This is link to issue related to read lock free cluster
https://github.com/orientechnologies/orientdb/issues/1600 .


On Fri, Mar 14, 2014 at 11:44 AM, Andrey Lomakin
<[email protected]>wrote:

> Hi Stefan,
>
> 1. Yes continues writes affects queries, now we can not avoid it but we
> will start to work on design of cluster where writes do  not affect reads.
> 2. Wal is purged now when it strikes maximum limit (about 4 GBs) but we
> are going to implement fuzzy checkpoints, when WAL will be truncated once
> operations are flushed to the disk.
> 3. Sorry , not now, in reality even if you do single thread inserts you
> push everything in memory and background thread sorts out everything by
> physical position and flush them to avoid I/O bottlneck, so in reality you
> work in multi thread mode any way.
>
> Sorry that all I write is "future plans" but they are not in far future
> any way.
>
>
> On Thu, Mar 13, 2014 at 12:04 PM, <[email protected]> wrote:
>
>> Hi Andrey,
>>
>> Thank you for taking the time to help, It's appreciated.
>>
>>
>>    1. What is the best way to make sure read-locking is not in the way
>>    for queries?
>>    What do you mean ?
>>
>> I imagine/think that the continuous writes affected the queries.
>> A typical database would issue a read-lock for querying that would
>> prevent writes on the target entities while the query results are prepared.
>>
>> I want to avoid that if possible but maybe it's a none issue with Orient
>> DB.
>>
>> 2. The WAL becomes quite big - when/how is it purged?
>>
>> 3. We are using a single thread for inserts in append-only mode - Is
>> there anything else I can take advantage of in that scenario?
>>
>> Regards,
>>   -Stefán
>>
>>
>>
>>
>> On Thursday, 13 March 2014 09:20:20 UTC, Andrey Lomakin wrote:
>>
>>> Hi,
>>>
>>> Should we ignore transactions and WAL because of the single threaded
>>>> inserts?
>>>
>>>
>>> No you should not. It is question of durability, not thread safety.
>>>
>>> What is the best way to make sure read-locking is not in the way for
>>>> queries?
>>>
>>>
>>> What do you mean ?
>>>
>>> What configuration would you recommend?
>>>
>>>
>>> The only question is usually  usage of second level and first level
>>> cache, if you do not use complex graph traverse, so I would suggest you to
>>> minimize heap memory usage and assign all RAM to the disk cache.
>>>
>>>
>>>
>>>
>>> On Thu, Mar 13, 2014 at 11:04 AM, Riccardo Tasso 
>>> <[email protected]>wrote:
>>>
>>>> I'm quite interested in this thread, since I'm in a similar case.
>>>>
>>>> Cheers,
>>>>    Riccardo
>>>>
>>>>
>>>> 2014-03-13 9:53 GMT+01:00 <[email protected]>:
>>>>
>>>> bump
>>>>>
>>>>>
>>>>> On Wednesday, 12 March 2014 12:31:24 UTC, [email protected]:
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> Now our event store is almost complete and working quite well on top
>>>>>> of OrientDB, thank you!
>>>>>>
>>>>>> I'm wondering what the best setup/configuration is for our case.
>>>>>>
>>>>>> Inserts:
>>>>>>
>>>>>>    - We have a single thread inserting into the database from a
>>>>>>    message queue
>>>>>>    - About 500 - 2000 event messages are inserted every second.
>>>>>>    - It batches the events messages and writes them in batches
>>>>>>    depending on batch_size or time_box, what ever comes first.
>>>>>>    - Each event message creates multiple vertexes/edges
>>>>>>
>>>>>> Queries
>>>>>>
>>>>>>    - Queries are handled directly with the REST API or with exposing
>>>>>>    a version of the Java SQL-API.
>>>>>>    - We need to be able to deal with a high number of concurrent
>>>>>>    queries (most of them are quite simple)
>>>>>>
>>>>>> The things I'm considering and need help with are:
>>>>>>
>>>>>>    1. Should we ignore transactions and WAL because of the single
>>>>>>    threaded inserts?
>>>>>>    2. What is the best way to make sure read-locking is not in the
>>>>>>    way for queries?
>>>>>>    3. What configuration would you recommend?
>>>>>>
>>>>>> Regards,
>>>>>>   -Stefán
>>>>>>
>>>>>  --
>>>>>
>>>>> ---
>>>>> You received this message because you are subscribed to the Google
>>>>> Groups "OrientDB" group.
>>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>>> an email to [email protected].
>>>>>
>>>>> For more options, visit https://groups.google.com/d/optout.
>>>>>
>>>>
>>>>  --
>>>>
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "OrientDB" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>> an email to [email protected].
>>>>
>>>> For more options, visit https://groups.google.com/d/optout.
>>>>
>>>
>>>
>>>
>>> --
>>> Best regards,
>>> Andrey Lomakin.
>>>
>>> Orient Technologies
>>> the Company behind OrientDB
>>>
>>>   --
>>
>> ---
>> You received this message because you are subscribed to the Google Groups
>> "OrientDB" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> For more options, visit https://groups.google.com/d/optout.
>>
>
>
>
> --
> Best regards,
> Andrey Lomakin.
>
> Orient Technologies
> the Company behind OrientDB
>
>


-- 
Best regards,
Andrey Lomakin.

Orient Technologies
the Company behind OrientDB

-- 

--- 
You received this message because you are subscribed to the Google Groups 
"OrientDB" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to