The Case For Software-Based IBM i HA/DR
July 25, 2022 Ash Giddings
Software-based replication surfaced in the 1990s and continues to be a very popular choice for IBM i shops of all sizes, across an array of industry sectors. But what are the benefits of using software solutions versus hardware alternatives for HA/DR? There are a number of benefits that are an outgrowth of an HA/DR strategy that go beyond HA/DR and where the software approach is better than the hardware approach.
The first one is business intelligence.
Due to its reliability, powerful integration capabilities and ability to secure, many choose Power Systems running IBM i as their main database server or business engine, housing some of the most critical data that businesses possess. Software-based HA/DR provides a method of keeping a copy of this data on an alternative server for business intelligence and reporting purposes. By separating BI from the main database, you also move the workload and associated demands on resources away from the production server, eliminating the impact it can have on more important production transactions. Software-based HA/DR solutions keep data synchronized in real-time so that all reporting is conducted on current data, imperative to accurate reporting. In addition, having a full copy or a subset of data available on an alternative server can also be beneficial for development or test purposes.
It’s never good practice to run BI over the data as is being used by production users and jobs. Running complex queries on transaction processing systems during business hours can have a profound impact on the throughput of the system, and that in turn can have adverse effects on the business. This is one of the reasons why data warehouses were invented in the first place. With software-based replication, you get a full complete system on which to do BI, and that can help drive the business.
The second important thing about software-based replication is that it requires no re-engineering of applications.
Software-based replication is built upon remote journaling, an integral part of the IBM i operating system for over 25 years, and when implemented requires no application re-engineering. Hardware-based replication in the form of solutions such as PowerHA is centered on the concept of clustering technology, with one partition having access to the data at any one time. This data, along with associated applications reside in an Independent Auxiliary Storage Pool (IASP) meaning a level of application re-engineering will need to be conducted prior to implementation.
For objects that cannot be housed in an iASP, PowerHA includes an administrative domain, although it’s worth noting that not every IBM i object can live in here and there are some limitations in the number of objects that can be handled this way. This domain is quite basic. Software-based replication can complement PowerHA installations and can seamlessly manage objects living outside of the iASP.
Third, with software-based replication, distance is not a problem.
With the advent of the cloud and before that the concept of co-locations, software-based replication really came to the forefront. The physical geographical location of servers and the distance between them bears no real relevance due to the ability of software-based replication to operate in either synchronous or asynchronous modes.
Synchronous replication is normally used for environments that are physically close where an acknowledgement is received before the next transaction is processed, with asynchronous mode designed to tolerate longer distances between the source and target partitions, and normally better suited to co-location, or the on-premise and cloud architecture mix. Asynchronous replication is generally faster as the remote journal communication transfer is initiated at the same time with the assumption that it will be successful, with verification and acknowledgment afterwards.
The fourth beneficial side effect of using software-based replication is that it can be used to speed up server migrations.
Software-based replication is incredibly versatile. Its use has historically been in the HA/DR space but in recent years businesses have adopted it as a way of migrating to new hardware without the cumbersome and lengthy save and restore approach. Data can be migrated without downtime and kept synchronized, thus reducing risk to the business. Maxava, for instance, provides short term licensing for server migration purposes, whether customers are moving to Power9 systems or Power10 systems – and whether they are moving across on premises machines or to those located in the cloud.
The other good thing about software-based replication is that it can be mixed with hardware-based replication, and there are times when this is a good idea.
Many IBM i shops have already invested in hardware-based replication solutions such as PowerHA and to a lesser extent, Db2 Mirror for i. While PowerHA has several configuration options depending on the geographical proximity of the nodes, Db2 Mirror for i is limited in the distance between the nodes, the greater the distance the larger the latency.
Software-based replication happily co-exists with hardware replication in active-passive environments in PowerHA or active-active setups in Db2 Mirror for i to copy the data to alternative servers, irrespective of location.
If you want backup flexibility, then the software approach to HA/DR is also useful.
Software-based HA/DR provides a replica copy of production data on a target partition, and as such many businesses take advantage of this by performing their backups on this target, away from the production server and eliminating the need for downtime. Prior to any saves being started, its commonplace for the remote journal apply sessions to be stopped and then restarted once the saves are complete. While the save is taking place, transactions are stored in their associated journal receivers. When backups are complete and apply sessions are resumed transactions are relayed and nothing is lost.
Let’s rattle off a few more benefits of software-based replication.
- Low Bandwidth Requirements: Logical replication is incredibly flexible: you can choose to replicate one file to pretty much everything the IBM i has to offer including database files, the Integrated File System, IBM MQ, and even QDLS. When it comes to high-transaction databases files that are potentially being written to many thousands of times an hour, logical replication really comes into its own by sending only changes to those files to the target sever. With solutions like PowerHA, you are replicating multiple blocks of data and as such the network bandwidth requirements are generally far greater.
- The Disaster Recovery Choice for Cloud: With many now looking to utilize the cloud for disaster recovery purposes, logical replication can be used to keep both on-premise and cloud-based partitions in-sync. The great advantage with using software-based replication in the cloud is that the target server can be run with absolute minimum resources, just enough for the replication to keep up. This keeps expenditure down and helps align expenditure with revenue. Should you need to invoke disaster recovery, vCPUs and memory can be increased to suit your requirements.
- Price Flexibility: Software-based HA/DR solutions such as Maxava HA can be purchased both via the traditional method with a perpetual license or the increasingly popular modern subscription model.
This article is provided to offer general information on matters on which the reader may wish to seek further guidance, and by its nature cannot be comprehensive and cannot be relied on as advice in respect to a specific environment or circumstances. Please consult your IT advisor or contact the writer for advice specific to your situation.
Ash Giddings is a product manager at Maxava and an IBM Champion 2022.
This content is sponsored by Maxava.
RELATED STORIES
In The IBM i Trenches With: IBM Champion Ash Giddings
PowerHA is really a fire and forget solution, and, being at the underlying iASP block level, generally less convoluted and complex to obtain a copy. But it is what it is: the main focus is having a ready copy in case of disaster and unplanned downtimes.
I hope as a full solution DB2 mirror (that is a software based replication according to the taxonomy of the article) will improve and come to a maturation allowing full syncing of the machines state (consider that we have IFS data too!).
Having two “really live” machines (without vary-on times), despite one can use the solution in a master – slave fashion, it is indeed useful to reduce time due to planned downtimes, like PTF, OS upgredes.
Hi Ash Giddings,
Thanks for sharing such a informative post regarding IBM i that will boost my knowledge. Great work! Thanks again.