6 ms·

But... you will lose data if your conflict resolution strategy is LWW, unless you do all writes to unique objects. Siblings+vclocks are provably equivalent to t

by aphyr 13y ago

But... you will lose data if your conflict resolution strategy is LWW, unless you do all writes to unique objects. Siblings+vclocks are provably equivalent to that strategy; they just allow you to garbage-collect unnecessary parts of the causal history more efficiently. Neither strategy frees you from having to define a commutative, associative, and idempotent merge function to perform a read.

jbellis 13y ago
The whole point is that "conflicting" updates to a single column is supposed to cause overwrites ("data loss"). If I wanted to keep multiple values in a column around I'd use a Map or a Set instead! Maybe the disconnect is that in Riak, you have to fetch the existing document before modifying it anyway, so there is a lot of "update-based-on-existing-document" code around. Cassandra is designed to encourage "blind," idempotent updates instead.
- aphyr 13y ago
  If I wanted to keep multiple values in a column around I'd use a Map or a Set instead! [edit: thanks iamalesky, correction on cell keys] If you're following along at home, Lists are implemented by choosing a UUID cell key for each distinct element of the collection; Maps and Sets use the key and value, respectively. The conflict resolution function applied at read time is, for Sets, set union, plus tombstones elements for deletion. Depending on the resolution strategy used, similar consistency anomalies to LWW may be present; e.g clock skew allows for various possible outcomes of logically concurrent mutation of a collection.
  - iamaleksey 13y ago
    In case you are talking about Cassandra collections (maps and sets), and not abstract maps/sets, then you are wrong. Only CQL3 Lists work like that - there are no UUIDs anywhere in CQL3 maps/sets implementations.
- brown9-2 13y ago
  Doesn't this assume that both writers want their value for the column to win, instead of one of them possibly deciding that it shouldn't write it's value if one already exists for the field?
  - jbellis 13y ago
    Correct. But, in that case vector clocks does not help you either [unless you are okay with both writers thinking they have "won" for an unbounded period of time]; what you really need is true linearizability: http://www.datastax.com/dev/blog/lightweight-transactions-in-cassandra-2-0 http://www.datastax.com/dev/blog/lightweight-transactions-in...
rbranson 13y ago
Agreed in general. I think the statement that "you will lose data" is a bit simplistic. Given enough chances, all systems will lose data. One could make a pretty strong argument that the default LWW approach used by Riak and Voldemort is quantifiably far less safe than the default approach in Cassandra, which works more like an LWW-Element-Set. I know this is changing with the CRDT work in Riak 2.0, which is very exciting. There ARE a large number of use cases, many of which are driving the demand for scale-out distributed DBs, where data IS immutable, with a requirement for ordered traversal over subsets of the data. The key/value+vclock approaches that I've seen make this either very difficult or very slow.
- aphyr 13y ago
  Sorry, I was speaking loosely. More formally: In a system which uses LWW as the conflict resolution strategy, there exist no circumstances under which you can guarantee that a value written to a given key will be causally connected to any future state of the system, unless all values written to that key are identical, or a strong external coordinator (e.g. Zookeeper) orders timestamps. If you have siblings and vclocks, you can recover that causal connection guarantee for arbitrary write patterns--at least over CRDTs. Since Cassandra did not (until today) offer transactional isolation for any type of multi-cell update, this means that--and we're speaking strictly in terms of safety here, not performance--Riak and Voldemort's consistency models were, prior to 2.0, a strict superset of Cassandra's. For instance, you can guarantee the visibility and transactional isolation of a write making multiple changes to a Riak object; I'm reasonably confident that you cannot achieve those guarantees in, say, a Cassandra collection without a Paxos transaction. You can certainly emulate Riak's consistency model by storing a distinct object for every write, and this is, as I understand it, what many Cassandra users do. The difference is in space consumption. Consider making four updates to an object. In Cassandra, you could write each update to a separate cell. In Riak, you might write them all to the same key: Cassandra Riak [update1] [update1|update2|update3|update4] [update2] [update3] [update4] To read from both Cassandra and Riak you need a merge function. Since neither provides ordering constraints, our merge must be associative, commutative, and idempotent in both cases. Cassandra Riak [update1]+ [update1|update2|update3|update4] [update2]+ | | | | [update3]+ +--------+-------+-------+ [update4]+ | | | V V [current value] [current value] The difference is in space. Vector clocks allow you to prune the causal history, meaning we can write back [current value], and as soon as a node sees that write, it can discard updates 1-4. In Cassandra, there is no causality tracking: you have to figure out how to do GC yourself, or punt. Cassandra Riak [update1] [merged value|update5] [update2] [update3] [update4] [update5] You can see how unbounded space might be a problem. From my conversations with DataStax, it sounds like users tend to write reducers which apply their merge function to compact some portion of the history. Which portion? Well, without causality tracking we'll leave that as an exercise to the reader. Cassandra Riak [update1-4] [merged value|update5] [update5] Does this look familiar? Yeah. It's the same concurrency model as the vector clocks this post is arguing against. You just have to do more work. Now, there are all sorts of practical efficiency constraints at play! For instance, Riak has ~50-100 bytes of overhead per key, and will start barfing if you go over 10 megabytes per key or so. And without being able to call list-keys, you wind up having to play all kinds of games with predictable keys, splitting datasets between multiple objects, and so on. Cassandra's IO throughput generally seems much higher than Riak's, and Cassandra has a much more efficient representation for wide values. It also offers better key ranges--but you also pay a per-cell overhead for every atomic chunk of state. Not so efficient if you were looking to store, say, big blocks of integers for your CRDTs. The great thing is--again speaking purely in terms of consistency--Cassandra 2.0 is now capable of a superset of Riak's operations! If correctly implemented, their Paxos operations support linearizable reads and writes, which is a way stronger class of consistency than the CRDT operations described above. I don't understand why jbellis is so upset when folks point out that LWW provides weak safety constraints--when their strongly-consistent operations now offer the highest level of transactional safety. Seems like we should be celebrating that achievement, because it opens up large classes of operations which were previously unsafe. :)

