We're announcing our Dedicated plans' beta today. Previously we've been offering these Dedicated virtual machines to a handful of production customers. We're opening the offering as a public beta in Amazon's EC2 East and as a limited alpha in Rackspace's ORD datacenter. You can find more about these plans including pricing at: http://mongolab.com/about/products
Compared to our Shared plans, these new Dedicated plans support much larger datasets and are designed for high availability and predictable latency. Shared plans are perfect for smaller datasets up to 20GB and lightweight production / pre-production uses. But when your service needs to scale to the next level, Dedicated plans are the way to go.
Dedicated plans, aptly named, have dedicated resources, RAM and VMs, that make contention less of an issue. Being a Cloud service, these Dedicated plans are built on modern virtual machines that offer excellent isolation between VM guests.
Several operational factors are improved in Dedicated. Highlights include:
- High-availability with automatic failover
- Improved read throughput and scalable read throughput by the creation of a Replica Set cluster of more than one member
- More administrative control like MMS are available (some MMS details here)
- Much larger maximum size of datasets
- Details are available at: https://mongolab.com/about/products/#/comparison
Prices start at $320/month for an X-Small (3.5GB RAM, 20GB storage) 1-node plan.
Pricing for dedicated is based on three main parameters.
- Amount of RAM.
- Number of nodes in the cluster.
- Pre-allocated storage.
Each of the plans include some included storage, typically over 2x the amount of RAM. Additional storage is available as an option.
We offer a 1-node Dedicated plan. Curious you might say? How do you handle automatic failover with just one node? Great question. For the 1-node plans, we have a hybrid model where the failover slave for the 1-node plan is a Shared instance. This is a fairly popular plan among our early customers as it provides a starting point for the benefits of the Dedicated plans.
Pricing details are here.
There are a few rules of thumb for choosing your plan:
- Fit the working set of data within RAM. The working set is approximately the data and indexes that would satisfy 80-90% of all queries. If a query or write cannot be satisfied in RAM, the latency for completion shoots up several orders of magnitude.
- For storage, it's important to note that indexes also consume storage (and working RAM). Additional storage will be necessary if you have data and indexes beyond the initial allocation. A certain amount of headroom is also recommended depending on your data growth rate. A 3-6 month bit of headroom is recommended. We'll notify you if we see that storage is at risk of being exhausted. Adding storage is easy, but does require some downtime.
- For cluster size, larger clusters improve read throughput. (Writes still must be processed by a single node). Modeling of your traffic will help with this sizing.
- Of course, you need to determine where your application server will be hosted at either EC2-east or Rackspace ORD.
Reminder: we're in beta. The service is sturdy enough for production on Amazon. We're improving the tooling around it to enter GA. Rackspace is not as mature, so we're designating it an alpha. So some parameters are subject to change on Rackspace. We're working on other providers as well.
How to order
- Sign up if you haven't already here:
- Input a credit card
- Send email to email@example.com with the desired configuration. (We're working on automating these steps).
We need to know:
- Which cloud
- Number of nodes
- Amount of RAM
- Amount of additional storage.
We'll get back to you shortly when the instance can be scheduled for provisioning, typically a business day.
We hope you'll enjoy the service.
(update: corrected some links 2012-02-01 11:05a)
I’m MongoLab’s VP of Sales and Marketing. And I’m here to serve our customers’ needs for MongoDB hosting in the cloud.