youâre migrating data to larger instance types or moving AZs or DCs, this
means that instances that are sending data can be turned off sooner saving
costs. An added cost benefit is that now you donât have to over provision the
instance. You get a similar streaming performance whether you use a i3.xl or an
i3.8xl provided the bandwidth is available to the instance.</p>
+
+<p><strong>Risk Reduction:</strong> There is a great reduction in the risk due
to Zero Copy Streaming as well. Since a Clusterâs recovery mainly depends on
the streaming speed, Cassandra clusters with failed nodes will be able to
recover much more quickly (5x faster). This means the window of vulnerability
is reduced significantly, in some situations down to few minutes.</p>
+
+<p>Finally, a benefit that we generally donât talk about is the
environmental benefit of this change. Zero Copy Streaming enables us to move
data very quickly through the cluster. It objectively reduces the number and
sizes of instances that are used to build Cassandra cluster. As a result not
only does it reduce Cassandraâs TCO (Total Cost of Ownership), it also helps
the environment by consuming fewer resources!</p>
+
+ </div>
+</div>
+
+ <hr />
+
+<footer>
+ <div class="container">
+ <div class="col-md-4 social-blk">
+ <span class="social">
+ <a href="https://twitter.com/cassandra";
+ class="twitter-follow-button"
+ data-show-count="false" data-size="large">Follow @cassandra</a>
+ <script>!function(d,s,id){var
js,fjs=d.getElementsByTagName(s)[0],p=/^http:/.test(d.location)?'http':'https';if(!d.getElementById(id)){js=d.createElement(s);js.id=id;js.src=p+'://platform.twitter.com/widgets.js';fjs.parentNode.insertBefore(js,fjs);}}(document,
'script', 'twitter-wjs');</script>
+ <a href="https://twitter.com/intent/tweet?button_hashtag=cassandra";
+ class="twitter-hashtag-button"
+ data-size="large"
+ data-related="ApacheCassandra">Tweet #cassandra</a>
+ <script>!function(d,s,id){var
js,fjs=d.getElementsByTagName(s)[0],p=/^http:/.test(d.location)?'http':'https';if(!d.getElementById(id)){js=d.createElement(s);js.id=id;js.src=p+'://platform.twitter.com/widgets.js';fjs.parentNode.insertBefore(js,fjs);}}(document,
'script', 'twitter-wjs');</script>
+
+ </span>
+ <a class="subscribe-rss icon-link" href="/feed.xml" title="Subscribe to
Blog via RSS">
+ <span><i class="fa fa-rss"></i></span>
+ </a>
+ </div>
+
+ <div class="col-md-8 trademark">
+ <p>© 2016 <a href="http://apache.org";>The Apache Software
Foundation</a>.
+ Apache, the Apache feather logo, and Apache Cassandra are trademarks of
The Apache Software Foundation.
+ <p>
+ </div>
+ </div><!-- /.container -->
+</footer>
+
+<!-- Javascript. Placed here so pages load faster -->
+<script
src="https://ajax.googleapis.com/ajax/libs/jquery/1.11.3/jquery.min.js";></script>
+<script src="./../../../../js/underscore-min.js"></script>
+<script
src="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.6/js/bootstrap.min.js";
integrity="sha384-0mSbJDEHialfmuBBQP6A4Qrprq5OVfW37PRR3j5ELqxss1yVqOtnepnHVP9aJ7xS"
crossorigin="anonymous"></script>
+
+
+
+<script type="text/javascript">
+ var gaJsHost = (("https:" == document.location.protocol) ? "https://ssl."; :
"http://www.";);
+ document.write(unescape("%3Cscript src='" + gaJsHost +
"google-analytics.com/ga.js' type='text/javascript'%3E%3C/script%3E"));
+
+ try {
+ var pageTracker = _gat._getTracker("UA-11583863-1");
+ pageTracker._trackPageview();
+ } catch(err) {}
+</script>
+
+
+ </body>
+</html>
Modified: cassandra/site/publish/blog/index.html
URL:
http://svn.apache.org/viewvc/cassandra/site/publish/blog/index.html?rev=1857225&r1=1857224&r2=1857225&view=diff
==============================================================================
--- cassandra/site/publish/blog/index.html (original)
+++ cassandra/site/publish/blog/index.html Wed Apr 10 01:16:07 2019
