Wednesday Feb 02, 2011
Myths Rumors Fud Hate NoSQL Cassandra vs hbase
So I have fallen many times to trolls and Cassandra and HBase debates. At times I have been that troll. It's time for me to set the record on what I think. This may surprise you.
- Data Model
- Cassandra
- You have a token ring. Each node takes a section of the Ring.
- Cassandra lets you decide between A Random Partitioner (hashing) and and Order Preserving Partitioner in order (by key).
- Fan boys say "Cassandra is more configurable then hbase because I have two partitioning choices"
- Critics say "If you use RandomPartitioner you can not range scan on keys (only the columns of a key)
- Practical people know you can use OOP and hash yourself.
- Hbase
- You have a regions. When regions grow large they split into sub regions.
- You only have the keys in byte order
- Critics say "If you are inserting based on timestamp one region gets overloaded"
- Fan boys say "Randomize your inserts if they are timestamps"
- Practical people know you can hash yourself
- Practical people know hbase has an advantage if you want to range scan on keys
- Practical people know you most like will build your own secondary indexes as range scan on key is not going to give your every SQL feature you can wet dream of.
- Setup
- Cassandra
- Cassandra is very easy to setup and get working out of the box
- Fan boys say "Cassandra is easier"
- Practical people know you are going to have to understand and tune anyway being able to boot up in seconds means not much
- Gurus know: Just bringing up a node and letting it auto bootstrap will likely not be what you want
- I can setup a 5 node cassandra cluster in 10 minutes (working with for <1 year)
- HBase
- Hbase lives on top hadoop which is both a blessing and curse
- Practical people know setting up hadoop is not easy
- It might take me about 2 hours to setup hadoop and hbase (working with almost 1-2 years) This setup includes checking all the various logs and web interfaces.
- Practical people know hbase has many moving parts (nn,2nn,jt (for counts) datanodes task, trackers, zookeepers, hbase, masters
- Practical people know its a bigger stack
- cluster size
- Cassandra
- Cassandra is actually effective at small sizes 1, 3, 5 nodes
- You do not get 'decent' scale out till ~7 nodes
- Practical people say "cool I can grow into this even at a small shop"
- Hbase
- Just too much set up and too many components to make sense at less then ~7 nodes
- Fan boys say "3 nodes ... too puny..." -- Believe it or not 3 computers is a lot to some people.
- Many people are looking to solve big data problems so node number of nodes being greater then 7 is not an issue for them. They have some iron to throw at the problem.
- Scale out
- Cassandra
- Cassandra allows your to move nodes and have new nodes take sections of the ring
- Fan boys say "Awesome"
- Practical people have noticed that you have to keep the cluster balanced
- Practical people know balancing the cluster DOES take administrative work
- Practical people know moving from 11-15 nodes is intensive (compaction, stream, cleanup)
- Practical people see joins sometime fail and have to do it again (annoyed)
- HBase
- Regions split and move quite frequent and automatically
- Fan boys say "awesome"
- Practical people see that they do not always move where you want
- Practical people see that automatic moving and splitting fails sometimes. 1/1000000 it might happens, and you may have to do surgery.
- Release cycle
- Cassandra
- The Cassandra code base moves pretty quickly and is pretty agile.
- Practical people have waiting a while for critical features 'efficient moves'
- There have been some recent releases that were followed very quickly by a bug fix release :(
- Hbase
- The hbase code base moves pretty quickly and is pretty agile.
- practical people get worried by a blurring number of jiras per release
- Practical people need a map to figure out the hadoop hbase version matrix and what works together. hadoop 0.20 append ?
- Consistency
- Cassandra
- Cassandra allows the user to chose consistency models
- Fans say "This is very flexible"
- Critics scoff and say "Eventual consistency" "Quorum READS"
- Hbase
- Offers one consistency model
- Practical people know Region server down data unreachable for time
- Fanboys say "use a caching layer" --( wait I need another component ?)
- Practical people might think this one consistency model is PERFECT for their use case
- Hadoop &
- Hbase
- Hadoop I love you but...
- You can't handle GIG-E correctly
- You can't handle different size drives correctly
- https://issues.apache.org/
jira/browse/HDFS-1601. <--no one noticed this before now? - Does anyone really give a fuck about random read besides my co worker Jay? https://issues.apache.org/jira/browse/HDFS-516 , https://issues.apache.org/jira/browse/HDFS-918 . Just seems like mission impossible to optimize the read or write path, but changing everything for kerberos security (that most people don't even need) is no problem.
