PostMark is a benchmark that demonstrates system performance for short-lived small files seen typically in Internet applications such as electronic mail, netnews and web-based commerce. The benchmark creates an initial pool of random text files of varying size. Once the pool has been created, the benchmark performs two types of transactions on the pool: (i) create or delete a file; (ii) read from or append to a file. The incidence of each transaction and its subtype are chosen randomly to eliminate the effect of caching and read-ahead.
Our experiments use a equal predisposition to each type of transaction as well as each subtype within a transaction. We performed 100,000 transactions on a pool of files whose size was varied from 1,000 to 25,000 in multiples of 5.
Table 5 depicts our results. As shown in the table, iSCSI generally outperforms NFS v3 due to the meta-data intensive nature of this benchmark. An analysis of the NFS v3 protocol messages exchanged between the server and the client shows that 65% of the messages are meta-data related. Meta-data update aggregation as well as aggressive meta-data caching in iSCSI enables it to have a significantly lower message count than NFS.
As the pool of files is increased, we noted that the benefits of meta-data caching and meta-data update aggregation starts to diminish due to the random nature of the transaction selection. As can be seen in Table 5, the number of messages relative to the file pool size increases faster in iSCSI than that in NFS v3. Consequently, the performance difference between the two decreases. However, as a side effect, the benchmark also reduces the effectiveness of meta-data caching on the NFS server, leading to higher server CPU utilization (see Section 5.4).