Virtualization Technology News and Information
Go Multi-Cloud, But Get Visibility Into Your Workflows

By Dave Wagner, SolarWinds

It can seem like a good idea to take advantage of your cloud provider's proprietary software features, services, and APIs: it can keep things simple, and the ramp-up time to market might feel faster and easier.

After the standard hiccups in getting started, your company-let's say it's in retail-uses these cloud software features to quantifiably improve customer service on the web and in brick-and-mortar locations. There might be a happy ending to this story, but it also means you're locked into your provider and have less leverage to negotiate price once your contract is up.

Like all expenses, cloud costs rise, and because you haven't architected your applications to be cloud-agnostic, you may now be in a difficult situation. And given COVID-19, the timing is terrible: you need operational maneuverability and you need to contain costs right away.

On the provider side, the cloud sector has steadily consolidated, leaving us with the "three horsemen": Amazon Web Services®, Google Cloud PlatformTM, and Microsoft® Azure®. Going with these providers, Oracle, IBM, or many of the other smaller competitors may help you gain slight incremental speed or flexibility for workload deployment, but as the number of platforms increases, the speed gains yield ever-diminishing returns. It's not about gaining even more speed (they're all quick). It's about getting a better price.

There's nothing inherently wrong with using your cloud provider's unique offerings, but proprietary features inhibit the interoperability and portability of applications. The platform providers do this purposefully: they're trying to differentiate themselves from their competition based on features other than speed and ease of adoption. It's not coincidental, however, that without these proprietary features, providers know the "stickiness" of their platforms is low. Customers can easily move from one cloud provider to another.

The Basis of an Effective Strategy

Effective cloud strategies generally call for companies to adopt a multi-cloud approach, using applications architected to be entirely independent of the platform providers. But taking on a multi-cloud strategy isn't a simple matter.

Multi-cloud gives customers as much leverage as possible when negotiating contracts for cloud capacity-you put one platform provider against another and have them fight it out for an optimal price.

Each business has specific cloud requirements. Still, with some areas in lockdown and other regions opening up for business, your hypothetical retail outfit may need to develop technologies capable of allowing your customers to opt for the increasingly popular BOPIS model: buy online and pick up in store. More than two million Target® customers have tried the feature, and Target says sales were up 1,000% for BOPIS this past April when compared with a year earlier.

This adaptation brings in much-needed revenue. But your retail outfit must still slash costs while investing in infrastructure and new apps to expand upon your ability to compete in a radically changed environment. It's something of a whirlwind.

In short, every dollar counts.

Think in Terms of Operational Flows

Your retail outfit has produced a new application allowing consumers to purchase a new TV via the BOPIS model. As the consumer arrives to pick up the item (wearing a mask and parked in a loading zone while their family waits in the car), this transaction should be seamless and fast. But if your cloud strategy isn't thought through, it may not be.

With some specific exceptions, you should put all the related apps-from a business and transactional workflow perspective-in the same cloud instance. Though it's possible to structure individual transactions to work across different cloud instances, doing so will be slower (physics) and perversely riskier (you now have two points of failure instead of one). If half of your data supports an application in Microsoft Azure and the other half supports one in Google Cloud, you can federate across. But this involves latency issues, and if both are required to complete the transaction, both must be available and perform well.

You'll be adding extra time because now you're going all the way out to one cloud and then back to another. At best, you're doubling the latency issue. And even though data transfer into any given cloud platform is typically free, data transfer out can become costly if large volumes of data are involved.

So, here's a rule of thumb: think in terms of operational flows. Every time you have to go across clouds, things get slower. This is especially important with respect to the data storage aspects of the application transaction workflow. Be careful how you architect your applications and data storage strategy.

With this said, some operations intentionally put data in different clouds as part of a replication strategy, whether it's for compliance or risk mitigation. This is done for specific reasons (and it's generally done in a 100% redundant manner, which reduces risk and significantly increases cost).

Having your information across an ever-growing number of clouds is a good data strategy.

But you must still monitor your environment and workloads. You need to know where your apps are slowing and where processes could be improved or eliminated for cost savings.

A Single Pane of Glass

To proactively manage multi-cloud performance, you need application performance management (APM). APM can help you monitor the full cloud application and infrastructure stack for application availability, performance, and throughput as well as the error rate of the transactions within the application.

Your operations are complex, but APM can give you broad and deep visibility into your app performance, from end-user experience to queries into database response times and all the cloud infrastructure in between. It can work across clouds, giving you single pane of glass visibility into your cloud operations.

At the bare minimum, standardize one APM solution for one cloud environment. Even better, use a single APM solution to monitor all your apps consistently across your cloud platforms instead of using one for Google Cloud and another for Azure.

Working with two APM solutions means you now have two dashboards and no transactional workflow across the elements spanning the environments those different APMs are monitoring. You have no way of looking at those operations holistically because you're having to shift between two different tools for each portion of the transaction workflow.

Because these are tough times, it's possible to fall back onto the myth that APM is challenging to implement and expensive. It isn't. In the past, APM required application expertise in the on-premises and hybrid world-this was back in the days of monolithic applications when only application developers could instrument applications for an APM solution.

Cloud implementations of APM don't require significant application-specific expertise. These applications are written in modern languages, using microservices deployed in containers and cloud services. Most of them can even be auto-instrumented: you simply configure a library automatically loaded at runtime, and the APM solution does the rest.

And APM is no longer expensive. This was only true in the old days when the original vendors of APM sold to application developers because they were building mission-critical custom apps.

APM is now less expensive because the underpinnings of most modern apps are open-source and commoditized. These building blocks have natural management hooks built into them, so APM tools can connect to them directly.

Speaking of expense, APM solutions are also highly valuable for your business when you must look for opportunities for cost efficiencies. They can identify the specific infrastructure resource capacity required to ensure acceptable transactional performance, allowing for cost consolidation efforts to be conducted at minimal service risk.

This odd era forces us to look for easy savings, but be careful with free applications: they could end up costing more than you expect. Instead, apply a strategy and then measure and monitor.


About the Author

David Wagner 

Dave Wagner is a senior manager of product marketing at SolarWinds with over 20 years of experience focused on IT performance and optimization. In his current role, Dave works closely with product management and corporate marketing to ensure alignment in go-to-market strategy and messaging for the SolarWinds application performance monitoring (APM) products. Prior to joining SolarWinds, Dave served as CTO of OpsDataStore, business development principal for TeamQuest, and vice president, marketing and sales at Solution Labs Inc. 
Published Wednesday, September 02, 2020 7:35 AM by David Marshall
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
<September 2020>