@@ -102,6 +102,15 @@
<ul class="blog-post-listing">
<li class="blog-post">
+ <h4><a href="/blog/2019/04/09/benchmarking_streaming.html">Even
Higher Availability with 5x Faster Streaming in Cassandra 4.0</a></h4>
+ <p>Posted on April 09, 2019 by The Apache Cassandra Community</p>
+ <p>Streaming is a process where nodes of a cluster exchange data in
the form of SSTables. Streaming can kick in during many situations such as
bootstrap, repair, rebuild, range movement, cluster expansion, etc. In this
post, we discuss the massive performance improvements made to the streaming
process in Apache Cassandra 4.0.</p>
+
+
+ <h5><a href="/blog/2019/04/09/benchmarking_streaming.html">Read more
»</a></h5>
+ </li>
+
+ <li class="blog-post">
<h4><a
href="/blog/2018/12/03/introducing-transient-replication.html">Introducing
Transient Replication</a></h4>
<p>Posted on December 03, 2018 by The Apache Cassandra Community</p>
<p>Transient Replication is a new experimental feature soon to be
available in 4.0. When enabled, it allows for the creation of keyspaces where
replication factor can be specified as a number of copies (full replicas) and
temporary copies (transient replicas). Transient replicas retain the data they
replicate only long enough for it to be propagated to full replicas, via
incremental repair, at which point the data is deleted. Writing to transient
replicas can be avoided almost entirely if monotonic reads are not required
because it is possible to achieve a quorum of acknowledged writes without
them.</p>
@@ -140,17 +149,17 @@ to ensure compliance with regulatory, se
<h5><a href="/blog/2018/08/21/testing_apache_cassandra.html">Read
more »</a></h5>
</li>
- <li class="blog-post">
- <h4><a
href="/blog/2018/08/07/faster_streaming_in_cassandra.html">Hardware-bound Zero
Copy Streaming in Apache Cassandra 4.0</a></h4>
- <p>Posted on August 07, 2018 by The Apache Cassandra Community</p>
- <p>Streaming in Apache Cassandra powers host replacement, range
movements, and cluster expansions. Streaming plays a crucial role in the
cluster and as such its performance is key to not only the speed of the
operations its used in but the clusterâs health generally. In Apache
Cassandra 4.0, we have introduced an improved streaming implementation that
reduces GC pressure and increases throughput several folds and are now limited,
in some cases, only by the disk / network IO (See: <a
href="https://issues.apache.org/jira/browse/CASSANDRA-14556";>CASSANDRA-14556</a>).</p>
-
-
- <h5><a
href="/blog/2018/08/07/faster_streaming_in_cassandra.html">Read more
»</a></h5>
- </li>
-
</ul>
+
+ <ul class="pager">
+
+
+ <li class="next">
+ <a href="/blog/page2/">Older Posts →</a>
+ </li>
+
+ </ul>
</div>
</div>
Added: cassandra/site/publish/blog/page2/index.html
URL:
http://svn.apache.org/viewvc/cassandra/site/publish/blog/page2/index.html?rev=1857225&view=auto
==============================================================================
--- cassandra/site/publish/blog/page2/index.html (added)
+++ cassandra/site/publish/blog/page2/index.html Wed Apr 10 01:16:07 2019
@@ -0,0 +1,177 @@
+<!DOCTYPE html>
+<html>
+
+
+
+
+<head>
+ <meta charset="utf-8">
+ <meta http-equiv="X-UA-Compatible" content="IE=edge">
+ <meta name="viewport" content="width=device-width, initial-scale=1">
+ <meta name="description" content="The Apache Cassandra database is the right
choice when you need scalability and high availability without compromising
performance. Linear scalability and proven fault-tolerance on commodity
hardware or cloud infrastructure make it the perfect platform for
mission-critical data. Cassandra's support for replicating across multiple
datacenters is best-in-class, providing lower latency for your users and the
peace of mind of knowing that you can survive regional outages.
