Samba on Scale: Difference between revisions

From SambaWiki
 
(39 intermediate revisions by the same user not shown)
Line 5: Line 5:
{| class="wikitable sortable"
{| class="wikitable sortable"
|-
|-
! Name !! Consistency Model || Transactions || C Client || Notes
! Name !! Type || Consistency Model || Transactions || C Client || Notes
|-
|-
| [https://wiki.samba.org/index.php/CTDB_and_Clustered_Samba ctdb] || Volatile DBs: [https://jepsen.io/consistency/models/sequential Basically Sequential Consistency]<br />Persistent DBs: [https://jepsen.io/consistency/models/linearizable Strict Serializability] || No<br/>Yes || yes ||
| [https://wiki.samba.org/index.php/CTDB_and_Clustered_Samba ctdb] || K/V || Volatile DBs: [https://arxiv.org/pdf/1512.00168.pdf Prefix-Linearizable?]<br />Persistent DBs: [https://jepsen.io/consistency/models/linearizable Strict Serializability] || No<br/>Yes || yes ||
|-
|-
| [https://etcd.io/ etcd] || [https://etcd.io/docs/v3.3/learning/api_guarantees/ Linearizability] || LWT ("mini") || yes ||
| [https://etcd.io/ etcd] || K/V || [https://etcd.io/docs/v3.3/learning/api_guarantees/ Linearizability] || LWT ("mini") || yes || [https://etcd.io/docs/v3.3/learning/why/ Why?]
|-
|-
| [https://docs.ceph.com/en/quincy/rados/api/librados-intro/ Ceph Rados] || [http://www.cs.nthu.edu.tw/~ychung/conference/2016_ICPADS.pdf Linearizability] || LWT via Object Operations || yes ||
| [https://docs.ceph.com/en/quincy/rados/api/librados-intro/ Ceph Rados] || K/V ||[http://www.cs.nthu.edu.tw/~ychung/conference/2016_ICPADS.pdf Linearizability] || LWT via Object Operations || yes ||
|-
|-
| [https://www.foundationdb.org/ FoundationDB] || K/V || [https://jepsen.io/consistency/models/strict-serializable Strict Serializability] || yes || yes || [http://charap.co/reading-group-foundationdb-a-distributed-unbundled-transactional-key-value-store/ Architecture review]
| [https://www.yugabyte.com/ YugabyteDB] || [https://youtu.be/w_zYYF3-iSo?t=2001 Single-key Linearizability with LWT] || yes || yes ||
|-
|-
| [https://www.yugabyte.com/ YugabyteDB] || SQL || [https://youtu.be/w_zYYF3-iSo?t=2001 Single-key Linearizability with LWT] || yes || yes ||
| [https://cassandra.apache.org/_/index.html Apache Cassandra] || Configurable: Linearizability with LWT<br />Upcoming: [https://thenewstack.io/an-apache-cassandra-breakthrough-acid-transactions-at-scale/ Strict Serializability] [https://cwiki.apache.org/confluence/display/CASSANDRA/CEP-15%3A+General+Purpose+Transactions?preview=/188744725/188744736/Accord.pdf]|| yes || [https://docs.datastax.com/en/developer/cpp-driver/2.16/topics/ yes] ||

|-
|-
| [https://cassandra.apache.org/_/index.html Apache Cassandra 1-4] || CQL || Configurable, highest:<br />[https://cassandra.apache.org/doc/latest/cassandra/architecture/guarantees.html#lightweight-transactions-with-linearizable-consistency Single-key Linearizability with LWT]|| yes || [https://docs.datastax.com/en/developer/cpp-driver/2.16/topics/ yes] ||
| [https://www.foundationdb.org/ FoundationDB] || [https://jepsen.io/consistency/models/strict-serializable Strict Serializability] || yes || yes ||

