Multi-cloud management: Challenges for tech, people, processes

Andy Jassy (AWS); Satya Nadella (Microsoft) and Thomas Kurian (Google Cloud)

When it comes to managing hybrid and multi-cloud environments there are many options but no easy path nor lack of challenges.

While cloud computing has been around in some form for more than a decade, tools to manage its current enterprise iterations from private, on-premises, or public locations are still evolving at a rapid rate. Gartner says that more than 90 vendors—including IBM and Red Hat, VMware, CloudBolt, Flexera, Scalr, Cisco, and Nutanix—offer varying degrees of cloud-management capabilities.

While there are many options, organisations struggle to effectively manage a multi-cloud environment, said Roy Ritthaler, vice president of product marketing, Cloud Management Business Unit, VMware.

“With workloads deployed in multiple public clouds, multi-cloud Kubernetes, private cloud/data centres and edge locations, most organisations find it challenging to get a unified view of the health of their environments as well as manage costs, ensure security and improve operational governance while automating core processes,” Ritthaler said.

This is not just a technology challenge, but also a people and processes challenge, Ritthaler said. “Lack of unified provisioning tools, siloed operational visibility, lack of holistic performance and cost insights and interoperability and integrations issues mean siloed resources, fragmented teams and management tool proliferation.

Multiple personas are involved—IT Operations, DevOps/ Developers, Finance and Line of Business (LOB) leaders—requiring extensive training, collaboration, and process changes as organisations embrace the cloud model.”

Recent IDC research found that most enterprises expect that they will need net-new multi-cloud-management tools to keep up with their emerging business and infrastructure-operations demands.

“Multi-cloud architectures are introducing a new wave of management complexity as developers and business groups implement cloud services and tools that best align with their application and business innovation road maps with limited regard for corporate preferences. The introduction of containers, microservices, and Kubernetes creates further complexity.” IDC stated.

Over the next two years, enterprise decision makers are expected to prioritise investments in analytics, performance monitoring and reporting, capacity optimisation, cost management, as well as automation and self-service to augment management capabilities for multi-cloud and governance, according to IDC.

“These management tools are deeply interconnected. Cost decisions must be made in the context of capacity requirements and application performance,” IDC stated.

There is also enterprise anxiety about application-development density that stretches across different cloud providers. A recent Enterprise Management Associates study said there are 2,316 Python libraries related to Amazon Web Services (AWS), Microsoft Azure, and Google Cloud that developers download approximately 13 million times per day to 112 different, mostly Linux-based, operating systems.

“While individual projects typically stay within the boundaries of a single cloud, EMA also sees an increasing number (approximately 10 per cent) of projects stretching across multiple clouds. The rapid growth of microservices increases this trend and at the same time emphasises the urgent need for a unified governance and management layer for both developers and IT operators to contribute to optimising release efficiency and operational reliability at the same time,” EMA stated.

Such a wide variety of projects has caused many customers to look for help managing workloads across many environments, which requires multiple consoles and tools, IDC stated.

“As enterprises manage multi-cloud environments and the number of consoles and tools grows, it’s common to experience challenges that stem from siloed data—an inevitable, common side effect of migrating applications running on legacy systems to disparate cloud environments,” said Briana Frank, director of product management with IBM Cloud.

“As enterprises move disconnected data from one cloud to another to be used by various applications, they often experience performance issues and significant cost increases up to 300 per cent, according to IBM research.”

Customers complain about using multi-cloud services saying they increased their costs dramatically due to data-transfer between clouds and increased IT staffing, said Douglas Gourlay, vice president and general manager of cloud-networking software at Arista.

Getting a unified view among clouds

Also, as enterprises move to multiple clouds they quickly find that each cloud provider is unique, which adds challenges to manage those environments, “like network architectures, features, and scale, which creates a steep learning curve for customers to operate in the cloud and creates operational challenges across existing environments like data centre and campus networks,” Gourlay said.

In Arista’s case, the company offers CloudEOS and CloudVision software that enable network connectivity and management capabilities between private or public clouds.

With CloudEOS customers can operate multiple public clouds with a consistent operating model for all network abstractions—using the same runbooks and processes they utilise to operate their existing data centre and campus networks, Gourlay said.

“CloudEOS telemetry, coupled with CloudVision provides the time-series storage and analysis of the network state of a customers’ multi-cloud network. This lets the customer go back and check why and how an issue happened and reduces the return-to-operations time while enabling rapid root-cause analysis on initial failure detection.”

