Switching HDR to RSS replication

This is a follow up to my earlier post on IIUG (July 19, 2017): HDR-Distance and Latency Our HDR distance is changing from 40 miles (NEAR) to 1200 miles (FAR) after 2 months. After testing with a set of simple update/delete statements on triggerred and non-triggerred tables I conclude that the latency is noticable (1.2x - NEAR vs 2.0x- FAR). HDR is ASYNC and there is no noticable checkpoint wait difference, logs complete slower with HDR, but transfer same second to the FAR location. RSS is your recommended option but yet to be tested here. Questions: I beleive HDR to RSS and vice versa is possible per this link: http://www-01.ibm.com/support/docview.wss?uid=swg21398379 My plan was to start with the status quo in production (HDR to FAR) and switch to RSS if intolerably slow. Is this advisable? If yes, I will start testing this scenario. Currently with HDR, LOG_INDEX_BUILDS is not enabled and LOG_STAGING_DIR is not set. Can both values be changed ( onmode -wf) with HDR ON before switching to RSS? System: HP-UX 11.31.ia64, Informix 12.10.FC6

Both can be set dynamically using onmode -wf Art Art S. Kagel, President and Principal Consultant ASK Database Management www.askdbmgt.com Blog: http://informix-myview.blogspot.com/ Disclaimer: Please keep in mind that my own opinions are my own opinions and do not reflect on the IIUG, nor any other organization with which I am associated either explicitly, implicitly, or by inference. Neither do those opinions reflect those of other individuals affiliated with any entity with which I am affiliated nor those of the entities themselves. On Wed, Aug 2, 2017 at 2:10 PM, MURALI PAZHAYANNUR <pmurali@ftportfolios.com > wrote: > This is a follow up to my earlier post on IIUG (July 19, 2017): > HDR-Distance > and Latency > > Our HDR distance is changing from 40 miles (NEAR) to 1200 miles (FAR) > after 2 > months. After testing with a set of simple update/delete statements on > triggerred and non-triggerred tables I conclude that the latency is > noticable > (1.2x - NEAR vs 2.0x- FAR). HDR is ASYNC and there is no noticable > checkpoint > wait difference, logs complete slower with HDR, but transfer same second to > the FAR location. > > RSS is your recommended option but yet to be tested here. > > Questions: > > I beleive HDR to RSS and vice versa is possible per this link: > http://www-01.ibm.com/support/docview.wss?uid=swg21398379 > > My plan was to start with the status quo in production (HDR to FAR) and > switch > to RSS if intolerably slow. Is this advisable? If yes, I will start testing > this scenario. > > Currently with HDR, LOG_INDEX_BUILDS is not enabled and LOG_STAGING_DIR is > not > set. Can both values be changed ( onmode -wf) with HDR ON before switching > to > RSS? > > System: HP-UX 11.31.ia64, Informix 12.10.FC6 > > > ************************************************************ > ******************* > Forum Note: Use "Reply" to post a response in the discussion forum. > >

With RSS there are some tricks that can be done to improve performance. There are two main bottlenecks with both HDR as well as with RSS. With HDR, the logs are sent in half-duplexed mode. That means that even in ASYNC mode, the logs sent must be ACKed from the secondary before the next log buffer will be sent. In ASYNC mode there are additional buffers which will be used to hold the logs prior to sending, but still the actual transfer is going to use half duplexed mode. That's one bottleneck. The second is in the actual apply of the logs on the secondary. If for some reason you run into that bottleneck then you will not be able to send an ACK back to the primary and will slow down the transfer of the log buffers. RSS has some characteristics which can be used to speed up the transfer of the logs. 1) The network transmission uses a full duplexed model. This means that the next buffer can be transmitted immediately. 2) RSS can use multiple network channels for data transmission. This can be an advantage because TCP itself uses a sliding window ACK system as a form of flow control. By using multiple network channels, you can send more data per second than a single network connection. This helps against the network bottleneck. 3) RSS can be configured to use a 1 second delay apply. That means that we can buffer more log pages without having to move them into the apply buffers. This means we have protection from running into the apply bottleneck. The log pages will tend to be on the secondary even if they haven't been applied just yet. Madison Pruet Retired and Loving it On Wednesday, August 2, 2017 2:11 PM, MURALI PAZHAYANNUR <pmurali@ftportfolios.com> wrote: This is a follow up to my earlier post on IIUG (July 19, 2017): HDR-Distance and Latency Our HDR distance is changing from 40 miles (NEAR) to 1200 miles (FAR) after 2 months. After testing with a set of simple update/delete statements on triggerred and non-triggerred tables I conclude that the latency is noticable (1.2x - NEAR vs 2.0x- FAR). HDR is ASYNC and there is no noticable checkpoint wait difference, logs complete slower with HDR, but transfer same second to the FAR location. RSS is your recommended option but yet to be tested here. Questions: I beleive HDR to RSS and vice versa is possible per this link: http://www-01.ibm.com/support/docview.wss?uid=swg21398379 My plan was to start with the status quo in production (HDR to FAR) and switch to RSS if intolerably slow. Is this advisable? If yes, I will start testing this scenario. Currently with HDR, LOG_INDEX_BUILDS is not enabled and LOG_STAGING_DIR is not set. Can both values be changed ( onmode -wf) with HDR ON before switching to RSS? System: HP-UX 11.31.ia64, Informix 12.10.FC6 ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.