Data Storage


MIGRATING WORKLOADS AND PERFORMANCE ISSUES IN THE PUBLIC CLOUD


When on-premises capacity runs short, public cloud tends to be the first option


infrastructure teams reach for. It is quick to provision, removes the hardware procurement problem, and sidesteps the question of what to do with an ageing estate. What remains unsettled is whether migrated workloads will perform as the business requires once they are live in production, or whether the recovery design has kept pace with where services now sit. Here, Steve Spittal, Technology Director at Pulsant, offers his insights.


M


emory costs have risen sharply, complicating hardware refresh for organisations still running significant


on-premises infrastructure. Businesses that moved workloads to public cloud are increasingly finding that production performance does not match what testing suggested. For infrastructure teams trying to add


headroom to existing environments, the memory market has created a genuine constraint. Gartner forecasts that DRAM prices will rise by 125% in 2026, with no meaningful correction expected before late 2027. Te Register reported in January that Samsung has already raised server memory prices by up to 60% and that, combined with 2025 increases, costs could nearly double by mid-year. AI infrastructure is consuming an outsized share of available memory supply, leaving enterprise hardware refresh competing for components at elevated prices with extended lead times. Running out of on-premises headroom is a


legitimate operational problem. Migrating to the cloud to escape it is a reasonable short-term response. Te difficulty is that urgency tends to compress the evaluation of whether public cloud


24 | September/October 2026


is actually the right long-term home for each workload being moved.


Production reveals what testing does not Performance problems aſter migration typically surface under real conditions rather than in controlled testing. Production volumes, live data, and the full web of service dependencies behave differently at scale. For latency-sensitive workloads in particular, the distance between where a service runs and where its users are located directly affects response times, a factor that pre-migration benchmarks rarely capture. Training workloads can stay in large, centralised


environments, but inference demands proximity to users and data. IDC research vice president Dave McCarthy observed in December that edge computing will be required to address latency and privacy as AI shiſts from training to inference. Deloitte’s 2026 technology outlook estimates inference will account for roughly two-thirds of all AI compute by year-end, up from a third in 2023. A workload can be well-provisioned and fully


managed in a hyperscale environment and still sit too far from the users and data it serves.


www.pcr-online.biz


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52