Main menu


Azure Cloud spends two more years on servers

Azure Cloud spends two more years on servers... Enterprises are also trending to extend the server lifecycle

Azure Cloud spends two more years on servers... Enterprises are also trending to extend the server lifecycle

For some time, cloud service providers have tended to overspend on hardware costs. This was because they had to deploy thousands or tens of thousands of servers in a data center the size of a football field as quickly as possible. But it seems that this hardware feast is over. At a recent performance meeting with analysts, Microsoft announced plans to extend the lifecycle of its cloud servers from four to six years.

Microsoft CFO Amy Hood explained, "Thanks to technological advances and increased operational efficiency of server and network equipment, the lifespan has been extended beyond the useful life." Cost savings aren't too bad either. Hood said that longer server purchase cycles could save $3.7 billion in 2023 alone.

Microsoft is not alone in making this move. Amazon Web Services and Google Cloud announced at a performance meeting in February that they had extended the hardware lifecycle by one more year.

In fact, many companies already have longer server refresh cycles than cloud service providers. According to IDC's 2021 survey, 10% of respondents have been using servers for critical workloads for 5 years or more, and 19% have been using servers for non-critical workloads for 5 years or more.

Although there is a big difference between a cloud service provider's data center and an enterprise data center, Microsoft's decision is worth consulting with enterprise data center experts as well. 

General enterprise hardware has warranties, but cloud service providers don't, according to Asishi Nadkarni, vice president of global infrastructure practices at IDC. Cloud service providers use a large amount of general-purpose hardware, and if there is a breakdown, to some extent, internal personnel repair or replace it. Companies usually sign technical support contracts with equipment vendors. Hyperscale data centers also use a redundancy model, but budget-constrained enterprises are much less resilient using redundancy.

In addition, changes in the way equipment is purchased and advanced technologies are also helping to extend the lifespan of a company's servers. Many enterprise data centers are changing the purchasing method from the CAPEX model in which servers are purchased directly to the OPEX model in which they are leased. Programs like HPE Greenlake and Dell Apex support this change.

In fact, today's servers are much more durable than they were 10 years ago. “I think the MTBF is very high,” Nadkhani said. This is particularly the case with component failures that make server operation impossible, such as hard disk failure or memory failure. As a result, the overall lifecycle of the server and overall uptime have increased significantly.” The replacement of hard disks with mechanical drives with SSDs also had a decisive effect on this. This is because the hard disk was a part with a higher probability of failure or failure than other parts.

Thermal management efficiency is also improved. “The fan design has improved and the system design has improved,” Nadkarni said. In particular, various cooling technologies such as water cooling are being used. So the chip doesn't fail that fast.”