Backing up databases and applications always is critically important. Still I can’t help but think that companies moving towards continuous availability solutions. For example, EMC has several different flavors of VPLEX for architecting continuous availability solutions. Using VPLEX Metro with Extended Oracle RAC is a solution that offers business the ability to keep the database and applications running even with the loss of a datacenter.
But this is my question: Do you see a trend or perhaps pressure to architect continuous availability solutions for your business?
I have been burning through the DataStax Cassandra training online (Cassandra is the most similar NoSQL database to Oracle that I have found so far), and they said something along these lines that I found compelling. An old-school, client-server database back end app like Oracle is notoriously difficult to make be CA. We end up with something like RAC, which has huge issues in terms of cost, complexity, overhead, etc.
NoSQL database like Cassandra are designed to be CA. Oracle will need to meet that in order to successfully compete with that.
Good question. I believe there will always be the need for traditional backup of databases and applications. In talking with customers it seems the number of applications that must be continuously available is growing too. I do also like the point that Jeff makes in that newer databases and applications are designed for CA. This really highlights how different the demands were from 20, 10, and even 5 years ago. I can see a time when most production applications have an underlying CA architecture.
Great question and discussion.
Good point. When I look at our portfolio, pretty much everything went into continuous availability compared to 5-6 years ago. We run several RAC Oracle hotels (ranging from 2 to 10 nodes, depending on purpose and tier) and very little dedicated Oracle DB hosts. Then again, this kind of solutions were not there before and only most important databases only would run in some sort of the OS cluster or have standby server solution. Why so? I see two reasons top of my head in my specific case;
1) business model has changed and we are living in 24/7 economy where each tier gets abused 24/7 and availability is must - it is only performance which is different
2) we went for oracle hotel concept where single RAC would host even up to 60 databases so we just use same approach for all of them
There is increased complexity for DBAs for sure, but once this is set it just works fine. Of course, it is not issue-less and never will be.
Great points and similar situation here. The VPLEX Metro with Extended Oracle RAC is one of the hottest solution at the local market in PRC as this is the most efficient way to provide active-active Oracle RAC solution:
- Less CPU resources required with less server cost and Oracle license cost
- Performance garanteed for Cache fusion and disk I/O latencies.
in my country there is only a few customers running Oracle RAC on VPLEX. Most of our accounts see great benefit of active-active datacenter, but as of today they prefer mutual active-standby configs (some applications run in primary DC, others in secondary DC) rather than truly active-active solution for each application. Most common reason for that is the complexity of RAC not only on infrastructure level, but integration of failover mechanisms with applications and challenge of efficient data partitioning to avoid unnecessary interconnect traffic between nodes.
Stretched RAC on Metro Vplex has a wow factor with the DBAs I speak to. Having said this I have had customers moving to the stretched RAC Metro Vplex and if fact convert this to single-instance Oracle on VMware stretching their VMware cluster between their data centers, and dropping RAC. With this they gained continuous availability, lowered their Oracle licensing cost, and captured all the functionality of VMware including continuous availability for both their application and database tiers.