|-
|-
| [https://cassandra.apache.org/_/index.html Apache Cassandra 5] || CQL || Upcoming:<br />[https://thenewstack.io/an-apache-cassandra-breakthrough-acid-transactions-at-scale/ Strict Serializability] [https://cwiki.apache.org/confluence/display/CASSANDRA/CEP-15%3A+General+Purpose+Transactions?preview=/188744725/188744736/Accord.pdf]|| yes || [https://docs.datastax.com/en/developer/cpp-driver/2.16/topics/ yes] ||
| [https://www.cockroachlabs.com/ CockroachDB] || [https://www.cockroachlabs.com/blog/consistency-model/#cockroachdbs-consistency-model-more-than-serializable-less-than-strict-serializability Serializable+], [https://www.cockroachlabs.com/blog/db-consistency-isolation-terminology/ Strong Partition Serializability] || LWT || yes || [https://jepsen.io/analyses/cockroachdb-beta-20160829 1]

|-
|-
| [https://www.cockroachlabs.com/ CockroachDB] || SQL || [https://www.cockroachlabs.com/blog/consistency-model/#cockroachdbs-consistency-model-more-than-serializable-less-than-strict-serializability Serializable+], [https://www.cockroachlabs.com/blog/db-consistency-isolation-terminology/ Strong Partition Serializability] || LWT || yes || [https://jepsen.io/analyses/cockroachdb-beta-20160829 Jepsen]
| [https://www.scylladb.com/ ScyllaDB] || [https://docs.scylladb.com/stable/using-scylla/lwt.html Single-key Linearizability with LWT] || LWT || yes ||
|-
|-
| [https://tikv.org/ tikv] || [https://tikv.org/docs/3.0/reference/faq/ Strong Consistency][https://github.com/tikv/tikv/issues/13956] || yes (Percolator based) || no (C++) ||
| [https://www.scylladb.com/ ScyllaDB] || CQL || [https://docs.scylladb.com/stable/using-scylla/lwt.html Single-key Linearizability with LWT] || LWT || yes || [https://docs.scylladb.com/stable/kb/dpdk-hardware.html Intel DPDK via Seastar]
|-
|-
| [https://pegasus.apache.org/ Apache Pegasus] || ? || ? || no (C++) ||
| [https://tikv.org/ tikv] || K/V || [https://tikv.org/docs/3.0/reference/faq/ Strong Consistency][https://github.com/tikv/tikv/issues/13956] || yes (Percolator based) || no (C++) ||
|-
|-
| [https://hbase.apache.org/ Apache HBase] || ? || ? || ? ||
| [https://pegasus.apache.org/ Apache Pegasus] || ? || ? || ? || no (C++) ||
|-
|-
| [https://ignite.apache.org/ Apache Ignite] || ? || ? || ? ||
| [https://hbase.apache.org/ Apache HBase] || ? || ? || ? || ? ||
|-
| [https://ignite.apache.org/ Apache Ignite] || ? || ? || ? || ? ||
|-
| [https://riak.com/ Riak] || K/V || [https://docs.riak.com/riak/kv/2.2.3/learn/concepts/strong-consistency.1.html Linearizability] || No || ? ||
|}
|}


Line 39: Line 46:
* [https://git.samba.org/?p=slow/samba.git;a=blob;f=python/samba/samba3/dbwrap_py_tdb.py;h=1931ff2f43a3a522d402d20dfc3306b184b037d8;hb=refs/heads/dbwrap_py TDB]
* [https://git.samba.org/?p=slow/samba.git;a=blob;f=python/samba/samba3/dbwrap_py_tdb.py;h=1931ff2f43a3a522d402d20dfc3306b184b037d8;hb=refs/heads/dbwrap_py TDB]


= Strict-serializability, can we get away with something weaker? =
= List of Opensource Clustered Filesystems =


* [https://www.cockroachlabs.com/blog/db-consistency-isolation-terminology/ How to talk about consistency and isolation in distributed DBs]
* [https://dbmsmusings.blogspot.com/2019/06/correctness-anomalies-under.html Correctness Anomalies Under Serializable Isolation]
* [https://kousiknath.medium.com/consistency-guarantees-in-distributed-systems-explained-simply-720caa034116 Consistency Guarantees in Distributed Systems Explained Simply]
* [http://muratbuffalo.blogspot.com/2022/08/strict-serializability-but-at-what-cost.html Strict-serializability, but at what cost, for what purpose?]
* [https://www.cockroachlabs.com/blog/consistency-model/ CockroachDB's consistency model]

