Measured on 12 October 2026About BsonJet
MongoDB GUI benchmark BsonJet vs. Compass vs. Studio 3T
"Fast" is easy to write. So we measured it: three MongoDB tools and mongodump on the same clean cloud machines, against the same servers, once inside a datacenter and once over an ordinary office connection (50 Mbit/s down, 10 Mbit/s up, 20 ms).
- Compass2.2 s
- Studio 3T19.6 s
- Compass not available
- Studio 3T129 s
- mongodump155 s
- Compass964 MB
- Studio 3T1,436 MB
A database with 50,000 collections
One database, 10,000 tenants with 5 collections each. Every tool starts closed: start it, connect, expand the database, type 09999 into the collection filter.
1 BsonJet waits 200 ms after the last key before it filters, so it does not filter on every letter. The filtering itself takes about 60 ms. Compass filters the list it already has as you type.
Copying a database to another server
43 collections, 7.34 million documents with indexes, from one server to another. The data goes through the client, as it does when you copy between two servers from your desk. Every copy was checked afterwards: document counts, indexes and dbHash of every collection identical to the source.
2 The same 43 collections with fewer documents, so that one run of the slowest tool stays under 7 minutes over a 10 Mbit/s upload.
BsonJet turns on network compression (zstd) by itself. We used the same connection string for every tool, the way you paste it, without compressors=zstd.
How we measured
- Machines
- Three fresh Vultr virtual machines in Frankfurt, each with 4 dedicated vCPUs, 16 GB RAM and NVMe: a Windows Server 2025 client with nothing but the tools, and two Ubuntu 24.04 servers.
- Servers
- MongoDB 9.0.2, a single-member replica set with authentication, installed with our install script.
- Tools
- BsonJet 1.0.6, MongoDB Compass 1.52.0, Studio 3T 2026.14.0 (trial), MongoDB Database Tools 100.19.1. Default settings, the newest versions on the day.
- Networks
- Datacenter: the private network between the machines, no limit. Office: the same network limited on the servers to 50 Mbit/s download, 10 Mbit/s upload and 20 ms round trip.
- Timing
- From the click to the result on the screen, frame by frame at 60 fps. Start: from the click on the taskbar icon to a window you can work in. Expand: from the click on the database to the list of its collections. Copy: the time each tool reports when it finishes, mongodump with mongorestore from the start of the command to its end.
- Memory
- The working set of all processes of the tool after expanding the database and filtering.
- Runs
- At least 3 runs of everything, one tool at a time, the others closed. The tables show the median.
Repeat it yourself
- gen-bench.js creates the same data every time.
WHAT=manymakes the 50,000 collections,WHAT=copythe database for the copy (SCALE=0.15for the smaller one). Run it withmongosh, the comments say how. - results.csv has every single run behind the tables.
- The office network on the servers:
tc qdisc replace dev <nic> root netem delay 20ms rate 50mbiton the source, the 10 Mbit/s upload limit on the incoming side of the target.
A word from the author
I make BsonJet, so read this with that in mind. I tried to give every tool a fair chance: clean machines, default settings, the same data and the same clicks. If you see a mistake, or a setting that makes another tool faster, write to [email protected] and I will measure it again and update this page.
MongoDB and MongoDB Compass are trademarks of MongoDB, Inc. Studio 3T is a trademark of its owner. BsonJet is not affiliated with either.