All of the components of Deadline are module; a few of them being optional (Pulse, Balancer). So, DB, Repo, Pulse, License Server could all be installed on one server, but the main reasons you might NOT want to do this are:
- RISK (Is your company management team happy with this level? Can client deadlines be missed slightly?)
- Size of your farm. (Fit for purpose).
So, to expand on those points. If this single server goes down, essentially your farm is dead. Is this an acceptable risk in your studio? For me, as a developer, running all these things on a single laptop with multiple testing VM’s, that is fine. With regard size; this is the major influencer on how you might go about setting up your farm to an ideal configuration given the size of your farm now and the near future?
All of the Deadline components can have redundancy, so you can install the mongoDB across multiple servers (see our docs on Replica Sets), when your farm is 2000+, you will want to consider mongoDB sharding as well (see our docs again here). Pulse and Balancer have a HA heartbeat failover (primary-secondaries) system.
Multiple License Servers can be built, but I personally prefer to take advantage of modern VM failover technology here and run a single license server, but have the VM infrastructure kit in place, to allow auto failover to another hypervisor host machine in case of failure. The “repository” is essentially a file server. You should install the “repo” on the fastest file server you have. This might well not be a lower spec machine (physical or virtual) that you have decided on installing the MongoDB on. If you are submitting your scene files to the Deadline Repository, then most definitely, you want a separate, fast file server, that is fit for purpose here.
How fast is fast enough? Well, I would answer with a question, what do you currently use as your main studio network file server to store your production files on? Is that fast enough given the number of machines/users you have now and potentially in the near future? Is this machine good enough to also provide the location to install the Repository as well? Of course, you can side-step this issue, by NOT submitting your scene (data) files with your Deadline job, but rather just reference them from their original location, which so many studios do take this approach. This other approach, lowers greatly the spec requirements for a repository, particularly on total disk capacity required (as 3D scene files are huge these days!). However, it will always be good to place your “repository” part of your Deadline setup on some really fast disks (RAID/SAS, SSD or better!). SSD prices have fallen greatly.
Don’t forget the usual backup. All the usual techniques will work here.
The official min spec requirements are here for DB:
docs.thinkboxsoftware.com/produc … l#database
and here for Repo:
docs.thinkboxsoftware.com/produc … repository
and this page has saved many a person, considering a farm spec overview:
docs.thinkboxsoftware.com/produc … iderations
Let us know if you have any further questions and I’d try my best to answer them. What size farm are you putting together?