On the cost side, a separate Arista offering—CloudEOS Edge—supports dynamic path selection at the network edge that lets customers assign paths for applications to use with an eye toward reducing data transfer/synchronisation cost.

With unified EOS and CloudVision deployments across data centre, campus, and multiple public clouds, customers can support and manage the their multi-cloud strategy without doubling or tripling their team size or affecting their budget plans, Gourlay said.

Troubleshooting

In managing multi-cloud environments, another issue customers face is distinguishing between application-performance issues vs. network problems.

“What we hear from customers is a sense of a loss of control when in a multi-cloud environment,” said  Kaustubh Das, vice president and general manager of Cisco’s Cloud and Compute product group. “It’s hard to predict with certainty the impact of newly provisioned cloud services on the network.”

Cisco offers a number of packages targeting the issue, such as its cloud-based Intersight management platform.

In addition, Cisco’s AppDynamics application management package and recently acquired ThousandEyes technology, which offers a cloud-based software package that analydes local and wide-area networks and internet performance. The package is designed to provide  broad visibility and let customers pinpoint cloud and non-cloud problems with applications and the network.

Earlier this year Cisco integrated its AppDynamics enterprise application information with Cisco Intersight Workload Optimizer.  With it, customers can manage a variety of infrastructure components such as servers, configuration and policy management as well as telemetry and analytics.

The idea, Das said, is to let application and infrastructure teams see a shared view of infrastructure dependencies that affect application performance, user experience, and business impact—all from one location.

“The IT and DevOps teams can work together, using a shared vocabulary to pinpoint the root cause for application degradation, proactively prevent issues in real-time, set policies, and automate responses to solve app issues on-premises or in the cloud, regardless of domain,” Das said.

According to Arista’s Gourlay most customers find it extremely difficult to troubleshoot a network issue in the public cloud due to a lack of information and visibility, especially when troubleshooting requires packet-level observability.

“I wish I could tell you that the apps and network IT folks work together to make multi-cloud work more effectively, but currently that’s not where we are at,” said Nabil Bukhari, CTO and Chief Product Officer at Extreme Networks.

Help from AI/ML

Stitching together data from physical on-premises deployments and multiple clouds and applying analytics to it is a challenge but one that is key to managing this kind of environment, Bukhari said.

Extreme offers the ExtremeCloud IQ package that offers a machine-learning and AI-driven cloud-management platform that simplifies onboarding, configuration, monitoring, managing, troubleshooting, alerting, and reporting for network infrastructure devices.

VMware’s Ritthaler said the company’s vRealize Cloud Management includes capabilities to visualise the entire network, both virtual and physical. It uses machine learning to build network and application boundaries and can perform full path analysis across VMs, containers, and across hybrid and multiple public clouds. This offers an easy way to troubleshoot VM-to-VM connectivity across multiple clouds.

“This network topology map also helps optimise network performance with proactive alerting and anomaly detection for firewall misconfigurations, spikes, capacity constraints, etc," Ritthaler said.

"All these capabilities are presented in a single network map, regardless of whether the network is virtual or physical or both, and in the context of the application boundaries to make sure the networking, security, infrastructure, and application teams are speaking the same language."

IBM’s Frank says Big Blue addresses this issue with its  Application Performance Management (APM) package that helps customers distinguish between application-performance issues and network issues across on-premises, cloud-based and hybrid workloads—all from a single dashboard.

“In a multi-cloud environment, a solution that works equally well both on-premises and across multiple clouds is key to gaining full visibility and eliminating siloes,” Frank said.

“IBM APM solutions measure application availability and performance, automate actions to quickly recover from application or network performance issues, and provide visibility and tools to diagnose and fix problems before they impact operations or end-users’ experience with the application environment.”

In-house knowledge

There are other challenges to managing a multi-cloud world, observers say. One big one is having the staff with the skillset to manage multivendor offerings, Extreme’s Bukhari said. “Data formats and APIs are different and you need a team that can understand all of those things.”

The introduction of containers, microservices, and Kubernetes creates further complexity, IDC stated.

“Getting the full benefit out of any multi-cloud management portfolio requires organisations to make trade-offs and strategic investment choices. In fast-moving technology environments, it can be difficult to fully anticipate the impact of new processes, methods, and tools,” IDC stated.