+">
+ <meta name="keywords" content="cassandra, apache, apache cassandra,
distributed storage, key value store, scalability, bigtable, dynamo" />
+ <meta name="robots" content="index,follow" />
+ <meta name="language" content="en" />
+
+ <title> - page 2</title>
+
+ <link rel="canonical" href="http://cassandra.apache.org/blog/page2/";>
+
+ <link rel="stylesheet"
href="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.6/css/bootstrap.min.css";
integrity="sha384-1q8mTJOASx8j1Au+a5WDVnPi2lkFfwwEAa8hDDdjZlpLegxhjVME1fgjWPGmkzs7"
crossorigin="anonymous">
+ <link rel="stylesheet" href="./../../css/style.css">
+
+
+
+ <link rel="stylesheet"
href="https://use.fontawesome.com/releases/v5.2.0/css/all.css";
integrity="sha384-hWVjflwFxL6sNzntih27bfxkr27PmbbK/iSvJ+a4+0owXq79v+lsFkW54bOGbiDQ"
crossorigin="anonymous">
+
+ <link type="application/atom+xml" rel="alternate"
href="http://cassandra.apache.org/feed.xml"; title="Apache Cassandra Website" />
+</head>
+
+ <body>
+ <!-- breadcrumbs -->
+<div class="topnav">
+ <div class="container breadcrumb-container">
+ <ul class="breadcrumb">
+ <li>
+ <div class="dropdown">
+ <img class="asf-logo" src="./../../img/asf_feather.png" />
+ <a data-toggle="dropdown" href="#">Apache Software Foundation <span
class="caret"></span></a>
+ <ul class="dropdown-menu" role="menu" aria-labelledby="dLabel">
+ <li><a href="http://www.apache.org";>Apache Homepage</a></li>
+ <li><a href="http://www.apache.org/licenses/";>License</a></li>
+ <li><a
href="http://www.apache.org/foundation/sponsorship.html";>Sponsorship</a></li>
+ <li><a
href="http://www.apache.org/foundation/thanks.html";>Thanks</a></li>
+ <li><a href="http://www.apache.org/security/";>Security</a></li>
+ </ul>
+ </div>
+ </li>
+
+
+ <li><a href="./../../">Apache Cassandra</a></li>
+
+
+
+
+ <li> - page 2</li>
+
+
+
+
+
+
+ </ul>
+ </div>
+
+ <!-- navbar -->
+ <nav class="navbar navbar-default navbar-static-top" role="navigation">
+ <div class="container">
+ <div class="navbar-header">
+ <button type="button" class="navbar-toggle collapsed"
data-toggle="collapse" data-target="#cassandra-menu" aria-expanded="false">
+ <span class="sr-only">Toggle navigation</span>
+ <span class="icon-bar"></span>
+ <span class="icon-bar"></span>
+ <span class="icon-bar"></span>
+ </button>
+ <a class="navbar-brand" href="./../../"><img
src="./../../img/cassandra_logo.png" alt="Apache Cassandra logo" /></a>
+ </div><!-- /.navbar-header -->
+
+ <div id="cassandra-menu" class="collapse navbar-collapse">
+ <ul class="nav navbar-nav navbar-right">
+ <li><a href="./../../">Home</a></li>
+ <li><a href="./../../download/">Download</a></li>
+ <li><a href="./../../doc/">Documentation</a></li>
+ <li><a href="./../../community/">Community</a></li>
+ <li>
+ <a href="./../../blog">Blog</a>
+ </li>
+ </ul>
+ </div><!-- /#cassandra-menu -->
+
+
+ </div>
+ </nav><!-- /.navbar -->
+</div><!-- /.topnav -->
+
+ <div class="content">
+ <div class="container">
+ <h2>Apache Cassandra Blog</h2>
+<p>Have something to share with the community? Let us know on the <a
href="http://cassandra.apache.org/community/#mailing";>mailing list</a>!</p>
+
+
+ <ul class="blog-post-listing">
+
+ <li class="blog-post">
+ <h4><a
href="/blog/2018/08/07/faster_streaming_in_cassandra.html">Hardware-bound Zero
Copy Streaming in Apache Cassandra 4.0</a></h4>
+ <p>Posted on August 07, 2018 by The Apache Cassandra Community</p>
+ <p>Streaming in Apache Cassandra powers host replacement, range
movements, and cluster expansions. Streaming plays a crucial role in the
cluster and as such its performance is key to not only the speed of the
operations its used in but the clusterâs health generally. In Apache
Cassandra 4.0, we have introduced an improved streaming implementation that
reduces GC pressure and increases throughput several folds and are now limited,
in some cases, only by the disk / network IO (See: <a
