Parity-encoded redundant disk arrays provide highly reliable, cost-eff
ective secondary storage with high performance for reads and large wri
tes. Their performance on small writes, however, is much worse than mi
rrored disks-the traditional, highly reliable, but expensive organizat
ion for secondary storage. Unfortunately, small writes are a substanti
al portion of the I/O workload of many important, demanding applicatio
ns such as on-line transaction processing. This paper presents parity
logging, a novel solution to the small-write problem for redundant dis
k arrays. Parity logging applies journalling techniques to substantial
ly reduce the cost of small writes. We provide detailed models of pari
ty logging and competing schemes-mirroring, floating storage, and RAID
level 5-and verify these models by simulation. Parity logging provide
s performance competitive with mirroring, but with capacity overhead c
lose to the minimum offered by RAID level 5. Finally, parity logging c
an exploit data caching more effectively than all three alternative ap
proaches.