Re: Channelling Madison Pruett
Madison, Is that Client Reroute must work with Stinger HADR? If we use db2 suspend + Shark or EMC split, can we use Client Reroute to reroute the client to the secondary server? Thanks, "mpruet" <mpruet@comcast.net> wrote in message news:40676184.0404271949.3ecdffa2@posting.google.com... > "rkusenet" <rkusenet@sympatico.ca> wrote in message news:<c6ma55$dosun$1@ID-75254.news.uni-berlin.de>... > > "Rob Vorbroker" <rvorbroker@yahoo.com> wrote in message > > news:c6m9hu$t7c$1@terabinaries.xmission.com... > > > > > > For those of us venturing out into the DB2 UDB world > > > can you give us a summary of tools similar to HDR and > > > ER (which we love so much). > > > > > > Trying to support 24x7x365 internet applications and > > > have a 'requirement' to make the DB2 UDB servers back > > > up in 5-10 minutes (or as close as possible) both from > > > a local perspective (HA/CMP?) as well as a DR (remote > > > site) perspective. > > > > > > Give me some good news... > > > What should my real expectations be with each > > > tool/scenario? > > > > This is what I know:- > > > > DB2 UDB currently support replication via SQL calls. Its throughput is > > not good. In the upcoming version (code names stinger), IBM is > > introducing MQ series based replication for higher throughput. MQ series > > will be bundled with DB2UDB. > > Let's distinguish between throughput and latency. ;-) > > The throughput of DPROP is good. However, the latency is higher that > ER. The key difference here is that DPROP works in a probing - cycle > model while ER works in a push transactional model. That means that > the DPROP apply is able to buddy-bunch the data being applied (which > does increase the total number of rows per hour), but at the cost of > some degree of latency. > > I do think that the upcoming replication solution in DB2 will be > considered as having similar latency as ER for the IDS customer that > is using ER as an HA solution (i.e. primary->target). > > > > > > AFAIK, HDR of Informix will be known as HADR (high availability disaster > > recovery). In the Stinger version, the secondary will not be in a read > > mode. It will be in a cold mode which will start up only when the primary > > is down.
I'm not sure I completely understand the question. Suspend + split sounds to me like a way to set up a second copy of a database using "split mirror". That will be one method by which one can initialize an HADR standby database (the other being Restore). Separately, another "Stinger" feature Client Reroute can be used to help with redirection of clients to an alternative server when the original is down. Use in combination with HADR is definitely anticipated, though not the only thing for which it can be deployed. I'm not clear on how the use of split mirror would be tied in with use of client reroute. Regards, -steve p. ------------------- Steve Pearson DB2 UDB Development Portland, OR, USA "Fan Ruo Xin" <fanruox@sbcglobal.net> wrote in message news:4Gikc.991$eH1.520758@newssvr28.news.prodigy.com... > Madison, > Is that Client Reroute must work with Stinger HADR? If we use db2 suspend + > Shark or EMC split, can we use Client Reroute to reroute the client to the > secondary server? > Thanks, > > "mpruet" <mpruet@comcast.net> wrote in message > news:40676184.0404271949.3ecdffa2@posting.google.com... > > "rkusenet" <rkusenet@sympatico.ca> wrote in message > news:<c6ma55$dosun$1@ID-75254.news.uni-berlin.de>... > > > "Rob Vorbroker" <rvorbroker@yahoo.com> wrote in message > > > news:c6m9hu$t7c$1@terabinaries.xmission.com... > > > > > > > > For those of us venturing out into the DB2 UDB world > > > > can you give us a summary of tools similar to HDR and > > > > ER (which we love so much). > > > > > > > > Trying to support 24x7x365 internet applications and > > > > have a 'requirement' to make the DB2 UDB servers back > > > > up in 5-10 minutes (or as close as possible) both from > > > > a local perspective (HA/CMP?) as well as a DR (remote > > > > site) perspective. > > > > > > > > Give me some good news... > > > > What should my real expectations be with each > > > > tool/scenario? > > > > > > This is what I know:- > > > > > > DB2 UDB currently support replication via SQL calls. Its throughput is > > > not good. In the upcoming version (code names stinger), IBM is > > > introducing MQ series based replication for higher throughput. MQ series > > > will be bundled with DB2UDB. > > > > Let's distinguish between throughput and latency. ;-) > > > > The throughput of DPROP is good. However, the latency is higher that > > ER. The key difference here is that DPROP works in a probing - cycle > > model while ER works in a push transactional model. That means that > > the DPROP apply is able to buddy-bunch the data being applied (which > > does increase the total number of rows per hour), but at the cost of > > some degree of latency. > > > > I do think that the upcoming replication solution in DB2 will be > > considered as having similar latency as ER for the IDS customer that > > is using ER as an HA solution (i.e. primary->target). > > > > > > > > > > AFAIK, HDR of Informix will be known as HADR (high availability disaster > > > recovery). In the Stinger version, the secondary will not be in a read > > > mode. It will be in a cold mode which will start up only when the > primary > > > is down. > >