ZStack Cloud 5.1.0

On April 30, 2024, ZStack Cloud officially released its latest version, ZStack Cloud 5.1.0, covering a number of important features described in details below.

Highlights

Allows elastic baremetal instances to connect with a new-generation high-performance distributed storage to enjoy a higher-performance and lower-latency storage service.
Supports IPv6 protocol in all LBs and HA in dedicated-performance LB, improving network security and business stability.
Allows you to modify the type and VLAN ID/VNI of L2 networks online, improving network O&M flexibility.
Supports hot migration and HA for VM instances with VF NICs attached, protecting business continuity.

Overview

VM Instance

Enhancements on the VM memory reclaim mechanism.
1. Allows you to enable/disable Memory Balloon and Kernel Same-Page Merging for a VM instance individually.
2. Provides a host memory monitoring that displays the memory reclaimed/shared/saved in a more intuitive way.
Improvements on VM HA mechanism, making HA reboots no longer occur after you stop VM instances manually or through a scheduled job.
Allows you to limit the migration rate during VM storage migration.
Supports hot migration and HA for VM instances with VF NICs attached.

Cloud Storage

Allows elastic baremetal instances to connect with new-generation high-performance distributed storage.
Supports the lifecycle and snapshot management of block storage volumes based on high-performance Vhost primary storage.
Allows you to hot migrate a VM instance across LocalStorage.

Cloud Network

Load balancing enhancements.
1. Adds support for IPv6 protocol in LB, improving network performance and security.
2. Dedicated-performance LB now supports HA, ensuring business stability.
3. Improvements on LB listeners.
A new network service, shared bandwidth, allowing multiple VM instances to share bandwidth resources.
Allows you to modify the type and VLAN ID/VNI of L2 networks online, improving the network O&M flexibility.
No longer uses 223.5.5.5 as the default DNS address for L3 networks.

Platform O&M

Allows you to obtain the monitoring data of more Cloud resources through SNMP.
Allows you to view HA log and auto-scheduling log details.
Volume export improvements.
1. Allows you to export information of data volumes and root volumes simultaneously.
2. Allows you to customize the information items to be exported.

VM Instance

Enhancements on the VM memory Reclaim Mechanism

In previous versions, ZStack Cloud has provided memory reclaim and sharing mechanisms that help ensure the efficient use of host memory through memory dynamic allocation and de-duplication.

Starting from 5.1.0, ZStack Cloud further enhances the memory reclaim and sharing mechanisms, making the mechanisms can take effect in a finer granularity and providing a more intuitive monitoring to displays the host memory changes.

Allows You to Enable/Disable Memory Balloon and Kernel Same-Page Merging for an individual VM Instance

Memory Balloon: The setting called Memory Reclaim in previous versions is renamed as Memory Balloon in 5.1.0. This mechanism monitors the host and VM memory utilization in real-time and ensures a high host memory use efficient by allocating/reclaiming memory to/from VM instances dynamically.
- Reclaim: If the VM workloads decreases, a part of its free memory is reclaimed to the host when the host memory utilization is higher than 80%. The reclaimed memory can be allocated to those VM instances with increasing workloads.
- Allocate: If the VM workloads increases, the host allocates more memory to it to ensure the business stability.
Kernel Same-Page Merging: A mechanism realizing the host memory sharing. This mechanism scans the VM memory page periodically, detecting whether multiple VM memory pages are mapped to physical memory pages with identical content and re-mapping these VM memory pages to one physical memory page. This allows multiple VM instances to share host memory and reduce resource wastes.

The two mechanisms mentioned above ensure a high use efficient of host memory. However, for VM instances that are running important businesses, these mechanisms may bring a performance influence. Starting from 5.1.0, you can enable/disable Memory Balloon or Kernel Same-Page Merging for a VM instance individually on its action list or details page. VM instances with Memory Balloon and Kernel Same-Page Merging disabled can enjoy more dedicated and stable memory spaces.

You can enable/disable these two mechanisms flexibly for different VM instances according to your business requirements, improving the physical memory use efficient as well as ensuing the stability of your core businesses.

Figure 1. Allows You to Set Only Memory Reclaim (Memory Balloon) for VM Individually before 5.1.0

Figure 2. Allows You to Set both Memory Balloon and Kernel Same-Page Merging for VM Individually in 5.1.0

Provides a Host Memory Monitoring that Displays the Memory Reclaimed/Shared/Saved in a more Intuitive Way

Starting from ZStack Cloud 5.1.0, you can view the host memory reclaimed/shared/saved through the Memory Balloon and Kernel Same-Page merging mechanisms in a more intuitive way on the host details page. The following metrics are displayed on the monitoring dashboard (Memory Reclaim):

Reclaimed through Memory Balloon: Displays the memory size reclaimed from VM instances with decreasing workloads.
Shared by VM Instances: Displays the memory size that is shared by multiple VM instances because of the Kernel Same-Page Merging mechanism.
Economized through KSM: Displays the memory size saved because of the Kernel Same-Page Merging mechanism.

Figure 3. Intuitive Host Reclaimed/Shared/Saved Memory Monitoring

Improvements on VM HA Mechanism, Making HA Reboots No Longer Occur after You Stop VM Instances Manually or Through a Scheduled Job

HA mode is a mechanism that defines whether a VM instance reboots automatically after it is stopped. In previous versions, ZStack Cloud supports two VM HA modes: None and NeverStop.

None: The VM instance does not reboot automatically after it is stopped, no matter the stopping is scheduled or unexpected.
NeverStop: The VM instance attempts to reboot automatically after it is stopped, no matter the stopping is scheduled or unexpected.

ZStack Cloud 5.1.0 improves the NeverStop HA mode, making it triggers an auto VM reboot only when the VM instance is stopped unexpectedly. If the VM instance is stopped manually or through a scheduled job, it does not automatically reboot even though its HA mode is set as NeverStop.

Here provides more details of the NeverStop mechanism after the improvements:

Unexpected VM Shut Down: If the VM instance is stopped unexpectedly because of it own errors, it attempts to reboot automatically. If the VM instance is stopped unexpectedly due to the faults of the compute, storage, or network it is residing on, it attempts to migrate to another host and reboot automatically.
Manual or Scheduled VM Shut Down: The VM instance does not reboot automatically if you stop, force stop, power off the VM instance manually or make it stopped through a scheduled job.

Figure 1. VM Instance Does Not Reboot Automatically After it is Stopped Manually

Figure 2. VM Instance Does not Reboot Automatically After it is Stopped Through a Scheduled Job