The Continuing Evolution of the Server Market

The primary motivator for Adaptec to re-enter the HBA market is due to the expanding use of HBAs in both high-density and custom-built server solutions.

The old model of server deployments is changing rapidly. The approach of using standard motherboard form factors has begun to give way to quad-node solutions. These types of servers, used in compute clusters and other deployments, provide more density and higher performance in a smaller space. There are many advantages to this approach, significantly reducing TCO and OpEX expenses.

These types of servers are becoming available from typical OEM suppliers, but require the slimmest applications for storage solutions. Quadrupling the amount of native ports available on their HBA's provides customers with significantly more ports without the need for additional expanders.

Expanders themselves can bring about a loss of performance for the overall solution. Expanders always incur a latency penalty and can adversely affect sequential throughput. Factoring the cost for procurement of expanders is just the beginning of an ongoing OpEX expenditure. The space required within the server can limit the storage density, and then the associated costs of powering and cooling the extra hardware add up quickly over the life of the device.

RAID controllers are losing favor in many deployments due to the tremendous capacity of HDDs. The available capacity of near-line and enterprise-class HDDs has expanded rapidly with technology advancements, bringing new problems into the datacenter. Failed RAID arrays can require an excessive amount of time to rebuild due to the larger size and slow speed of HDDs. In many cases, the length of a rebuild on a large RAID array can require so much time that the odds of another drive failing during this process is mathematically assured.

This guarantees a loss of data in RAID 5 deployments, and in some cases will require an additional layer of parity to negate the loss of data during the inevitable rebuild. RAID 6 lowers the available capacity of the solution and hampers performance. This also does little to address the long rebuild times or the compromised performance during these rebuild periods.

Another solution for this problem lies in simply discarding the array and rebuilding it from a backup or replication scheme. This provides administrators with a faster means of bringing the array back on line, and in many cases, this will negate the need for a RAID controller.

As HBA's begin to make more sense in many applications typically served by RAID controllers, there will be an accompanying expansion in the HBA market.