DGraph as a Service (Managed DB)


#1

Hi all,

TLDR: Would you be interested in a “DGraph as a Service”? https://dgraphcloud.io :zap:

We’ve struggled a lot back then for installing DGraph, until we finally got it and use it in production.

We wished there was a Managed DGraph service that we could have used right away.

Since there is a LOT of issues about installing/running DGraph, we have decided to use all that we’ve learned bringing DGraph to production and create such a service.

This post is to see how interested users would be in such a service.

Here is the website: https://dgraphcloud.io

We would appreciate any comment/critique/insult/feature request :slight_smile:

Cheers :clinking_glasses:


(Rus) #2

YES , DGraph aaS sounds fine, question being to what client capacity ,

meaning more than http// connections.

My interest is to run StackOverflow Clone for hundreds of separate client bases.

without the content. Blank canvas, with an App interface (using Google certificates)

Each with a specific provisioning service , where questions / answers are about that service.

When can u show an instance, with pricing structure for Kubernetes HUB style POD per client.

Naturally takes time to grow … we presume < 1GB traffic / day to start for months, expect TB

Thanks for the reach out …

Rus Talisin


#3

Thanks for your interest @talis!

We do not have pricing plans to present yet. There will most probably be different plans, ranging from a free “sandbox” one with very limited resources, all the way to High Throughput/High Resources plans with SLA.

For starting we will be focusing on those low resources plans not meant for production load, so we can iterate, stabilize, and act on feedback.


(Rus) #4

Naturally, low resource plans are fine … DGraph s best pal is Badger,

All keen to see how an all Go merge pans out without recourse to C++ libs.

Quite a story on lack of profiling / general coding darkness , when mixed libraries were used

Congratulations on the success of Badger.

As a hardware eng, back in this field from protracted Med studies, only programming i did was ASM

I’m talking AD2100 / 21060 DSP and Texas 9900 way back…

Unfort never liked C or C++ (fake reusable promise) or abstraction, it was and still is SLOW

yet i retain skill , that may transpose across abstraction, allowing both sides to work…

I have 2 IBM P8 servers over in the US, transactional memory AND isolated PCIe x16 controllers

after spending loads on E5 2690 Servers, i get 2 P8 's in hope to arbitrate memory in ASM on an

Atomic time slice basis, to carry out burst weighted table correlations.

Something not possible with Intel , not sure if the virtual mem management can be altered on P8,

at least 2 isolated PCIe on chip controllers ( separate to the Bridge) provide arbitration control.

This success is highly geared.

If u like to compare results running Dgraph on transactional memory (no pinning) i can ship

the 2 servers to your location, or have same installed upper NY.

Docker seems to have shed some weight into a runtime image , without VM baggage.

I want to flex a concept that may deploy thru https://github.com/galileo/docker-binaries

for the client side … plan for 10000 client / month takeup. (likely 12months gestation)

with supervisory Server client accounting.

Happy for a go-slo , as I’m abstraction deprived… not completely starved, tho getting close

and trump has not assisted … many items sit in NY, by grace of datacentre owner.

Any road ahead for an Overflow clone to play with?

Regards , with Anticipation

Rus Talisin