= List of Opensource Clustered Filesystems =


{| class="wikitable sortable"
{| class="wikitable sortable"
Line 63: Line 77:
= Links =
= Links =


* [https://www.microsoft.com/en-us/research/wp-content/uploads/2011/10/ConsistencyAndBaseballReport.pdf Replicated Data Consistency Explained Through Baseball]
* [https://fauna.com/blog/comparison-of-scalable-database-isolation-levels A Comparison of Scalable Database Isolation Levels]
* [https://fauna.com/blog/comparison-of-scalable-database-isolation-levels A Comparison of Scalable Database Isolation Levels]
* [https://fauna.com/blog/demystifying-database-systems-correctness-anomalies-under-serializable-isolation Correctness Anomalies Under Serializable Isolation]
* [https://fauna.com/blog/demystifying-database-systems-correctness-anomalies-under-serializable-isolation Correctness Anomalies Under Serializable Isolation]
* [https://accelazh.github.io/storage/Linearizability-Vs-Serializability-And-Distributed-Transactions Linearizability vs Serializability, and Distributed Transactions]
* [https://docs.yugabyte.com/preview/faq/comparisons/ Comparison of a few SQL and no-SQL DBs]
* [https://docs.yugabyte.com/preview/faq/comparisons/ Comparison of a few SQL and no-SQL DBs]
* [https://li-boxuan.medium.com/distributed-transaction-in-database-from-epaxos-to-accord-6de7999ad08e Summary of Paxos, Multipaxos, Raft, Epaxos and ACCORD]
* [https://li-boxuan.medium.com/distributed-transaction-in-database-from-epaxos-to-accord-6de7999ad08e Summary of Paxos, Multipaxos, Raft, Epaxos and ACCORD]
Line 75: Line 91:
* [https://www.youtube.com/watch?v=w_zYYF3-iSo Comparing Distributed Transaction Architectures for the Cloud Era]
* [https://www.youtube.com/watch?v=w_zYYF3-iSo Comparing Distributed Transaction Architectures for the Cloud Era]
* [https://pdos.csail.mit.edu/6.824/index.html MIT Distributed Systems Course]
* [https://pdos.csail.mit.edu/6.824/index.html MIT Distributed Systems Course]
* [http://www.vldb.org/pvldb/vol7/p181-bailis.pdf Highly Available Transactions: Virtues and Limitations]
* [https://storage.googleapis.com/pub-tools-public-publication-data/pdf/45855.pdf Spanner, TrueTime and The CAP Theorem]
* [https://paxos.systems/ Paxos Made Moderately Complex]

Latest revision as of 13:03, 6 May 2023

List of scalable Key/Value Stores

For an overview of Consistency Models see Jepsen or Wikipedia. Note: LWT are defined as multiple operations on a single key.

Name Type Consistency Model Transactions C Client Notes
ctdb K/V Volatile DBs: Prefix-Linearizable?
Persistent DBs: Strict Serializability
No
Yes
yes
etcd K/V Linearizability LWT ("mini") yes Why?
Ceph Rados K/V Linearizability LWT via Object Operations yes
FoundationDB K/V Strict Serializability yes yes Architecture review
YugabyteDB SQL Single-key Linearizability with LWT yes yes
Apache Cassandra 1-4 CQL Configurable, highest:
Single-key Linearizability with LWT
yes yes
Apache Cassandra 5 CQL Upcoming:
Strict Serializability [1]
yes yes
CockroachDB SQL Serializable+, Strong Partition Serializability LWT yes Jepsen
ScyllaDB CQL Single-key Linearizability with LWT LWT yes Intel DPDK via Seastar
tikv K/V Strong Consistency[2] yes (Percolator based) no (C++)
Apache Pegasus ? ? ? no (C++)
Apache HBase ? ? ? ?
Apache Ignite ? ? ? ?
Riak K/V Linearizability No ?

Rapid Samba dbwrap backend prototyping

A new dbwrap backend called dbwrap_py that calls into Python for the database operations. Available backends:

Strict-serializability, can we get away with something weaker?

List of Opensource Clustered Filesystems

Name Built on-top of Object Storage Notes
Ceph yes
GlusterFS no
JuiceFS yes
MooseFS no
ObjectiveFS yes
Lustre no
BeegFS no

Links