Hi,
This is because the applications by design are using the same database.
One server is writing to the Data1 when another one is reading from Data1
for example.
I'm trying to move the load from a single node and cut the number of servers
I have currently.
In Active-Passive one server is stand by and it is waste of resources, so I
installed another instance with different storage and have now VDB01 and VDB02
Now from the 4 servers I can get rid of one and use the VDB02 as a subscriber
VDB01 - Publisher - 11k tx/sec DL380 G2 2 XEON CPUs 4.6GB RAM
SRV01 - DL380 G3 2 CPUs XEON Subscriber only Read 22k to 33k tx/sec -
critical needs to have a failover. 3GB RAM, 50% CPU
SRV02 - Subscriber - for reports - not very busy but on a report hit goes up
to 100% CPU for some minutes. 5GB RAM
DIS01 - Distributor - 10-250 tx/sec, 3GB RAM 25%-50%CPU normally Push
Transactional Replication running non-stop.
Reason we have replication to SRV01 is that on the LAN using ADO to VDB01 it
takes 2-3 seconds more time which is critical.
Now if I skip the replication and use VDB01 only on DL380s Cluster without
SAN and GBit LAN it will crash.
Now we have SAN storage 1GBLAN and GBit switches.
VDB01,VDB02,+1 one additional DL380 G3 probably for distributor, if the LAN
speed is not fast enough, we will have another SRV01 in the picture.
Currently Im adding SRV01 to the VDB02.
I want to prove that this is not enough for the above situation, that is why
Im looking for a better solution using less servers and avoiding replication.
Regards,
C.Dimitrov
[quoted text, click to view] "Mike Epprecht (SQL MVP)" wrote:
> Hi
>
> If both have to be read/write, replication is the only way.
>
> What are you trying to achive with having 2 writable copies of the same DB
> on a cluster?
>
> Regards
> Mike
>
> "C.Dimitrov" wrote:
>
> > Hi,
> >
> > We have one cluster with 2 virtual databases and name instances VDB01\DB1
> > and VDB02\DB2. They use different storage on a SAN. Machines are DL380s with
> > 2 CPUs. OS: Windows 2003 Enterprise Server, SQL 2000SP3a.
> > Now both of them have the same database Data1, Data2 and Data3
> >
> > Existing scenario:
> > We have 1 machine for Distributor with Transactional Push replication from
> > Cluster A(2 DL380s one of them is standby) with Publications to standalone
> > SQL server B because of netowork speed from server B to Cluster A. All
> > togather 4 machines.
> >
> > Im trying to find better scenario using Active/Active so that both nodes are
> > in use without having a replication.
> > I know that is not possible in W2K Cluster to recognize to which node to
> > send the SQL transactions. Now is this possible in W2K3 Cluster?
> >
> > Is it possible to configure instant replicaiton between them without having
> > distributor sitting on another machine considering LAN speed and no VIA
> > available?
> > What is the best practice achiving VDB01 and VDB02 devices to have the same
> > data?
> >
> > Regards,
