Latency Numbers Every Programmer Should Know
Jonas Bonér (based on work by Peter Norvig and Jeff Dean from 2012):
L1 cache reference 0.5 ns Branch mispredict 5 ns L2 cache reference 7 ns 14x L1 cache Mutex lock/unlock 25 ns Main memory reference 100 ns 20x L2 cache, 200x L1 cache Compress 1K bytes with Zippy 3,000 ns 3 us Send 1K bytes over 1 Gbps network 10,000 ns 10 us Read 4K randomly from SSD* 150,000 ns 150 us ~1GB/sec SSD Read 1 MB sequentially from memory 250,000 ns 250 us Round trip within same datacenter 500,000 ns 500 us Read 1 MB sequentially from SSD* 1,000,000 ns 1,000 us 1 ms ~1GB/sec SSD, 4X memory Disk seek 10,000,000 ns 10,000 us 10 ms 20x datacenter roundtrip Read 1 MB sequentially from disk 20,000,000 ns 20,000 us 20 ms 80x memory, 20X SSD Send packet CA->Netherlands->CA 150,000,000 ns 150,000 us 150 ms
Colin Scott has a page that helps visualize how these types of numbers have changed over time (Hacker News).
Mohammad Zeya Ahmad has an informative post [archive] that answers that question. He has a list of how much time various common operations take. That’s interesting but what make his list stand out is that he draws conclusions from his results.
For example, SSDs are about 30 times faster than HDDs so if you have a high performance disk-based task, it makes sense to use SSDs. Of course, there are reasons to prefer HDDs but if performance is your controlling metric, SSDs are probably your best choice.
For each group of comparable metrics, Ahmad offers an actionable suggestion. Those groups range from CPU versus Cache and Memory speeds to network transfer times.
Previously:
- Python Numbers Every Programmer Should Know
- C Is Not a Low-level Language
- Squeezing the Most Out of Bluetooth
- Performance Comparisons of Common Operations, 2016 Edition
- Transatlantic Ping Faster Than Sending a Pixel to the Screen