href="https://issues.apache.org/jira/browse/CASSANDRA-14556";>CASSANDRA-14556</a>).</p>
+
+
+ <h5><a
href="/blog/2018/08/07/faster_streaming_in_cassandra.html">Read more
»</a></h5>
+ </li>
+
+ </ul>
+
+
+ <ul class="pager">
+
+ <li class="previous">
+ <a href="/blog/">← Newer Posts</a>
+ </li>
+
+
+ </ul>
+
+ </div>
+</div>
+
+ <hr />
+
+<footer>
+ <div class="container">
+ <div class="col-md-4 social-blk">
+ <span class="social">
+ <a href="https://twitter.com/cassandra";
+ class="twitter-follow-button"
+ data-show-count="false" data-size="large">Follow @cassandra</a>
+ <script>!function(d,s,id){var
js,fjs=d.getElementsByTagName(s)[0],p=/^http:/.test(d.location)?'http':'https';if(!d.getElementById(id)){js=d.createElement(s);js.id=id;js.src=p+'://platform.twitter.com/widgets.js';fjs.parentNode.insertBefore(js,fjs);}}(document,
'script', 'twitter-wjs');</script>
+ <a href="https://twitter.com/intent/tweet?button_hashtag=cassandra";
+ class="twitter-hashtag-button"
+ data-size="large"
+ data-related="ApacheCassandra">Tweet #cassandra</a>
+ <script>!function(d,s,id){var
js,fjs=d.getElementsByTagName(s)[0],p=/^http:/.test(d.location)?'http':'https';if(!d.getElementById(id)){js=d.createElement(s);js.id=id;js.src=p+'://platform.twitter.com/widgets.js';fjs.parentNode.insertBefore(js,fjs);}}(document,
'script', 'twitter-wjs');</script>
+
+ </span>
+ <a class="subscribe-rss icon-link" href="/feed.xml" title="Subscribe to
Blog via RSS">
+ <span><i class="fa fa-rss"></i></span>
+ </a>
+ </div>
+
+ <div class="col-md-8 trademark">
+ <p>© 2016 <a href="http://apache.org";>The Apache Software
Foundation</a>.
+ Apache, the Apache feather logo, and Apache Cassandra are trademarks of
The Apache Software Foundation.
+ <p>
+ </div>
+ </div><!-- /.container -->
+</footer>
+
+<!-- Javascript. Placed here so pages load faster -->
+<script
src="https://ajax.googleapis.com/ajax/libs/jquery/1.11.3/jquery.min.js";></script>
+<script src="./../../js/underscore-min.js"></script>
+<script
src="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.6/js/bootstrap.min.js";
integrity="sha384-0mSbJDEHialfmuBBQP6A4Qrprq5OVfW37PRR3j5ELqxss1yVqOtnepnHVP9aJ7xS"
crossorigin="anonymous"></script>
+
+
+
+<script type="text/javascript">
+ var gaJsHost = (("https:" == document.location.protocol) ? "https://ssl."; :
"http://www.";);
+ document.write(unescape("%3Cscript src='" + gaJsHost +
"google-analytics.com/ga.js' type='text/javascript'%3E%3C/script%3E"));
+
+ try {
+ var pageTracker = _gat._getTracker("UA-11583863-1");
+ pageTracker._trackPageview();
+ } catch(err) {}
+</script>
+
+
+ </body>
+</html>
Modified: cassandra/site/publish/feed.xml
URL:
http://svn.apache.org/viewvc/cassandra/site/publish/feed.xml?rev=1857225&r1=1857224&r2=1857225&view=diff
==============================================================================
--- cassandra/site/publish/feed.xml (original)
+++ cassandra/site/publish/feed.xml Wed Apr 10 01:16:07 2019
@@ -1,5 +1,82 @@
-<?xml version="1.0" encoding="utf-8"?><feed
xmlns="http://www.w3.org/2005/Atom"; ><generator uri="https://jekyllrb.com/";
version="3.4.3">Jekyll</generator><link
href="http://cassandra.apache.org/feed.xml"; rel="self"
type="application/atom+xml" /><link href="http://cassandra.apache.org/";
rel="alternate" type="text/html"
/><updated>2019-03-13T12:32:18+13:00</updated><id>http://cassandra.apache.org/</id><title
type="html">Apache Cassandra Website</title><subtitle>The Apache Cassandra
database is the right choice when you need scalability and high availability
without compromising performance. Linear scalability and proven fault-tolerance
on commodity hardware or cloud infrastructure make it the perfect platform for
mission-critical data. Cassandra's support for replicating across multiple
datacenters is best-in-class, providing lower latency for your users and the
peace of mind of knowing that you can survive regional outages.
