Cyber Infrastructure 6.0 – What's New?
What's New: Vaults
Improved object management in S3 storage
Cyber Infrastructure 6.0 introduces support for S3 Lifecycle—managing the life cycle of objects in a bucket, allowing you to configure automatic deletion of objects or versions of objects according to specified conditions and schedules—and S3 Tagging—managing object tags to support Bucket Policy.
Managing Object Tags to Support Bucket Policy
Object tag management provides support for setting rights to actions with buckets, objects, and object groups. For users, this enables the organization of a mandatory management model: setting access levels to storage resources for different categories of users. For example, you can provide read-only access, read and write access, etc.
Adding and removing policies is done using standard methods of interacting with object storage via the S3 API, for example using compatible client applications – CyberDuck, boto3, s3cmd, etc.
Policies are described in a document in JSON format. The PutObjectTagging, GetObjectTagging, DeleteObjectTagging, PutBucketTagging, GetBucketTagging and DeleteBucketTagging methods are implemented.
S3 Lifecycle Support
S3 Lifecycle support provides management of storing objects in S3 storage. We support deleting objects/versions after a specified number of days or on a specified date, as well as storing no more than n versions of an object in a bucket with versioning.
For users, this means easier management of the lifecycle of objects in S3 storage. The PutBucketLifecycleConfiguration, GetBucketLifecycleConfiguration and DeleteBucketLifecycleConfiguration methods have been implemented.
Future versions of Cyber Infrastructure will feature the ability to move objects from one class of storage to another class of storage.
Support for external storage systems for virtual machines
Importing external storage systems allows you to place virtual machine disks on external classic storage systems connected via the iSCSI protocol (support for the Fibre Channel protocol is planned for version 6.5). For each external storage system, you can create several connections with unique parameters and create new virtual machine storage policies based on them. It is possible to use the unique functionality of the storage system by setting connection parameters.
This version of the product supports the following storage systems:
Huawei:
OceanStor V3 V300R006C60/V300R006C61
OceanStor V5 V500R007C61/V500R007C70/V500R007C71
OceanStor V6 6.1.3/6.1.5/6.1.6
OceanStor Dorado V3 V300R002C10/V300R002C20
OceanStor Dorado V6 6.1.0/6.1.2/6.1.3/6.1.5/6.1.6
HPE:
3PAR StoreServ 7000/8000
P10,000
P7000
F-Class
For information on configuring Huawei storage systems, please refer to documentation of driver version 2.6.4and about setting up HPE 3PAR storage systems – in driver documentation.
Users now have the option to choose storage for virtual machines: Cyber Infrastructure storage or external storage.
Bulk configuration of storage disks
Cyber Infrastructure 6.0 has an expanded disk management interface. It now allows you to select a group of disks to assign roles to or reset a previously assigned configuration.
This can significantly reduce disk management time, especially when deploying a cluster with a large number of disks and nodes.
What's New: Virtualization
Implementation of EVC in a Computing Cluster
Version 6.0 introduces the ability to manage CPU emulation settings for virtual machines via a web panel, as well as the ability to enable and disable protection against the Spectre and Meltdown hardware vulnerabilities.
The CPU emulation mode of a virtual machine determines which model will be assigned and what additional features will be available to the virtual CPU. The CPU emulation mode set at the compute cluster level is used by default for newly created virtual machines. It can be overridden at the individual virtual machine level if necessary.
EVC functionality for configuring virtual machine CPU emulation allows you to choose a balance between the ability to migrate VMs between cluster nodes and performance in a cluster of mixed hardware. There is support for host model, passthrough, and custom emulation modes with a point selection of additional instructions.
Controlling nested virtualization support for virtual machines should be used if you need the ability to create virtual machines within virtual machines. Nested virtualization is not supported for the qemu64 CPU model.
What's New: High Availability
Setting up control panel failover
In version 6.0, the control panel can be replicated across three, five, or seven nodes in high availability mode, which corresponds to single-, dual-, and triple-node cluster fault tolerance. If a control server fails or becomes unreachable over the network, an instance of the control panel on another server will take over the control panel service so that it remains accessible over the same dedicated IP address. Failed control servers can be replaced in groups of one, two, or three nodes.
High Availability Service Settings
Cyber Infrastructure 6.0 implements hardware failure threshold management at which services are migrated. It is possible to adjust the failure threshold for starting the migration of S3 and VM services, as well as adjust the failure threshold for starting data recovery.
New high availability settings enable you to improve the fault tolerance of storage and compute services.
Expanded capabilities for creating a metro cluster
The new version of the product now provides full metro cluster configuration from the control panel. Replication scheme options of 4 replicas and efficient redundant coding schemes with an increased number of parity blocks (3+5, 7+5, etc.) and fault tolerance are included. Configuration of the cluster sensitivity level to node failure and support for writing to storage after complete loss of redundancy are implemented.
These innovations provide users with the ability to create highly available metro clusters.
Improvements and refinements
With each new version of the product, we not only expand its capabilities, but also improve and refine existing functions.
Improvements in working with NVMe drives
Cyber Infrastructure 5.5 introduced the ability to use NVMe Namespaces functionality. NVMe Namespaces allows a single physical disk to be divided into multiple smaller logical disks. This can significantly increase the performance of a data storage cluster by dividing NVMe disks into namespaces and using the resulting logical disks to store data fragments. The performance gain is due to the fact that the same physical disk is used by multiple data fragment services, although this increases the consumption of RAM, CPU cores, and disk space due to the greater number of data fragment services per physical disk.
In version 6.0, support for splitting physical disks into logical disks has been improved. The loss of space when splitting disks into namespaces has been reduced from 9% to 1% (for 960 GB, 1.92 TB, 3.2 TB disks), and the speed of splitting disks has been increased.
Adaptive Memory Manager
Version 6.0 implements an adaptive memory manager on nodes. The service monitors the RAM load profile and defragments it in advance to improve the response of all system services. This reduces the likelihood of memory overflow (Out of Memory, OOM) and increases the limit of the withstandable load on the storage during random access. For customers, this means more stable operation of the entire system and increased performance of computing services.
More resources for VM
Cyber Infrastructure 6.0 introduces the ability to create virtual machines with high resource requirements — the number of virtual CPUs and memory. Now a virtual machine can have up to 4 TB of RAM and up to 240 virtual CPUs.
How to update to a new version
Cyber Infrastructure 5.5 can be updated to version 6.0 in the Settings > Update section. Node reboot is not required.
What we will do in version 6.5
Our product continues to actively develop and here is a far from complete list of functions that will appear in Cyber Infrastructure version 6.5: increased speed of loading large objects into S3, acceleration of redundant coding operations and formation of checksums on the Intel platform, support for connecting external storage systems via the Fibre Channel interface, S3 Bucket Notification, a new S3 interface, support for Transition and Tagging in S3 Lifecycle, copying disk configurations between nodes and many other improvements.
On the day of the release of Cyber Infrastructure 6.0, July 11, 2024, we held a webinar where we talked in detail about the new features in the product, shared plans for the next release, and answered numerous questions from listeners. In addition, colleagues from Depo Computers talked about the new hardware and software complex DEPO Atlas Alpha — a solution for server virtualization and deployment of virtual workstations, built using Cyber Infrastructure. Recording of the access event Here.
