cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jonathan Haddad <...@jonhaddad.com>
Subject Re: apache cassandra for trading system
Date Sat, 26 Mar 2016 16:23:58 GMT
A few points here...

I don't know why you feel you need to separate your data into 2 clusters.
Cassandra will deal just fine with all your data in a single cluster.  You
can use TTLs to keep trade data for as long as you need to and it will get
purged out automatically.

Having 1 node won't give you HA.

If you're going to use an RDBMS for backup purposes, then why not just use
that if it can handle all the writes?  Your proposed setup feels a lot more
complicated than it needs to be making tradeoffs that don't buy you
anything.

If it was me I'd just put all the data in your Cassandra cluster, in a
single keyspace across 3 nodes (giving you HA), and not bother with the
RDBMS or splitting today's data from historical.

On Sat, Mar 26, 2016 at 8:39 AM Vero Kato <vero.katos@gmail.com> wrote:

> thanks, we are going to have 500 000 msgs per day, historical database
> will be growing every day, but we also want one cluster with one day of
> data. most of the time we will be pushing data into database, but we also
> want to execute client's queries and allow them to fetch historical trades
> and orders. we also need to run some analytics  for reporting on one of day
> data cluster. the cluster should be HA 99% of time as we will have RMDS
> backup solution also for trades.
>
> On Fri, Mar 25, 2016 at 9:33 PM, Jack Krupansky <jack.krupansky@gmail.com>
> wrote:
>
>> How much data do you expect to have - amount added per day, number of
>> days kept, amount of data per trade, number of tradable securities (or
>> whatever you are trading?
>>
>> What kind of access patterns do you expect? Like append, any updates,
>> what queries, any analytics.
>>
>> Generally, queries and access patterns will drive the data modeling, and
>> also impact how much data you can realistically place on each node.
>>
>> What are your HA (High Availability) requirements?
>>
>>
>> -- Jack Krupansky
>>
>> On Fri, Mar 25, 2016 at 2:40 PM, Jonathan Haddad <jon@jonhaddad.com>
>> wrote:
>>
>>> You can use keyspaces with multiple data centers to get what you want.
>>>
>>> That said, if you're going to use only 1 node, I don't think Cassandra
>>> is the right fit for you.
>>>
>>>
>>> http://rustyrazorblade.com/2013/09/cassandra-faq-can-i-start-with-a-single-node/
>>>
>>> On Fri, Mar 25, 2016 at 11:09 AM Vero Kato <vero.katos@gmail.com> wrote:
>>>
>>>> hi, we are building trading system and we want to use cassandra as our
>>>> database.
>>>> we want this set-up:
>>>> one node which stored one day set of data which will be running on the
>>>> same host as trading application
>>>> two nodes which store all data (for the current date and historical)
>>>> which will be running on other boxes.
>>>>
>>>> can you please suggest how easily it will be to have this set-up? i.e.
>>>> we have three nodes, but one of them (on the same box where our app
>>>> located) will be discarding all historical data which are older than one
>>>> day?
>>>>
>>>>
>>>> thanks
>>>>
>>>
>>
>

Mime
View raw message