- Cassandra
- Does not use hadoop!
- Haters say "Moving data between nodes takes a long time"
- Practical people say "MMAP and direct io support" and none of ^
- API & RPC
- Cassandra
- Thrift and only thrift
- Critics say "Thrift sucks on PHP"
- (I personally think thift is awesome)
- Hbase
- pure java client,
- Critics say rest needs another component
- Critics say thrift needs another component
- Every fucker that writes a half ass "why XXX is better blog"
- You don't know shit after 2 days
- You are going to get shot to shit on either #hbase or #cassandra
- Remember there are other use cases in the world besides yours
- Commercial products that want to hate!
- Don't be mad because open source has a leg up on you and your the imitators!
- Volt-db I see your gmail adds with your volt v cassandra benchmark (hate)
- Hyper Table I see YOUR gmail adds with your hyper table vs hbase benchmark (hate)
- Administration
- Cassandra
- Admins are happy with one log
- Log is under control in terms of events
- Admnistration is more then just a log file and the wiki is skimpy
- Best practices are skimpy too.
- Schema design info is skimpy too.
- hbase & hadoop & zk & ...
- Practical people see ALOT of log files and a lot of cryptic hadoop messages
- Practical people see this actually takes a team of people!
- Facebook messaging & U
- Hbase
- Makes sense. Many nodes!
- Makes sense. Many hadoop/hbase committers!
- Makes sense. Really really really big scale!
- Haters point out 'still uses memcache db'
- Cassandra
- At the time did NOT have online schema updates (does now in 7)
- see section on scale out!
- Try to get them next go around!
Let me cap this by saying both products are awesome and have many awesome people heading them up. I would mind working with either tool for any project. LISTEN carefully....THERE ARE MANY CASES WHERE I WOULD CHOSE ONE OR THE OTHER! I am not going to sit here and tell you XXXX is best.
Posted at 03:56PM Feb 02, 2011 by edwardcapriolo in General | Comments[128]
Posted by Confluence: Focused Crawler on October 29, 2011 at 07:52 AM EDT #
Posted by Confluence: Focused Crawler on October 29, 2011 at 07:55 AM EDT #
Posted by Confluence: Focused Crawler on October 29, 2011 at 07:57 AM EDT #
Posted by Confluence: Focused Crawler on October 29, 2011 at 07:59 AM EDT #
Posted by Confluence: Focused Crawler on November 01, 2011 at 03:49 AM EDT #
Posted by Confluence: Focused Crawler on November 01, 2011 at 04:13 AM EDT #
Posted by Confluence: Focused Crawler on November 01, 2011 at 04:14 AM EDT #
Posted by jaaanp on June 05, 2012 at 12:35 AM EDT #
Posted by Confluence: Kimberly Duffy on June 18, 2012 at 11:02 PM EDT #
Posted by Confluence: Kimberly Duffy on June 19, 2012 at 01:26 PM EDT #
Posted by Confluence: Kimberly Duffy on June 20, 2012 at 12:41 PM EDT #
Posted by Confluence: Kimberly Duffy on June 20, 2012 at 12:52 PM EDT #
Posted by Confluence: Kimberly Duffy on June 20, 2012 at 08:29 PM EDT #
Posted by Confluence: Kimberly Duffy on June 22, 2012 at 02:09 PM EDT #
Posted by prada ?? on April 20, 2013 at 08:53 PM EDT #
Posted by baidu censor on April 21, 2013 at 06:39 AM EDT #
Posted by CheaP Oakley Juliet on May 09, 2013 at 12:54 AM EDT #
Posted by OaklEy Big TacO SunglaSSES on May 09, 2013 at 12:54 AM EDT #
Posted by FakE OaklEy EyEpatcH 2 on May 09, 2013 at 12:54 AM EDT #
Posted by FakE OaklEy EyEpatcH on May 09, 2013 at 12:54 AM EDT #
