Re: Scalebility of Inite Threads

[afedotov <al...@gmail.com>]

Hi,

Please properly subscribe to the mailing list so that the community can
receive email notifications for your messages.
To subscribe, send an empty email to user-subscribe@ignite.apache.org and
follow simple instructions in the reply.

I've some questions to clarify:
1. What do you mean by "Ignite threads"? Is it the number of
threads/clients reading/writing data from/into the cache?
2. Are you checking reads or writes, or both?
3. What is the CPU and memory utilization in case of 1 thread, 16 threads,
and 32 threads?
4. Do you observe high GC pauses under increasing load?

In general, every client thread implies that a request will be executed in
a separate thread on the server nodes.
Increasing the number of clients/threads you need to take into account such
parameters as the number of physical CPU cores,
amount of memory and its speed, network speed on the physical machines running
Ignite server nodes.
If, under some load, you become bound by, for example, CPU then response
time for each thread will increase, but overall throughput should grow.
Actually, it's a common trade-off between response time and throughput.

As well performance differs when you are referencing mostly the same set of
data or different sets.

If you are executing mostly reads then please check if enabling
auxiliary on-heap
cache
<https://apacheignite.readme.io/docs/memory-configuration#section-on-heap-caching>
will
improve the things. In this case, an eviction policy
<https://apacheignite.readme.io/docs/evictions#section-entry-based-eviction-for-on-heap-cache>
should be configured
properly to avoid OOM.






Kind regards,
Alex.

On Sat, Aug 26, 2017 at 10:05 PM, sandeep jain [via Apache Ignite Users] <
ml+s70518n16430h64@n6.nabble.com> wrote:

> We have a scenario where we need to have around 600 Million keys in cache.
> We have partitioned the cache and followed the ignite manual on performance
> tuning. We are observing that the performance degrades when you are using
> increasing the number of threads. With one thread the speed we got was 5000
> records per sec and with increasing the number of threads to 16, speed
> reduced to 1800 and then with further increased the threads to 32, the
> speed reduced to 1200 records per sec. When analyzed the stack, we could
> see a lot of threads in the waiting state and only a few threads which were
> in the runnable state. Others threads were in waiting state. Suspect if
> there is potential issue in synchronization which is resulting in dipping
> down the performance. Also could be that I am doing things incorrectly. Any
> suggestions are welcome and appreciated.
>
> ------------------------------
> If you reply to this email, your message will be added to the discussion
> below:
> http://apache-ignite-users.70518.x6.nabble.com/
> Scalebility-of-Inite-Threads-tp16430.html
> To start a new topic under Apache Ignite Users, email
> ml+s70518n1h65@n6.nabble.com
> To unsubscribe from Apache Ignite Users, click here
> <http://apache-ignite-users.70518.x6.nabble.com/template/NamlServlet.jtp?macro=unsubscribe_by_code&node=1&code=YWxleGFuZGVyLmZlZG90b2ZmQGdtYWlsLmNvbXwxfC0xMzYxNTU0NTg=>
> .
> NAML
> <http://apache-ignite-users.70518.x6.nabble.com/template/NamlServlet.jtp?macro=macro_viewer&id=instant_html%21nabble%3Aemail.naml&base=nabble.naml.namespaces.BasicNamespace-nabble.view.web.template.NabbleNamespace-nabble.view.web.template.NodeNamespace&breadcrumbs=notify_subscribers%21nabble%3Aemail.naml-instant_emails%21nabble%3Aemail.naml-send_instant_email%21nabble%3Aemail.naml>
>




--
View this message in context: http://apache-ignite-users.70518.x6.nabble.com/Scalebility-of-Inite-Threads-tp16430p16449.html
Sent from the Apache Ignite Users mailing list archive at Nabble.com.