Real-Time Visualizations
A major benefit of Coherence is its design and configuration flexibility to meet a wide variety of high performance computing applications. But how do you know you have the best design and most efficient Coherence implementation for your application mix? How can you quickly see what is happening in the data grid as your data load increases in real-time? Is the design and use of Coherence optimized across different application cycles?
Evident ClearStone Live provides a set of dashboards and visualizations that provides data grid insights as follows:
Data Grid Health Dashboard
The Data Grid Health dashboard provides a macro view of the health and status of a data grid/cluster. This visualization is user configurable on the fly and can display any of the supported Key Performance Indicator (KPIs) metrics.
Data Grid Health:
Common Questions Addressed/Key Performance Indicators (KPIs)
- How many storage nodes are being used vs. all available nodes
in the data grid?
- What’s the total memory allocation vs. used memory across the
data grid?
- How many TCP-Extend sessions are currently active?
- How many proxy nodes do I have?
- What is my total storage memory available and how much is
being used?
Event Viewer
The Event Viewer displays a summary of data grid level events with timestamp, severity, and description. This includes node events (e.g. when a nodes joins or leaves the data grid, or when a node status changes), cache events (e.g. when a cache is created or its status changes), TCP Extend events (e.g. when a proxy node joins/leaves data grid), TCP-Extend session (e.g. when a client joins/leaves the data grid), etc. The events are time correlated so that events can be related to relevant changes in other KPI’s (“cause and effect”).
Event Viewer:
Common Questions Addressed/Key Performance Indicators (KPIs)
- Are any SEVERE or WARNING events occurring?
- Did any storage or proxy nodes leave the cluster and if so which nodes?
- What are the most recent events in the cluster?
Events can be color coded so that more severe events are highlighted in the event viewer table.
click to enlarge
Cluster Performance Charts
This multi-chart visualization trends an entire cluster’s capacity in two dimensions (node counts and storage memory utilization). It allows the user to simultaneously and synchronously view and compare two time-series charts, each with different cluster KPIs, in real-time. The top chart supports KPI’s related to node counts (supply) and the bottom chart support KPIs related to a cluster’s memory utilization (demand). These charts are annotated by the cluster level events to illustrate the cause and effect of data grid changes over time.
Cluster Performance Charts:
Questions Addressed/Key Performance Indicators (KPIs)
- Have there been any changes in storage or proxy nodes over the
past N hours?
- How is available memory affected as storage nodes are added or removed form the cluster?
- What’s the memory utilization over time?
- At what points does memory capacity become saturated?
Cache Performance Charts
This visualization provides views of the performance and properties for named cache in a data grid. The performance data is presented in a tree–table and a time-series chart. Users can organize the rows in the table by grouping on a specific named cache attribute. Users can sort, filter, and adjust the columns to be displayed in the table and select a KPI to trend for one or more named caches over time.
Cache Performance Charts:
Common Questions Addressed/Key Performance Indicators (KPIs)
- What’s the level of activity for a named cache?
- What are the number of objects in each cache and how much memory is being consumed for each cache by service?
- How many nodes are associated with each cache per service?
Additional columns that can be selected include: Status, Storage Enabled, Persistence Type, # of nodes, queue delay, queue size, Total Gets, Puts, Hits, Misses, Average Gets ms, Puts ms, Hits ms, Miss ms, Hit Probability, Total Get Time ms and Put Time ms
Data Grid HeatMap
The Data Grid HeatMap visualization provides a versatile view of active named caches. A heat map display rectangles (“cells”) that can be arranged, sized and colored to graphically reveal underlying data patterns and allow users to easily recognize complicated data cache relationships that are otherwise not obvious. Each cell in the heat map represents an individual named cache partition. Named caches can be grouped by any of the following attributes: Physical host, member, service type, cache type, persistence type, etc. The cell size is proportional the measurement of a particular, configurable cache KPI.
Data Grid HeatMap:
Common Questions Addressed/Key Performance Indicators (KPIs)
- How are named caches distributed across the different nodes in a cluster?
- What’s the memory utilization of the named caches across the cluster?
- Which caches have the most objects?
- Which caches are consuming more than N (configurable) GB’s of memory? i.e. who are the biggest consumers?
KPI supported include: # of nodes and size (total objects), Memory used, Total Gets, Hits, Misses, Puts, and many more…
Data Grid Session Dashboard
This dashboard provides a view of the active TCP-extend sessions in the data grid. For each session, it displays a real-time trend of the session’s behavior. Specifically, it displays network-oriented metrics like bytes and messages transferred. The sessions can be filtered and group by client or proxy network addresses. This dashboard can be configured to view active session information, (network volume and statics), top N-clients (MB sent/received), and session history (client connect time, aggregated duration, MB and messages sent/received, etc.). Once a session has disconnected, the summary of a TCP-Extend session will be displayed in the disconnected sessions view.
Data Grid Session Dashboard:
Common Questions Addressed/Key Performance Indicators (KPIs)
- How many concurrent sessions is managed by each proxy node?
- Who are top clients?
- Does sufficient network capacity exist to handle the load?
- What are the most active sessions?
- Are there any “stale” sessions?
- Are there any sessions with unusual loads?
How many and what are the characteristics of all sessions (including expired sessions) over the past 24 hrs?
Data Grid Node Dashboard
This dashboard provides real-time monitoring and analysis of data grid members. The user can select from any available node via a drop down menu. The node property view displays the node’s runtime cluster properties. The main visualization window contains the real-time visualizations for up to four simultaneous charts, graphs, and/or tables visualization for the selected node).
Data Grid Session Dashboard:
Common Questions Addressed/Key Performance Indicators (KPIs)
- How are the named caches partitioned for a specific node?
- What’s the memory utilization for a specific node?
- What’s the network utilization (bytes/packets transferred) for a
specific node?
- What’s the frequency of access for each named cache for a
specific node?
- What’s the session connectivity trend for a specific TCP-Extend node?
What’s the total TCP-Extend network utilization for a specific TCP-Extend node?
click to enlarge
Node HeatMap
The Node HeatMap is an analysis tool used to visualize all the cluster members of the monitored cluster. This type of visualization can be helpful in identifying the hot spots in the cluster from a VM perspective. The cells in the heat map represent a cluster member (JVM).
Node HeatMap:
Common Questions Addressed/Key Performance Indicators (KPIs)
- Are any of my nodes running out of Memory?
- Are my publisher and receiver success rate falling below a threshold?
- Is memory utilization consistent across the cluster?
click to enlarge
|