Posted by FakE OaklEy FuEl CEll on May 09, 2013 at 12:54 AM EDT #
Posted by AXijNhUc on May 09, 2013 at 02:02 AM EDT #
Posted by chanel bags on May 09, 2013 at 04:40 AM EDT #
Posted by seo software on May 09, 2013 at 07:19 AM EDT #
Posted by ???? on May 09, 2013 at 09:57 AM EDT #
Posted by wow gold on May 10, 2013 at 06:48 AM EDT #
Posted by miu miu bags on May 10, 2013 at 06:48 AM EDT #
Posted by lebron 10 on May 10, 2013 at 06:48 AM EDT #
Posted by cheap soccer cleats on May 10, 2013 at 06:48 AM EDT #
Posted by lacoste outlet on May 10, 2013 at 06:48 AM EDT #
Posted by burberry bags on May 10, 2013 at 06:48 AM EDT #
Posted by michael kors bags on May 10, 2013 at 11:27 AM EDT #
Posted by mulberry bags on May 10, 2013 at 11:27 AM EDT #
Posted by hermes bags on May 10, 2013 at 11:28 AM EDT #
Posted by michael kors outlet on May 10, 2013 at 11:28 AM EDT #
Posted by pig on May 10, 2013 at 04:28 PM EDT #
Posted by coach outlet on May 10, 2013 at 04:28 PM EDT #
Posted by prada bags on May 10, 2013 at 04:28 PM EDT #
Posted by nike france on May 10, 2013 at 04:28 PM EDT #
Posted by mulberry outlet on May 11, 2013 at 01:07 AM EDT #
Posted by coach bags on May 11, 2013 at 01:07 AM EDT #
Posted by diablo 3 gold on May 11, 2013 at 01:07 AM EDT #
Posted by mulberry sale on May 11, 2013 at 01:07 AM EDT #
Posted by cheap oakley sunglasses on May 11, 2013 at 01:07 AM EDT #
Posted by chanel handbags on May 11, 2013 at 01:07 AM EDT #
Posted by prada handbags on May 11, 2013 at 10:24 AM EDT #
Posted by celine bags on May 11, 2013 at 10:24 AM EDT #
Posted by Oakley sunglasses sale on May 11, 2013 at 04:58 PM EDT #
Posted by Oakley sunglasses sale on May 11, 2013 at 04:58 PM EDT #
Posted by Oakley Sunglasses Cheap on May 11, 2013 at 04:58 PM EDT #
Posted by Cheap Oakleys on May 11, 2013 at 04:58 PM EDT #
Posted by Oakley sunglasses Outlet on May 12, 2013 at 03:50 AM EDT #
Posted by oaklEys ouTlet on May 12, 2013 at 03:50 AM EDT #
Posted by oakley sunglass for sale on May 12, 2013 at 03:50 AM EDT #
Posted by longchamp bags on May 12, 2013 at 04:21 AM EDT #
Posted by cheap jordans on May 12, 2013 at 04:21 AM EDT #
Posted by louis vuitton bags on May 12, 2013 at 04:21 AM EDT #
Posted by cheap ray ban sunglasses on May 12, 2013 at 04:21 AM EDT #
Posted by cheap ray ban sunglasses on May 12, 2013 at 04:22 AM EDT #
Posted by CHEAP OAKLEY on May 12, 2013 at 12:51 PM EDT #
Posted by Chanel Outlet on May 12, 2013 at 11:14 PM EDT #
Posted by celine outlet on May 13, 2013 at 03:03 AM EDT #
Posted by CELINE OUTLET on May 13, 2013 at 03:03 AM EDT #
Posted by chaussures christian louboutin on May 13, 2013 at 07:19 AM EDT #
Posted by LOUIS VUITTON OUTLET on May 13, 2013 at 07:19 AM EDT #
Posted by hermes replica on May 13, 2013 at 07:19 AM EDT #
Posted by Cheap Louis Vuitton Bags on May 13, 2013 at 07:19 AM EDT #
Posted by air max 90 pas cher on May 13, 2013 at 10:13 PM EDT #
Posted by maillot de foot pas cher on May 13, 2013 at 10:13 PM EDT #
Posted by jeremy scott adidas on May 13, 2013 at 10:34 PM EDT #
Posted by CHEAP PROM DRESSES on May 13, 2013 at 10:34 PM EDT #
Posted by Nike Pas Cher on May 14, 2013 at 03:08 AM EDT #
Posted by louis vuitton replica on May 14, 2013 at 03:09 AM EDT #
Posted by christian louboutin outlet on May 15, 2013 at 05:43 AM EDT #
Posted by cHEap OaklEy SunglaSSES on May 15, 2013 at 07:50 AM EDT #
