Storage for DBAs: Take a look in your data centre at all those humming boxes and flashing lights. Ignore the storage and networking gear for now and just concentrate on the servers. You probably have many different models, with different types and numbers of CPUs and DRAM inside. My question is, which CPUs are the most … Continue reading The Most Expensive CPUs You Own
Oracle RAC is sold as a high-availability solution, but its real cost – in licenses, complexity and the hidden assumption that losing a node doesn’t count as an outage – is rarely made explicit.
Oracle database licensing is eye-wateringly expensive per CPU core – and storage costs, often seen as prohibitive, represent a surprisingly small fraction of the total three-year spend.
Putting Oracle redo logs on SSD is a common reflex response to slow databases – but it usually misdiagnoses the problem, and SSDs handle write-heavy sequential workloads worse than you might expect.
IOPS figures dominate flash storage marketing, but for databases latency is the metric that matters – high IOPS at unpredictable latency deliver no real-world value.
NAND flash development is driven by the consumer market – not the enterprise. The question to ask any flash vendor isn't "how fast?" but "where's your innovation?"
Not every database benefits from flash storage – knowing when it matters requires understanding how much I/O your workload generates, how random it is and how much latency is already costing you.
Disk I/O forces a choice between random and sequential access – and that choice defines whether latency compounds or disappears. Flash makes the distinction irrelevant.
Latency, IOPS and bandwidth are the three properties that define any storage system – understanding how they relate to each other is the first step to knowing what your database actually needs.
Database performance problems are rarely solved by faster CPUs alone – the real issue is imbalance between resources, and disk latency is the silent bottleneck that flash storage changes.
flashdba 