-</subtitle><entry><title type="html">Introducing Transient
Replication</title><link
href="http://cassandra.apache.org/blog/2018/12/03/introducing-transient-replication.html";
rel="alternate" type="text/html" title="Introducing Transient Replication"
/><published>2018-12-03T21:00:00+13:00</published><updated>2018-12-03T21:00:00+13:00</updated><id>http://cassandra.apache.org/blog/2018/12/03/introducing-transient-replication</id><content
type="html"
xml:base="http://cassandra.apache.org/blog/2018/12/03/introducing-transient-replication.html";><p>Transient
Replication is a new experimental feature soon to be available in 4.0. When
enabled, it allows for the creation of keyspaces where replication factor can
be specified as a number of copies (full replicas) and temporary copies
(transient replicas). Transient replicas retain the data they replicate only
long enough for it to be propagated to full replicas, via incremental repair,
at which point the data is deleted. Writing to transi
ent replicas can be avoided almost entirely if monotonic reads are not
required because it is possible to achieve a quorum of acknowledged writes
without them.</p>
+<?xml version="1.0" encoding="utf-8"?><feed
xmlns="http://www.w3.org/2005/Atom"; ><generator uri="https://jekyllrb.com/";
version="3.4.3">Jekyll</generator><link
href="http://cassandra.apache.org/feed.xml"; rel="self"
type="application/atom+xml" /><link href="http://cassandra.apache.org/";
rel="alternate" type="text/html"
/><updated>2019-04-10T13:13:00+12:00</updated><id>http://cassandra.apache.org/</id><title
type="html">Apache Cassandra Website</title><subtitle>The Apache Cassandra
database is the right choice when you need scalability and high availability
without compromising performance. Linear scalability and proven fault-tolerance
on commodity hardware or cloud infrastructure make it the perfect platform for
mission-critical data. Cassandra's support for replicating across multiple
datacenters is best-in-class, providing lower latency for your users and the
peace of mind of knowing that you can survive regional outages.
+</subtitle><entry><title type="html">Even Higher Availability with 5x Faster
Streaming in Cassandra 4.0</title><link
href="http://cassandra.apache.org/blog/2019/04/09/benchmarking_streaming.html";
rel="alternate" type="text/html" title="Even Higher Availability with 5x Faster
Streaming in Cassandra 4.0"
/><published>2019-04-09T20:00:00+12:00</published><updated>2019-04-09T20:00:00+12:00</updated><id>http://cassandra.apache.org/blog/2019/04/09/benchmarking_streaming</id><content
type="html"
xml:base="http://cassandra.apache.org/blog/2019/04/09/benchmarking_streaming.html";><p>Streaming
is a process where nodes of a cluster exchange data in the form of SSTables.
Streaming can kick in during many situations such as bootstrap, repair,
rebuild, range movement, cluster expansion, etc. In this post, we discuss the
massive performance improvements made to the streaming process in Apache
Cassandra 4.0.</p>
+
+<h2 id="high-availability">High Availability</h2>
+<p>As we know Cassandra is a Highly Available, Eventually Consistent
database. The way it maintains its legendary availability is by storing
redundant copies of data in nodes known as replicas, usually running on
commodity hardware. During normal operations, these replicas may end up having
hardware issues causing them to fail. As a result, we need to replace them with
new nodes on fresh hardware.</p>
+
+<p>As part of this replacement operation, the new Cassandra node streams
data from the neighboring nodes that hold copies of the data belonging to this
new nodeâs token range. Depending on the amount of data stored, this process
can require substantial network bandwidth, taking some time to complete. The
longer these types of operations take, the more we are exposing ourselves to
loss of availability. Depending on your replication factor and consistency
requirements, if another node fails during this replacement operation, ability
will be impacted.</p>
+
+<h2 id="increasing-availability">Increasing
Availability</h2>
+<p>To minimize the failure window, we want to make these operations as
fast as possible. The faster the new node completes streaming its data, the
faster it can serve traffic, increasing the availability of the cluster.