by aphyr 13y ago

jbellis 13y ago

The whole point is that "conflicting" updates to a single column is supposed to cause overwrites ("data loss"). If I wanted to keep multiple values in a column around I'd use a Map or a Set instead! Maybe the disconnect is that in Riak, you have to fetch the existing document before modifying it anyway, so there is a lot of "update-based-on-existing-document" code around. Cassandra is designed to encourage "blind," idempotent updates instead.

aphyr 13y ago
If I wanted to keep multiple values in a column around I'd use a Map or a Set instead! [edit: thanks iamalesky, correction on cell keys] If you're following along at home, Lists are implemented by choosing a UUID cell key for each distinct element of the collection; Maps and Sets use the key and value, respectively. The conflict resolution function applied at read time is, for Sets, set union, plus tombstones elements for deletion. Depending on the resolution strategy used, similar consistency anomalies to LWW may be present; e.g clock skew allows for various possible outcomes of logically concurrent mutation of a collection.
- iamaleksey 13y ago
  In case you are talking about Cassandra collections (maps and sets), and not abstract maps/sets, then you are wrong. Only CQL3 Lists work like that - there are no UUIDs anywhere in CQL3 maps/sets implementations.
brown9-2 13y ago
Doesn't this assume that both writers want their value for the column to win, instead of one of them possibly deciding that it shouldn't write it's value if one already exists for the field?
- jbellis 13y ago
  Correct. But, in that case vector clocks does not help you either [unless you are okay with both writers thinking they have "won" for an unbounded period of time]; what you really need is true linearizability: http://www.datastax.com/dev/blog/lightweight-transactions-in-cassandra-2-0 http://www.datastax.com/dev/blog/lightweight-transactions-in...

