Glossary comparison
Scale-out vs Scale-up: what is the difference?
The difference
Scale-up joins a limited number of accelerators with very high-bandwidth, low-latency links, often inside a rack, so they behave almost like one large device. Scale-out links many such domains through a slower but much larger switched network.
Scale-out connects many separate nodes or domains over a data centre network, trading bandwidth and latency for size. Scale-up enlarges one domain with much faster, shorter links.
The two terms
Scale-out
Adding capacity by connecting many independent nodes over a switched network such as InfiniBand or Ethernet. It scales to very large sizes but imposes higher latency and lower bandwidth per accelerator than scale-up.
Scale-up
Enlarging a single coherent compute domain by connecting more accelerators over a short, very high bandwidth link so they behave as one large device. It is bounded by the reach of that link technology.
Side by side
| Field | Scale-out | Scale-up |
|---|---|---|
| Stands for | Not an abbreviation | Not an abbreviation |
| Type | Term | Term |
| Domain | Compute & AI Hardware | Compute & AI Hardware |
| Group | Interconnect and fabric | Interconnect and fabric |