Principal Engineer, Apple Cloud Object Storage (ACOS), Storage Infrastructure

AppleSeattle

Posted 2 months agoStill open, checked 2h ago

About the job

ACOS is Apple's geo-distributed, exabyte-scale object storage service. It serves billions of requests per day for iCloud, Apple Music, Apple TV, Apple Maps, and Apple's internal data and analytics platforms. It provides 11-nines of durability and 4-nines of availability.

The ACOS Storage Infrastructure team owns how data is encoded, placed, verified, repaired, and reclaimed inside a storage cluster. This includes erasure and replicated encoding schemes, data placement and rebalancing optimization, compaction, scrubbing and repair, durability modeling, performance tiering, and the supply chain that keeps available capacity in front of customer demand. In practice, this team determines what storage costs, how it survives in the face of software and hardware failures, and how fast it is served. This is a role for an engineer who has designed, shipped, and operated the data path of a storage system at exabyte scale.

The next big challenge here is to support the industry's transition to much denser drives. This means capacity per drive is growing far faster than IOPS per drive, and maintenance operations — sealing, compaction, scrubbing, repair, rebalancing, cross-region replication — consumes that increasingly scarce IOPS budget. This is made harder by the growing demand for high-performance storage tiers by Data and AI workloads

You will own the technical agenda for solving that, across four areas:

* Storage Efficiency. Drive cost per usable byte down further: reduce replication factor without conceding the durability bar, reclaim storage overheads for compaction and other processes, and make placement maximize both space and IOPS utilization. Land support for denser and heterogeneous drive generations (CMR, HAMR, SMR) without letting IOPS-per-TB become the binding constraint on the fleet.
* Storage Integrity. Guarantee 11-nines of durability in production in the face of software and hardware failures. Own detection and repair end to end: balance detection latency against the IOPS cost of scrubbers and ensure correctness of the encode/decode/repair/rebalance paths.
* Storage Performance. Deliver differentiated performance tiers. That means improving tail latency and IOPS for the standard performance tier, while also introducing high-performance SSD-backed tiers for data and AI workloads.
* Capacity Planning and Management. Make supply closely track customer demand. Own demand forecasting, drive-generation transitions, capacity on-boarding and decommissioning, quota and rate-limit commitments, and the release value when the system is oversubscribed.

Your influence will extend well beyond ACOS. You will be the person Apple's storage organization, its hardware engineering partners, and its executives rely on for judgment on encoding, durability, and fleet efficiency.