4 ms·

how big is the market of people who have money to burn? or is it that some people mistakenly believe that bigger devices are more cost-effective? this part of

how big is the market of people who have money to burn? or is it that some people mistakenly believe that bigger devices are more cost-effective? this part of the release is interesting: https://documents.westerndigital.com/content/dam/doc-library/en_us/assets/public/western-digital/product/platforms/rapidflex/product-brief-rapidflex-c2110-interposer.pdf https://documents.westerndigital.com/content/dam/doc-library...

choilive 2y ago
This is targeted for enterprise/cloud/hyperscale customers and density is often a big deal in those applications. I know that 61.44TB SSD have been selling very well exactly for this reason.
wmf 2y ago
Drive bays cost money so in some cases it could be cheaper to use a smaller number of more expensive drives.
jandrewrogers 2y ago
There are applications and workloads where storage density is a key consideration, and the set of data models isn't that small as a subset of the market. Networked storage is not a substitute for NVMe when working with very large storage volumes due to the poor storage bandwidth. For some workloads you also cannot substitute a single high-density server with multiple low-density servers for effectively the same reason (low bandwidth to remote storage). For these workloads, ultra-high density storage can definitely be cost effective within reason. It bears keeping in mind that most popular open source software is not designed to be effective for storage densities anywhere near this high. If you are working with data at this density then you are likely using one of the closed source storage engines designed to work at this scale running straight off of the raw block device.
supersour 2y ago
This is also useful for power-constrained systems - 128T x 35W is less power than 2 x 64T x 35W which gives a larger power budget for whatever h100-type device you have connected to these drives
estebank 2y ago
Picture a company with its own on-prem bare metal servers. Over the years they have had space to expand but projections show that soon either a new data center will be needed, either in addition or to move to. Doing so would require a huge capex to either migrate the existing hardware to the bigger space or to reachitect their topology to work on a multi-DC setting. Or you can look at the oldest servers in the fleet, that are scheduled to be decommissioned in the next N years, that have HDDs that are a small fraction of the capacity of the new ones. If you crunch the numbers, it can start to make sense to just say "we just pay the premium for physical density and increase the time before we're in the same position by M years". If M is large enough, people will jump on that option as fast as you can draft the Purchase Order.

choilive 2y ago

This is targeted for enterprise/cloud/hyperscale customers and density is often a big deal in those applications. I know that 61.44TB SSD have been selling very well exactly for this reason.

wmf 2y ago

Drive bays cost money so in some cases it could be cheaper to use a smaller number of more expensive drives.

jandrewrogers 2y ago

There are applications and workloads where storage density is a key consideration, and the set of data models isn't that small as a subset of the market. Networked storage is not a substitute for NVMe when working with very large storage volumes due to the poor storage bandwidth. For some workloads you also cannot substitute a single high-density server with multiple low-density servers for effectively the same reason (low bandwidth to remote storage). For these workloads, ultra-high density storage can definitely be cost effective within reason. It bears keeping in mind that most popular open source software is not designed to be effective for storage densities anywhere near this high. If you are working with data at this density then you are likely using one of the closed source storage engines designed to work at this scale running straight off of the raw block device.

supersour 2y ago

This is also useful for power-constrained systems - 128T x 35W is less power than 2 x 64T x 35W which gives a larger power budget for whatever h100-type device you have connected to these drives

estebank 2y ago

Picture a company with its own on-prem bare metal servers. Over the years they have had space to expand but projections show that soon either a new data center will be needed, either in addition or to move to. Doing so would require a huge capex to either migrate the existing hardware to the bigger space or to reachitect their topology to work on a multi-DC setting. Or you can look at the oldest servers in the fleet, that are scheduled to be decommissioned in the next N years, that have HDDs that are a small fraction of the capacity of the new ones. If you crunch the numbers, it can start to make sense to just say "we just pay the premium for physical density and increase the time before we're in the same position by M years". If M is large enough, people will jump on that option as fast as you can draft the Purchase Order.