Towards this goal, Cassandra 4.0 saw the addition of <a
href="https://en.wikipedia.org/wiki/Zero-copy">Zero Copy</a>
streaming. For more details on Cassandraâs zero copy implementation, see this
<a
href="../../../2018/08/06/faster_streaming_in_cassandra.html">blog
post</a> and <a
href="https://issues.apache.org/jira/browse/CASSANDRA-14556">CASSANDRA-14556</a>
for more information.</p>
+
+<h2 id="talking-numbers">Talking Numbers</h2>
+<p>To quantify the results of these improvements, we, at Netflix,
measured the performance impact of streaming in 4.0 vs 3.0, using our open
source <a
href="https://github.com/Netflix/ndbench">NDBench</a>
benchmarking tool with the CassJavaDriverGeneric plugin. Though we knew there
would be improvements, we were still amazed with the overall results of a
<strong>five fold increase</strong> in streaming performance. The
test setup and operations are all detailed below.</p>
+
+<h3 id="test-setup">Test Setup</h3>
+<p>In our test setup, we used the following configurations:</p>
+<ul>
+ <li>6-node clusters on i3.xl, i3.2xl, i3.4xl and i3.8xl EC2 instances,
each on 3.0 and trunk (sha dd7ec5a2d6736b26d3c5f137388f2d0028df7a03).</li>
+ <li>Table schema</li>
+</ul>
+<div><pre>
+CREATE TABLE testing.test (
+ key text,
+ column1 int,
+ value text,
+ PRIMARY KEY (key, column1)
+) WITH CLUSTERING ORDER BY (column1 ASC)
+ AND bloom_filter_fp_chance = 0.01
+ AND caching = {'keys': 'ALL', 'rows_per_partition': 'NONE'}
+ AND comment = ''
+ AND compaction = {'class':
'org.apache.cassandra.db.compaction.LeveledCompactionStrategy'}
+ AND compression = {'enabled': 'false'}
+ AND crc_check_chance = 1.0
+ AND dclocal_read_repair_chance = 0.1
+ AND default_time_to_live = 0
+ AND gc_grace_seconds = 864000
+ AND max_index_interval = 2048
+ AND memtable_flush_period_in_ms = 0
+ AND min_index_interval = 128
+ AND read_repair_chance = 0.0
+ AND speculative_retry = '99PERCENTILE';
+</pre></div>
+
+<ul>
+ <li>Data size per node: 500GB</li>
+ <li>No. of tokens per node: 1 (no vnodes)</li>
+</ul>
+
+<p>To trigger the streaming process we used the following steps in each
of the clusters:</p>
+<ul>
+ <li>terminated a node</li>
+ <li>add a new node as a replacement</li>
+ <li>measure the time taken to complete streaming data by the new node
replacing the terminated node</li>
+</ul>
+
+<p>For each cluster and version, we repeated this exercise multiple
times to collect several samples.</p>
+
+<p>Below is the distribution of streaming times we found across the
clusters
+<img
src="/img/blog-post-benchmarking-streaming/cassandra_streaming.png"
alt="Benchmark results" title="Benchmark results"
/></p>
+
+<h3 id="interpreting-the-results">Interpreting the
Results</h3>
+<p>Based on the graph above, there are many conclusions one can draw
from it. Some of them are</p>
+<ul>
+ <li>3.0 streaming times are inconsistent and show high degree of
variability (fat distributions across multiple samples)</li>
+ <li>3.0 streaming is highly affected by the instance type and
generally looks generally CPU bound</li>
+ <li>Zero Copy streaming is approximately 5x faster</li>
+ <li>Zero Copy streaming time shows little variability in its
performance (thin distributions across multiple samples)</li>
+ <li>Zero Copy streaming performance is not CPU bound and remains
consistent across instance types</li>
+</ul>
+
+<p>It is clear from the performance test results that Zero Copy
Streaming has a huge performance benefit over the current streaming
infrastructure in Cassandra. But what does it mean in the real world? The
following key points are the main take aways.</p>
+
+<p><strong>MTTR (Mean Time to Recovery):</strong> MTTR is a
KPI (Key Performance Indicator) that is used to measure how quickly a system
recovers from a failure. Zero Copy Streaming has a very direct impact here with
a <strong>five fold improvement</strong> on performance.</p>
+
+<p><strong>Costs:</strong> Zero Copy Streaming is ~5x
faster. This translates directly into cost for some organizations primarily as
a result of reducing the need to maintain spare server or cloud capacity. In
other situations where youâre migrating data to larger instance types or
