Hi Andrew,
We have observed that following method(segment.access()) blocks ignite
data caching using data streamers(For single ignite instance). This limits
our resource utilization i.e. CPU, MEM are not fully utilized. How can we
avoid this blocking so that we can get maximum performance out
Hi Andrey,
Attached is the code we have used for bench marking. Is there any tuning
that we can apply to get better performance out of ignite single instance
further?
Also we have attached logs taken from our tool where we varied
datastreamer parallelism from 1 to 16(default). In this
Hi Rish,
DataStreamer does not have any threads, DataStreamPool is used for incoming
batches processing.
DataStreamer split you data into batches which size is equal to streamer
perNodeBufferSize parameter.
When batch become full, a task is send to a node the buffer belongs to.
I have tried writing while loop which continuously inserts data(same entry)
with increamenting cache key(so that there is unique key). Without
datastrmr.addData() while loop is generating data at 200K msgs/sec(Data
generation rate is much more than caching rate). Is there any blocking done
by
Hi Rishi,
Is it possible ignite put faster than you preparing data for put?
Or may be you change same entry each time and there is a contention on this
entry?
On Thu, Jun 29, 2017 at 7:44 AM, rishi007bansod
wrote:
> Hi,
> With datastreamer threads = 16(default)
Hi,
With datastreamer threads = 16(default) am able to get upto 80 K
msgs/sec throughput. As I have 56 core machine(with hyperthreading), I have
tried increasing this pernodeparallelprocessing parameter of datastreamer to
56. But still no increase in throughput is observed. Also out of 5600%