GSE Conference – what I learned on Wednesday

Last time, I was talking about the sessions I attended on the Tuesday at the excellent GSE UK Conference in the first week of November. This time, I want to tell you what I learned from the sessions I attended on the Wednesday.

At the 10:15 session, I was speaking in the AI stream, looking at the brain and what we mean by ordinary intelligence, before people start talking about artificial intelligence. The session was well-received, and I was asked to give it again at lunch time to some people who were unable to attend.

After the coffee break, I saw IBM’s Lih M Wang’s presentation entitled, “AI for IT Resiliency Use Cases”. She started by suggesting that we are in a new era of computing. The challenges facing IT Operations include:

• Digital transformation with exponential business growth / cost. There are billions of transactions per day with unpredictable resource demands. And there are millions of Log and SMF records per day, but which indicators shouldn’t be missed?
• Complexity of business applications across hybrid cloud. There are multiple components across platforms making it difficult to isolate problems. And there’s the impact of any changes, eg hardware / software / application changes.
• Knowledge and skills gap for IBM zSystems. There’s limited cross-domain SME compared with number of systems managed. Plus, people need to know about the topology, inter-relationships, and dependencies.

Lih said that customers are asking: “Can AI-ML help?” They are looking for early warnings or sick symptoms. They also need to identify anomalous behaviour. Anomaly Analytics on IBM zSystems can: transform unstructured (SYSLOG) data into insights; turn performance metric (SMF) data into operational dashboards; and accelerate problem prevention by leveraging Machine Learning. IBM’s maxim is: Proactive, Prevent, Optimize.

Lih explained the difference between threshold monitoring and a Machine Learning (ML) baseline model. Basically, threshold monitoring is static, whereas ML can recognize what’s significant at much lower levels of activity. Using IBM Z Anomaly Analytics with Watson (ZAA), Machine Learning, and Enterprise Data Warehouse (EDW), data can be collected. Then, running typical workloads, the model can be trained. Thirdly, it can be scored by comparing models of expected behaviour with metrics. Visualization of the metrics shows how well the model runs on its own. Lih then showed how this worked with various scorecards for CICS, Db2, IMS, and MQ. Using colours, it becomes very easy to see where anomalies are occurring.

Real-time insights mean that the system will generate events when specified metrics exceed an anomaly threshold. Events will be shown in the main Problem Insights panel along with other events such as key single message events. Customer can select which of the KPIs to monitor for events and the threshold to use. Events can be forwarded to event monitors such as Watson AIOps. Selecting the Evidence column will take you to the scorecard with that KPI and time period open. Lih went on to give some customer examples.

I wanted to see BMC Software's Dave McCain's presentation called, "How can we use AI and user behaviour for better security monitoring". Unfortunately, I had a meeting. I hope to catch it another time.

The last AI stream session of the day was IBM Champion’s Henri Kuiper’s Jeopardy game. Jeopardy is a game that gives you the answer and you have to come up with the right question. For example, “This British mathematician and computer scientist is often considered the father of theoretical computer science and artificial intelligence. The answer is Alan Turing. Or, how about, “The 1956 workshop held at Dartmouth College which is often considered the birth of AI as a field”. The answer is, “What is the Dartmouth Workshop?”. Try this one, “This type of Machine Learning algorithm is inspired by the structure and function of the human brain and is used for tasks like image and speech recognition”. The answer is, “What is a neural network?”

There were other questions about unsupervised learning, GPT3, transfer learning, OpenAI, and so much more about AI and its history.

All-in-all, the GSE conference provided great education, brilliant company, and isn’t to be missed. I heartily recommend it to everyone who has an interest in mainframes. See you at the next one!