Reason 2: Marketing metrics are not your metrics.
F or years, the main performance indicators for purpose-built backup appliances have been the “ingest performance.” The maximum performance published on most vendors’ datasheets are the highest backup rates that the appliance can achieve in a very controlled environment. Usually, these rates are achieved starting with no data on the appliance, so that there is no deduplication occurring. Here again, there are a number of factors that influence the maximum ingest performance. These include how much redundancy is in the data being backed up, the number of backup servers and streams, and the age of the system.
As data redundancy increases, the ingest performance will decline. Generally, data redundancy levels reach a constant state if there are new data sources, that is, new types of data being backed up. This generally occurs during the first month of use, but it will change if new data sources are added to the backup.
The number of backup servers and streams is generally constant (remember the example in reason 1). If the customer adds more servers and streams, this can increase the ingest performance. Removing backup servers and streams decreases performance.
The age of the system generally decreases performance until a steady state is reached. Steady state is achieved when the system is backing up and deleting the same amount of data each day or week. There are a number of reasons for this, but the primary ones are housekeeping processes, such as space reclamation and the spread of the data over the disks in the appliance.
Other factors that affect ingest performance include replication, encryption, Path-to-Tape (PTT) (a.k.a tape out), and restores. At steady state, customers normally see 20% – 30% lower ingest performance than they experience on the first day of use. This is a similar issue as described earlier for the deduplication ratio. Ingest performance is not real life. A more interesting comparison would be to look at daily capacity.
Daily capacity is a measure of sustained performance — in other words, how much data the appliance can backup and delete in a day, including housekeeping tasks. This is what it must perform at steady state. Daily rates are not published, as they also vary on a case-by-case basis and are difficult to standardize. But this is an important figure to understand during a proof-of-concept phase. Daily capacity is much closer to what a customer will need, compared to needing maximum ingest performance. Daily capacity measurement should include the load of replication, tape out, space reclamation, and some restores to measure the “real life” appliance performance (in TB/day).
To learn more about our industry-leading deduplication systems, visit us at www.quantum.com/dxi.
Stay tuned for Reason #3 this week!