moving AZs or DCs, this means that instances that are sending data can be
turned off sooner saving costs. An added cost benefit is that now you donât
have to over provision the instance. You get a similar streaming performance
whether you use a i3.xl or an i3.8xl provided the bandwidth is available to the
instance.</p>
+
+<p><strong>Risk Reduction:</strong> There is a great
reduction in the risk due to Zero Copy Streaming as well. Since a Clusterâs
recovery mainly depends on the streaming speed, Cassandra clusters with failed
nodes will be able to recover much more quickly (5x faster). This means the
window of vulnerability is reduced significantly, in some situations down to
few minutes.</p>
+
+<p>Finally, a benefit that we generally donât talk about is the
environmental benefit of this change. Zero Copy Streaming enables us to move
data very quickly through the cluster. It objectively reduces the number and
sizes of instances that are used to build Cassandra cluster. As a result not
only does it reduce Cassandraâs TCO (Total Cost of Ownership), it also helps
the environment by consuming fewer
resources!</p></content><author><name>The Apache Cassandra
Community</name></author><summary type="html">Streaming is a process where
nodes of a cluster exchange data in the form of SSTables. Streaming can kick in
during many situations such as bootstrap, repair, rebuild, range movement,
cluster expansion, etc. In this post, we discuss the massive performance
improvements made to the streaming process in Apache Cassandra
4.0.</summary></entry><entry><title type="html">Introducing Transient
Replication</title><link href="http://cassandra.apache.org/blog/2018/12/0
3/introducing-transient-replication.html" rel="alternate" type="text/html"
title="Introducing Transient Replication"
/><published>2018-12-03T21:00:00+13:00</published><updated>2018-12-03T21:00:00+13:00</updated><id>http://cassandra.apache.org/blog/2018/12/03/introducing-transient-replication</id><content
type="html"
xml:base="http://cassandra.apache.org/blog/2018/12/03/introducing-transient-replication.html";><p>Transient
Replication is a new experimental feature soon to be available in 4.0. When
enabled, it allows for the creation of keyspaces where replication factor can
be specified as a number of copies (full replicas) and temporary copies
(transient replicas). Transient replicas retain the data they replicate only
long enough for it to be propagated to full replicas, via incremental repair,
at which point the data is deleted. Writing to transient replicas can be
avoided almost entirely if monotonic reads are not required because it is
possible to achieve a quorum of acknow
ledged writes without them.</p>
<p>This results in a savings in disk space, CPU, and IO. By deleting
data as soon as it is no longer needed, transient replicas require only a
fraction of the disk space of a full replica. By not having to store the data
indefinitely, the CPU and IO required for compaction is reduced, and read
queries are faster as they have less data to process.</p>
Added:
cassandra/site/publish/img/blog-post-benchmarking-streaming/cassandra_streaming.png
URL:
http://svn.apache.org/viewvc/cassandra/site/publish/img/blog-post-benchmarking-streaming/cassandra_streaming.png?rev=1857225&view=auto
==============================================================================
Binary file - no diff available.
Propchange:
cassandra/site/publish/img/blog-post-benchmarking-streaming/cassandra_streaming.png
------------------------------------------------------------------------------
svn:mime-type = application/octet-stream
Added:
cassandra/site/src/img/blog-post-benchmarking-streaming/cassandra_streaming.png
URL:
http://svn.apache.org/viewvc/cassandra/site/src/img/blog-post-benchmarking-streaming/cassandra_streaming.png?rev=1857225&view=auto
==============================================================================
Binary file - no diff available.
Propchange:
cassandra/site/src/img/blog-post-benchmarking-streaming/cassandra_streaming.png
------------------------------------------------------------------------------
svn:mime-type = application/octet-stream
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
