{"id":37745,"date":"2019-10-31T22:19:28","date_gmt":"2019-10-31T19:19:28","guid":{"rendered":"https:\/\/prohoster.info\/blog\/balansirovka-nagruzki-v-openstack-chast-2\/"},"modified":"2019-10-31T22:19:28","modified_gmt":"2019-10-31T19:19:28","slug":"balansirovka-nagruzki-v-openstack-chast-2","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/administrirovanie\/balansirovka-nagruzki-v-openstack-chast-2","title":{"rendered":"Load Balancing in OpenStack (Part 2)","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

In the last article<\/a><\/noindex> We discussed attempts to use Watcher and presented a testing report. Such tests are periodically conducted for load balancing and other critical functions of large corporate or service provider clouds.<\/p>\n

The high complexity of the task may require multiple articles to describe our project. Today, we publish the second article in the series, dedicated to balancing virtual machines in the cloud. <\/p>\n

A Bit of Terminology<\/h3>\n

\nVmWare introduced the DRS (Distributed Resource Scheduler) utility for load balancing in the virtualization environment they developed and offer. <\/p>\n

As stated by searchvmware.techtarget.com\/definition\/VMware-DRS<\/a><\/noindex>
\n\"VMware DRS (Distributed Resource Scheduler) is a utility that balances compute workloads with available resources in a virtual environment. The utility is part of a virtualization package called VMware Infrastructure.<\/p>\n

With VMware DRS, users define rules for distributing physical resources among virtual machines (VMs). The utility can be configured for manual or automatic management. VMware resource pools can be easily added, removed, or reorganized. If desired, resource pools can be isolated among different business units. If the workload on one or more virtual machines suddenly changes significantly, VMware DRS redistributes the virtual machines among physical servers. If the overall workload decreases, some physical servers may be temporarily powered down as the workload is consolidated.<\/i><\/p>\n

Why is Load Balancing Necessary?<\/h3>\n

\n<\/a><\/noindex>
\nIn our opinion, DRS is an essential feature of the cloud, though this does not mean that DRS must be used always and everywhere. Depending on the purpose and needs of the cloud, there can be different requirements for DRS and balancing methods. There may be situations where balancing is not needed at all, or even harmful.<\/p>\n

To better understand where and for which clients DRS is needed, let's consider their goals and objectives. Clouds can be divided into public and private. Here are the main differences between these clouds and the goals of clients.<\/p>\n

Private Clouds \/ Large Corporate Clients
\nPublic Clouds \/ Medium and Small Businesses, Individuals<\/p>\n

Primary Criteria and Goals of the Operator
\nProviding a reliable service or product
\nReducing the cost of services in a competitive market<\/p>\n

Service requirements
\nReliability at all levels and in all system components<\/p>\n

Guaranteed performance<\/p>\n

Prioritization of virtual machines into several categories\u00a0<\/p>\n

Information and physical data security<\/p>\n

SLA and round-the-clock support
\nMaximum ease of obtaining the service<\/p>\n

Relatively simple services<\/p>\n

The responsibility for data lies with the client<\/p>\n

VM prioritization is not required<\/p>\n

Information security at the level of standard services, responsibility lies with the client<\/p>\n

There may be failures<\/p>\n

No SLA, quality is not guaranteed<\/p>\n

Email support<\/p>\n

Backup is not mandatory<\/p>\n

Client features
\nA very wide range of applications.<\/p>\n

Legacy applications inherited in the company.<\/p>\n

Complex customized architectures for each client.<\/p>\n

Affinity rules.<\/p>\n

Software operation without stoppage in 7x24 mode.\u00a0<\/p>\n

\u2018Hot\u2019 backup solutions.<\/p>\n

Predictable cyclic client load.
\nStandard applications \u2013 load balancing, Apache, WEB, VPN, SQL<\/p>\n

Application downtime is possible for a period of time<\/p>\n

Arbitrary distribution of VMs in the cloud is allowed<\/p>\n

Backup by the client<\/p>\n

Statistically averaged load predicted with a large number of clients.<\/p>\n

Implications for architecture
\nGeo-clustering<\/p>\n

Centralized or distributed storage systems<\/p>\n

Reservable SRK
\nLocal data storage on computing nodes<\/p>\n

Balancing objectives
\nEven distribution of load<\/p>\n

Maximum application responsiveness\u00a0<\/p>\n

Minimum latency for balancing<\/p>\n

Balancing only in case of explicit necessity<\/p>\n

Taking part of the equipment out for preventive maintenance
\nReducing service costs and operator expenses\u00a0<\/p>\n

Disabling part of resources in case of low load<\/p>\n

Energy savings<\/p>\n

Reducing personnel costs<\/p>\n

We draw the following conclusions for ourselves:<\/h3>\n

\nFor private clouds,<\/b>provided to large corporate clients, DRS may be applied considering the restrictions:<\/p>\n