#community-help

Discussing RAM, CPU, and Hosting Options for Project

TLDR Bill asked about plans, domains and performance disparity with Typesense versus Digital Ocean. Kishore Nallan and Jason provided insights about network and disk performances, comparisons and prices.

Powered by Struct AI

1

Jan 05, 2023 (11 months ago)
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
01:57 PM
Hello, is the 2GB Ram - 2vcpu - 4hr burst with CDN - 3 nodes plan, suitable for production (at the start of a project)? In addition, can we use custom domain with CNAME?
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
02:57 PM
Depends on your traffic profile. Many people don't have enough search traffic to really need dedicated cpu. In such cases burst enough to handle the odd cpu surges like during indexing.
Jason
Photo of md5-8813087cccc512313602b6d9f9ece19f
Jason
03:38 PM
You should be able to CNAME the Typesense hostnames This will cause https very issues
03:40
Jason
03:40 PM
You should be able to setup a proxy via Cloudflare DNS to setup custom hostnames in front of your Typesense hostnames
03:41
Jason
03:41 PM
Jan 14, 2023 (11 months ago)
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:03 PM
Hello, we had tested typesense in a Digital Ocean droplet (1 vcpu - 2 gb ram) and in your cloud with same specs and we launched a load test. In our implementation in DO we could have 45-50 concurrent reqs but in Typesense's cloud we got max 7-10 concurrent reqs. Why is there so much different? Shouldn't be the same or even better in your cloud?
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
12:06 PM
We don't have a 1 vcpu configuration: what exact configuration did you use? Also were both instances from the same region? Latency from your location to the server has an impact.
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:11 PM
*1 gb ram. - 2vcpu (2hr burst)
12:11
Bill
12:11 PM
I had tested it in the same region (Frankfurt) and the requests in Typesense's cloud was ~ 50% lower
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
12:13 PM
Can't compare these burst instance raw performance with a DO configuration. The burst instance network performance is not that great. The other major difference is that DO has SSD disk which also affects disk access times.
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:14 PM
What disk has Typesense as default?
12:14
Bill
12:14 PM
We are interested to deploy typesense in your cloud but unfortunately there is so much difference in performance
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
12:14 PM
EBS. Higher configuration we offer high performance disk which is local SSD.
12:16
Kishore Nallan
12:16 PM
Typesense uses AWS. I suspect that once you cross the 4 GB sizing, performance should match especially with high performance disk enabled.
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:16 PM
So in order to have identical disk speed as DO we have to go for the package -> " 4 GB - 2 vcpus (dedicated) " priced on 475$/month at least?
12:16
Bill
12:16 PM
Yes with high performance disk enabled
12:17
Bill
12:17 PM
• 3 nodes
12:18
Bill
12:18 PM
*$518.40 / month with CDN
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
12:19 PM
Perhaps even the 4 GB, 2 cpu (4 hour burst) might also match, but can't say without benchmarking. The SSD factor comes into play when documents are large. For small docs it makes less difference.
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:22 PM
Without the SSD in Typesense's cloud the concurrent/reqs are 50-60% lower than Digital Ocean Droplet's SSD. So in order to achieve the same performance in a 3 node cluster we have to select -> $518.40 / month with CDN
12:24
Bill
12:24 PM
Why a "4 GB, 2 cpu (4 hour burst) setup" in Typesense Ckoud should match a "1 GB Ram - 2 vCPU" in DO?
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
12:24 PM
SSD is a factor but can't comment without benchmarking with/without on your dataset for the same configuration to say how much difference it makes.
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:25 PM
We had 2K records - 10 fields in each object (eg. color: blue)
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
12:26 PM
AWS will pack smaller number of VMs on a large box with 4G configuration than DO would in 1G configuration. Noisy Neighbors affect performance.
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:26 PM
So it has nothing to do with the dataset. i think it's the lag because you use EBS which are working over netwrok instead of local SSDs
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
12:26 PM
But all these are conjectures
12:27
Kishore Nallan
12:27 PM
It's possible
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:29 PM
Okay, so in order to have same performance we have to select at least -> "4gb ram - 2vcpu 4hrs burst) at 259.20$/month 3 node region
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
12:31 PM
Yes, assuming your benchmarks doesn't have any gotchas that sounds correct.
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:35 PM
Okay thank you Kishore

1

Typesense

Lightning-fast, open source search engine for everyone | Knowledge Base powered by Struct.AI

Indexed 3015 threads (79% resolved)

Join Our Community

Similar Threads

Discussing Typesense Cloud's SSDs, NVMe, and Resources Needed

A asked about Typesense's storage type and configuration possibilities. Jason shared that they use SSDs and suggested NVMe SSDs for high-availability instances. They discussed server resources needed for specific user cases and briefly touched on DDoS protection via Cloudflare.

4

33
31mo

Optimizing Typesense Implementation for Large Collections

Oskar faced performance issues with his document collection in Typesense due to filter additions. Jason suggested trying a newer Typesense build and potentially partitioning the data into country-wise collections. They also discussed reducing network latency with CDN solutions.

5

67
11mo

Addressing Typesense Server Issues and Optimization Needs

Robert had an issue with a 'stuck' typesense server. Jason and Kishore Nallan gave advice on handling writes, configuration for high search volumes, and running multiple typesense instances. They also recommended monitoring CPU usage and updating the server version for bug fixes.

1

30
14mo

Enhancing Vector Search Performance and Response Time using Multi-Search Feature

Bill faced performance issues with vector search using multi_search feature. Jason and Kishore Nallan suggested running models on a GPU and excluding large fields from the search. Through discussion, it was established that adding more CPUs and enabling server-side caching could enhance performance. The thread concluded with the user reaching a resolution.

3

140
1mo

Inquiry and Troubleshooting of Typesense Cloud

Alex is addressing memory usage, import, and bandwidth issues with Typesense Cloud. Jason helped determine the issues and offered solutions, including in-place upgrades and potential use of Cloudflare for security.

12
26mo