You're ready for the last model, model 4? This kicks it up a notch by adding GDPS Continous Availability. GDPS CA consists of two separate sites. Both sites run these same applications and have access to the same data. If one site fails, the workload can failover to the other site within seconds. An average recovery time is 1-2 minutes, and data lasts for 3-5 seconds. The goal of GDPS CA is to have two separate sites, both running the same applications using workloads balancing and read/write workload to execute on only one site although read-only is available on both sites. By doing this, the benefits include using automation for successful recovery, achieving workload distribution between the two sites with this resulting in zero data loss. The GDPS Continuous Availability solution does not require the techniques that GDPS products rely on. However, the GDPS CA solution can also incorporate them for three or more data centers for continuous availability and disaster recovery. Both software and automation are used to achieve the solution. The GDPS Continuous Availability Zero Data Loss solution or ZDL can support zero data loss for an unplanned outage in the active site. Note in the following diagram that both sites must be located within supported Metro Mirror distances. Considering the following scenario in regard to this picture, active Site1 suffers an outage, the latest updates are available on disk in Site2, and then the synchronous data can be replicated into target in the normal or non-ZDL system. The ZDL solution only supports workloads that are using Db2 data. Now to implement model 4, considerations include two GDPS controllers which are independent monoplexed systems with network access to all the systems in the environment. A common consideration for all replication server products is that of cross-site bandwidth. Insufficient bandwidth typically leads to an increase in the latency, RPO and might result in other desirable conditions such as IBM MQ queues filling up. IBM Multi-site Lifeline, this runs on the primary controller and communicates with Server Application State Protocol or SASP. To continue with considerations, the Active/Active, AA solution is on the workload level, which is an aggregation of software, data, and network activity. Db2 must be setup in data sharing mode even if there is only a single Db2 member in the data sharing group. Workload management is a key component of the z/OS operating system. It provides dynamic workload management within a single z/OS system or across multiple images within a parallel sysplex environment. Since the introduction of goal-oriented workload management, the ISPF-based, WLM administrative application or AA can define, view, install, and activate WLM service policies. GDPS CA includes much more than data replication. It also includes many other aspects of the environment such as sysplex, automation, network, workload routing, workload management, testing processes, and planned and unplanned outage scenario testing. Most installations do not have all of these skills readily available. It's rare to find a team with this range of skills across many implementations. The most successful GDPS projects are those projects in which IBM and client skills form a unified team to perform the implementation. GDPS CA provides advantages for continuous availability. Workloads benefit from a very low RTO, recovery time objective. Remember that the RTO is not just the time that is needed to bring up a backup server, it also includes the time to recover the data, prepare the physical facilities, switch the network, and bring up the operating system, databases, management tools, and finally the applications. There are some planned outages which just can't be avoided. Examples of software migration, upgrading Db2 or migrating z/OS from one version and release to another. Hardware upgrades can involve moving input/output devices or even channel cards from one site to another. Just like upgrading Db2, there are many applications which can be avoided for a planned outage. With these examples, switching any effective workloads to one site or the other is defined as upgraded or changed. In a GDPS CA solution scenario, the GDPS CA concept consists of having two sites that are separated by virtually unlimited distances, running the same applications and having the same data to provide cross-site workload balancing and continuous availability and disaster recovery. This change represents a fundamental paradigm shift from a failover model to a continuous availability model. Locate does the RS2 right below Sysplex A in the diagram here. The only available leg for HyperSwap is RL1, and it is enabled for HyperSwap. After HyperSwap is initiated and completed, RS2 becomes the primary DASD, and DASD RS3 takes over as soon as possible. The results to zero data loss, HyperSwap duration in the second level, and Site1 is still up and active. Triggers let application scripts run a function automatically when a certain event, like opening a document occurs. Installable triggers offer more capabilities and simple triggers but must be activated before use. For both types of triggers, application scripts pass the triggered function and event object that contains information about the context in which the event occurred. For example, if there are two sysplexes, sysplex A and sysplex B, then to switch scripts would be coded. Those are unplansite_sysplexB which GDPS would initiate automatically if it performs an unplanned site switch from sysplex A to sysplex B, and then unplansite_ sysplex A, which GDPS would initiate automatically if it performs an unplanned site switch from sysplex B to sysplex A. The results are zero data loss. HyperSwap duration in the second level, an RTO can achieve switched duration, seconds or minutes, depending on the number of workloads and Site2 workloads are activated. GDPS Continuous Availability, is a software-based asynchronous mirroring solution that is designed to maintain a consistent copy of data within IMS, within Db2, and within VSAM data at virtually unlimited distances with minimal impact to application response time. Since it's a software-based solution, GDPS Continuous Availability is disk independent. Both an active standby configuration and inactive query configuration to offload read-only workload to the recovery site supported. Application development services can help gain better efficiency and provide amazing user experience in critical applications. Testing is critical for software replication. Performance tuning and optimization is usually a repetitive approach to making and then monitoring modifications to an application and it's database. It could involve adjusting the configuration of the database and server or making changes to the applications. It is recommended that a maintenance policy for all sites are developed and implement a proactive maintenance strategy whereby regular maintenance is applied. When switching between sites, it's best to utilize manual instead of automated response for control. A disaster planning document should be developed in conjunction with the business continuity plan. Priorities and recovery time objectives for information technology should be developed during the business impact analysis. Technology recovery strategy should be developed to restore hardware, applications and data in time to meet the needs of the business recovery. Now the chart on this slide shows how the layers relate to the four resiliency deployment models. Notice that model 1 is a high risk, an impact to the customers business. It's estimated it would take eight plus hours to recover from an outage. For the system to be totally restored, there are many days involved. By following the models, model 4 shows a minimum impact, minutes for recovery and also minutes, but total restoration.
