getty
System efficiency has long been a guiding principle for modern tech organizations looking to control costs, streamline operations and improve performance. But when outages, cyber incidents or infrastructure failures occur, systems that have been optimized too aggressively can leave companies with little room to adapt.
Resilience requires a different kind of planning—one that accounts for disruption before it happens, even when that means accepting some added cost, complexity or operational friction. Below, members of Forbes Technology Council share the tradeoffs tech teams should consciously make to strengthen resilience and reduce operational risk.
To prioritize resilience, companies must sometimes sacrifice speed. AI is billed as an efficiency accelerator, but if that comes with cutting corners, the result can lead to risks down the road. Prioritizing architectural and security fundamentals like disaster recovery, process redundancies, multifactor authentication and other safeguards will reduce risk in the long-term, even if speed isn’t priority No. 1. - Bryan Krieger, RapidScale
Shift the focus from uptime to recovery time. Many teams obsess over preventing failure but overlook how fast they can recover from it. A system that will fail cleanly and recover in minutes beats one that hardly ever fails but fails catastrophically or takes days to untangle. Fast, predictable recovery matters most. - Amy Gu, Dynamsoft
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Many organizations optimize for efficiency through consolidation, which introduces complexity and increases systemic risk during outages. Teams should consciously trade some efficiency for resilience by adopting hybrid, redundant architectures that distribute workloads across on-premises edge and cloud environments, which reduces single points of failure and enables continuous operations. - Luiz Domingos, Mitel
Prioritize control, visibility and redundancy over pure efficiency. Traditional network architectures often rely on fewer opaque network paths, which limits insight and flexibility during outages. Investing in diverse connectivity and direct interconnection gives teams greater control over how data moves, enabling faster rerouting and more predictable performance when disruption occurs. - Ivo Ivanov, DE-CIX AG
You need circuit breakers. The financial markets learned this the hard way: Efficient systems optimized for speed will outrun their own safeguards. The same is true in software. The conscious tradeoff is engineering deliberate halts that feel like waste until they’re the only thing standing between a hiccup and a catastrophe. - Jeromee Johnson, Tellus
Trade speed for recovery. Systems optimized for maximum throughput often fail hardest because they lack recovery paths. Building in slower release gates, rollback mechanisms and controlled degradation may reduce peak performance, but it ensures systems can fail safely and recover quickly. This turns outages into manageable events instead of business disruptions. - Nirab Kumar, Odyssey Logistics
Every system optimized to the edge has no room to fail gracefully. Leaders must consciously over-provision through backup systems, failover paths and redundant infrastructure that sits idle until it isn’t. Efficiency asks, “What can we cut?” Resilience asks, “What can we not afford to lose?” Build the backup before you need it. The outage will come. The only question is whether you’re ready. - Aruna Veerappan, Upwork
Trade seamless edge automation for deliberate friction. Hyper-optimized AI cascades failures instantly by stripping away operational speed bumps. Establish a “human red-line contract”—hardcoded thresholds where edge nodes pause, hold state and degrade to a human in the loop before executing drastic shifts. A slower, deliberate failure is always a recoverable one. - Deep Narayan Mishra, Walmart Inc.
Trade centralized efficiency for blast radius control. Some systems consolidate dependencies, allowing failures to cascade quickly. Introduce segmentation, isolation and independent failure domains, even with added cost and complexity. Resilience is not preventing failure; it is containing it so outages stay isolated instead of becoming enterprisewide disruptions. - Lane Sullivan, Concentric AI
Trade tight coupling for modular isolation. Highly optimized systems share dependencies to reduce latency and cost, but that creates cascade failures. Designing for isolation with clear boundaries may add overhead, but it limits blast radius and enables faster recovery when failures occur. - Pawan Anand, Persistent Systems
The convenience of having a single provider with highly integrated systems to handle multiple operations should be sacrificed in favor of resilience. Even the biggest tech giants have had outages in the past that affect every service they provide. It is less convenient to use multiple providers, but when one of them has an issue, you get to keep part of your operations running. - Julius Černiauskas, Oxylabs
Using parsed or sharded distributed storage will preserve progress across departments even during an outage, as these systems are fault-tolerant. You are not choosing resilience over efficiency; you are choosing to decentralize to make your efficiency more resilient. - Daniel Keller, InFlux Technologies Limited (FLUX)
In process industries, meeting safety requirements often means accepting higher CapEx and OpEx by adding process analyzers and measurement redundancy instead of focusing only on efficiency. Hydrogen is a good example. Reliable, real-time data is needed, even at higher cost and complexity, to keep operations safe during failures and allow scale-up. - Gregory Shahnovsky, Modcon Systems Ltd.
Trade peak efficiency for optionality. Standardizing on a single optimized system lowers costs but increases fragility. Introducing redundancy, vendor diversity or failover paths may seem inefficient, but it preserves continuity. Resilience isn’t free—it’s a conscious decision to pay for flexibility before you’re forced to. - Prajkta Waditwar, Box Inc.
Diversify your bottom-line systems. While business continuity/disaster recovery plans are important, when an outage hits, you can be more resilient if your technology and business plans are diversified. Say your business drops to 80% capacity for 24 hours while BC/DR plans are carried out. This partial capacity allows you to continue operating and generating revenue while you work to restore 100% capacity. - WaiJe Coler, InfoTracer
The real tradeoff isn’t efficiency versus resilience; it’s reversible versus irreversible failure. A payments app can safely halt for hours: transactions queue, and users come back. An autonomous vehicle, however, cannot safely halt. Map every system according to whether downtime causes recoverable inconvenience or permanent harm, then spend your resilience budget accordingly. - Lihong Wang, Freeport Markets
Tech teams should build human bench depth, not just system redundancy. The most efficient org chart is one expert per system, with no overlap—and it’s also the most fragile. When the engineer who built the legacy auth service quits, your multiregion failover doesn’t help. The military cross-trains for a reason. Why don’t other orgs do the same? Cut the deputy, and you’ll meet them at 3 a.m. - Dan Sorensen, Nexus Security Advisors
The tradeoff most teams refuse to make is coverage for speed. They cut test coverage to hit deployment targets. That is not optimization; it is deferral. Resilience means investing in verification before an incident, not after. When outages hit, the majority of organizations blame the engineer closest to the code. The shortcut made that inevitable. It always does. - Shubha Govil, Sauce Labs
Ignoring cost resilience trade-offs wastes money without improving durability. Move from expensive, traditional replication to ultra-resilient architecture (active-active single master, multimaster, multiregion and multicloud) by first rationalizing cost and resilience SLAs. Regulatory needs, like country-specific multicloud critical service deployments, also shape design, not arbitrary choices. - Karthik Ranganathan, Yugabyte
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。