Posted by CHEap OaklEy FrOgSkinS on May 15, 2013 at 07:50 AM EDT #
Posted by FakE OaklEy SunglaSSES on May 15, 2013 at 07:50 AM EDT #
Posted by FakE oaklEys on May 15, 2013 at 07:51 AM EDT #
Posted by ???? on May 16, 2013 at 04:25 AM EDT #
Posted by IPHONE 5 BLACK on May 17, 2013 at 12:10 PM EDT #
Posted by LOUIS VUITTON EN LIGNE on May 17, 2013 at 12:10 PM EDT #
Posted by Cheap Christian Louboutin on May 17, 2013 at 12:10 PM EDT #
Posted by cheap louis vuitton bags,cheap louis vuitton handbags on May 17, 2013 at 12:10 PM EDT #
Posted by NIKE FREE RUN on May 17, 2013 at 12:10 PM EDT #
Posted by PAS CHER LOUIS VUITTON on May 17, 2013 at 06:51 PM EDT #
Posted by Ray Ban Sunglasses Outlet on May 18, 2013 at 12:13 AM EDT #
Posted by Louis Vuitton Wallet on May 18, 2013 at 12:13 AM EDT #
Posted by jimmy choo shoes on May 18, 2013 at 12:13 AM EDT #
Posted by ?????? on May 18, 2013 at 01:28 PM EDT #
Posted by Christian Louboutin Shoes on May 18, 2013 at 08:42 PM EDT #
Posted by Ralph Lauren Polo on May 19, 2013 at 01:54 AM EDT #
Posted by cheap oakley on May 19, 2013 at 01:54 AM EDT #
Posted by chanel handbags on May 19, 2013 at 07:08 AM EDT #
Posted by ?? on May 19, 2013 at 07:08 AM EDT #
Posted by replica oakleys on May 20, 2013 at 04:19 AM EDT #
Posted by fake oakleys on May 20, 2013 at 04:19 AM EDT #
Posted by Oakley sunglasses sale on May 20, 2013 at 04:19 AM EDT #
Posted by Oakley Outlet on May 20, 2013 at 04:19 AM EDT #
Posted by Cheap Beats By Dre on May 20, 2013 at 10:21 AM EDT #
Posted by CHEAP OAKLEY SUNGLASSES on May 20, 2013 at 10:23 AM EDT #
Posted by google on May 20, 2013 at 11:52 AM EDT #
Posted by monster dr dre on May 20, 2013 at 10:44 PM EDT #
Posted by eEvrsUEV on May 21, 2013 at 02:15 PM EDT #
Posted by lunette oakley pas cher on May 21, 2013 at 09:43 PM EDT #
Posted by FakE OaklEy Flak JackEt on May 22, 2013 at 02:35 AM EDT #
Posted by CHEap OaklEy DiSpatcH II on May 22, 2013 at 02:35 AM EDT #
Posted by Oakley oil Rig Sunglasses on May 22, 2013 at 02:35 AM EDT #
Posted by VIBRAM FIVE FINGERS CLASSIC on May 22, 2013 at 04:33 AM EDT #
Posted by vibram five fingers on May 22, 2013 at 04:33 AM EDT #
Posted by Cheap Ray Ban Sunglasses,Fake Ray Ban Sunglasses,Replica Ray Ban Sunglasses on May 22, 2013 at 04:33 AM EDT #
Posted by vibram five fingers women shoes on May 22, 2013 at 04:33 AM EDT #
Posted by RAYBAN SUNGLASSES on May 22, 2013 at 11:41 AM EDT #
Posted by mulberry sale on May 22, 2013 at 11:41 AM EDT #
Posted by ???????? on May 22, 2013 at 11:41 AM EDT #
Posted by Oakley Sunglasses Outlet on May 23, 2013 at 09:00 AM EDT #
Posted by Cheap Oakley Sunglasses on May 23, 2013 at 09:00 AM EDT #
Posted by Bred 4s on May 23, 2013 at 03:55 PM EDT #
Posted by Cool Grey 9s on May 23, 2013 at 03:55 PM EDT #
Posted by Thunder 4s on May 23, 2013 at 03:55 PM EDT #
Posted by Order Jordans 2013 Online on May 23, 2013 at 03:55 PM EDT #
Posted by Bred 11s on May 23, 2013 at 03:55 PM EDT #
Posted by Fire Red 5s Restock on May 23, 2013 at 03:55 PM EDT #
Posted by Jordan 2013 For Sale on May 23, 2013 at 03:55 PM EDT #
Posted by oakleys outlet on May 24, 2013 at 04:56 AM EDT #
Posted by nike blazer pas cher on May 24, 2013 at 06:09 AM EDT #
Posted by nike blazers low on May 24, 2013 at 06:09 AM EDT #
Posted by cambridge satchel company on May 24, 2013 at 06:09 AM EDT #
Posted by Jordan 11 Low For Sale on May 24, 2013 at 08:58 AM EDT #