• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
Austin cassandra meetup
 

Austin cassandra meetup

on

  • 1,437 views

Slides used for my presentation to the Austin Cassandra Meetup where I discuss how Cassandra fits in to Rackspace Cloud Monitoring.

Slides used for my presentation to the Austin Cassandra Meetup where I discuss how Cassandra fits in to Rackspace Cloud Monitoring.

Hint: It's just a small part.

Statistics

Views

Total Views
1,437
Views on SlideShare
1,432
Embed Views
5

Actions

Likes
0
Downloads
0
Comments
0

4 Embeds 5

http://www.linkedin.com 2
http://tweetedtimes.com 1
https://si0.twimg.com 1
https://twitter.com 1

Accessibility

Upload Details

Uploaded via as Microsoft PowerPoint

Usage Rights

CC Attribution License

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment
  • Self serviceExtensive dashboard12 platformsDeep analysis
  • Secondary, nice to have, but not critical to monitoring.

Austin cassandra meetup Austin cassandra meetup Presentation Transcript

  • How we use CassandraGary Dusbabek@gdusbabek
  • We’re hiringhttp://jobs.rackspace.com/
  • http://www.flickr.com/photos/ollesvensson/4252196844/Goals CM Overview Control Cluster Data Cluster
  • CM OverviewThousands of serversPre-existing solutionsLessons learned from CloudkickInternal versus externalMillions of checks http://www.flickr.com/photos/jean_koulev/2697677595/
  • Cassandra is the leastinteresting part
  • TerminologyEntity Something with an IP address or host name
  • TerminologyCheck Tied to an Entity Is an action Produces metrics
  • Features Remote checksCollectors in 5 DCs Processing in 3 DCsAlerting Notifications
  • All REST all the time More Ops friendlyFeatures Metrics Agent https://github.com/racker/virgo
  • Future Automation Prediction Support hooksAgent expansion Correlation AggregationEntity Spanning
  • 1,000 Words
  • Control ClusterMetadataStateThree datacentersHigh RFWide rowsEasy dump & loadhttps://github.com/racker/cassandra-syncer
  • Data ModelRich but simpleObjects used together stored together Simple parent-child relationsOne row per customer (tenant)Composite column names
  • Data Model Good: Single Parent/Child Acyclic
  • Data Model Bad: Complex Cyclic
  • Data Model As Columns Easy slicing
  • Y U NO MySQL?
  • Control ClusterAPI server is Node.jsJavascript ORM library• Define object model in JS• Read/write entire objects• Never think about CQLnode-cassandra-clienthttps://github.com/racker/node-cassandra-client
  • Control Cluster
  • The fun starts Data here Cluster
  • Data ClusterGoal: Fast graphsTime series dataFewer data pointsOK to shave resolutionRecent data is most important
  • LocatorIdentifies a single metric check identifier + nameE.g. my:check:id:ttfb
  • GranularityFull, 5m, 20m, 60m, 240m, 1440m
  • Rollup ConceptsSlot (Range)• Pegged at 4032 slots• One slot is a range of seconds (varies with granularity)• metrics_locator CF• Key is granularity name + slot num• Columns index keys in rollup tables
  • Keyed by asciiBigint column names Blob column values JDBC Rollups
  • Full Resolution! Arrival• time, name, several metrics• metric = name, type, value• Compute locator and slot• Insert metrics col=timestamp, value=encoded metric• Single Cassandra APPLY BATCH;
  • Rollups• Two types – Rollup all metrics from timeX to timeY – Rollup a single metric from timeX to timeY – Times may span multiple slots (ranges)• Use rollups to produce rollups – E.g.: use 20m data points to create 60m point. – Store number of data points with rollup
  • Rollups• Gotchas! – Do not want to rollup a coarse range when finer range that feeds data to it is scheduled for rollup shortly 60m | | | … 20m | | | | | | | |… . 5m |||||||||||||||||||||||||||||… – Mind the “tail” during datapoint queries (calculate rollups on the fly)
  • It ScalesRollup operations are idempotent* Simplifies availabilityRollups are easily parallelized Hash partition the locator space
  • But…What if data arrives after rollup is performed?More than 24hrs late: don’t care, forget itElse treat normally: slots are scheduled for rollups as they age
  • http://www.flickr.com/photos/nateone/3768979925 / HBase@gdusbabek
  • Image Creditssoccer http://www.flickr.com/photos/ollesvensson/4252196844/apples http://www.flickr.com/photos/jean_koulev/2697677595/ bored http://www.flickr.com/photos/aidanmorgan/3709199304/ clock http://www.flickr.com/photos/szift/3196084839/hallway http://www.flickr.com/photos/orinrobertjohn/4354716077/ car http://www.flickr.com/photos/ellenm1/3541270451control http://www.flickr.com/photos/germanium/117612088/model http://www.flickr.com/photos/mikeschinkel/2703438152/ robot http://www.flickr.com/photos/kb35/430976324/swings http://www.flickr.com/photos/pierrebedat/1095337445/cluster http://www.flickr.com/photos/inl/5097547405/ dish http://www.flickr.com/photos/mrpbps/2862208028/ sand http://www.flickr.com/photos/backkratze/3480338854/ slots http://www.flickr.com/photos/andresrueda/2925383781 sushi http://www.flickr.com/photos/basykes/4348613931/ airplane http://www.flickr.com/photos/fhashemi/72489620/ scale http://www.flickr.com/photos/puuikibeach/4765115333 hourglass http://www.flickr.com/photos/aidanmorgan/2331754875/ street http://www.flickr.com/photos/nateone/3768979925/