rbranson 13y ago

Agreed in general. I think the statement that "you will lose data" is a bit simplistic. Given enough chances, all systems will lose data. One could make a pretty strong argument that the default LWW approach used by Riak and Voldemort is quantifiably far less safe than the default approach in Cassandra, which works more like an LWW-Element-Set. I know this is changing with the CRDT work in Riak 2.0, which is very exciting. There ARE a large number of use cases, many of which are driving the demand for scale-out distributed DBs, where data IS immutable, with a requirement for ordered traversal over subsets of the data. The key/value+vclock approaches that I've seen make this either very difficult or very slow.

aphyr 13y ago
Sorry, I was speaking loosely. More formally: In a system which uses LWW as the conflict resolution strategy, there exist no circumstances under which you can guarantee that a value written to a given key will be causally connected to any future state of the system, unless all values written to that key are identical, or a strong external coordinator (e.g. Zookeeper) orders timestamps. If you have siblings and vclocks, you can recover that causal connection guarantee for arbitrary write patterns--at least over CRDTs. Since Cassandra did not (until today) offer transactional isolation for any type of multi-cell update, this means that--and we're speaking strictly in terms of safety here, not performance--Riak and Voldemort's consistency models were, prior to 2.0, a strict superset of Cassandra's. For instance, you can guarantee the visibility and transactional isolation of a write making multiple changes to a Riak object; I'm reasonably confident that you cannot achieve those guarantees in, say, a Cassandra collection without a Paxos transaction. You can certainly emulate Riak's consistency model by storing a distinct object for every write, and this is, as I understand it, what many Cassandra users do. The difference is in space consumption. Consider making four updates to an object. In Cassandra, you could write each update to a separate cell. In Riak, you might write them all to the same key: Cassandra Riak [update1] [update1|update2|update3|update4] [update2] [update3] [update4] To read from both Cassandra and Riak you need a merge function. Since neither provides ordering constraints, our merge must be associative, commutative, and idempotent in both cases. Cassandra Riak [update1]+ [update1|update2|update3|update4] [update2]+ | | | | [update3]+ +--------+-------+-------+ [update4]+ | | | V V [current value] [current value] The difference is in space. Vector clocks allow you to prune the causal history, meaning we can write back [current value], and as soon as a node sees that write, it can discard updates 1-4. In Cassandra, there is no causality tracking: you have to figure out how to do GC yourself, or punt. Cassandra Riak [update1] [merged value|update5] [update2] [update3] [update4] [update5] You can see how unbounded space might be a problem. From my conversations with DataStax, it sounds like users tend to write reducers which apply their merge function to compact some portion of the history. Which portion? Well, without causality tracking we'll leave that as an exercise to the reader. Cassandra Riak [update1-4] [merged value|update5] [update5] Does this look familiar? Yeah. It's the same concurrency model as the vector clocks this post is arguing against. You just have to do more work. Now, there are all sorts of practical efficiency constraints at play! For instance, Riak has ~50-100 bytes of overhead per key, and will start barfing if you go over 10 megabytes per key or so. And without being able to call list-keys, you wind up having to play all kinds of games with predictable keys, splitting datasets between multiple objects, and so on. Cassandra's IO throughput generally seems much higher than Riak's, and Cassandra has a much more efficient representation for wide values. It also offers better key ranges--but you also pay a per-cell overhead for every atomic chunk of state. Not so efficient if you were looking to store, say, big blocks of integers for your CRDTs. The great thing is--again speaking purely in terms of consistency--Cassandra 2.0 is now capable of a superset of Riak's operations! If correctly implemented, their Paxos operations support linearizable reads and writes, which is a way stronger class of consistency than the CRDT operations described above. I don't understand why jbellis is so upset when folks point out that LWW provides weak safety constraints--when their strongly-consistent operations now offer the highest level of transactional safety. Seems like we should be celebrating that achievement, because it opens up large classes of operations which were previously unsafe. :)