- Products
- Learn
- Local User Groups
- Partners
- More
On-Premise SD-WAN Management
Register HereWhat's New in R82.10?
Watch Here AI Security Masters E8:
Claude Mythos: New Era in Cyber Security
CheckMates Go:
The Moment Before
Hi,
We have a pair of 15400 appliances that are currently configured with 4 cores for the SND instances and 8 cores for CoreXL instances. These appliances come with 2 physical CPUs of 8 cores each. As such, 4 of our cores are currently unassigned.
Now, if I try to increase the amount of CoreXL instances to 12, I end up with 6 CoreXL instances on each physical CPU. This leaves 2 cores per physical CPU for the SND, but I was told by someone that the SND cores should not be split across multiple physical CPUs. Is that really the case?
Ideally, we need to keep 4 cores for the SND because going lower (2 for example) increases the CPU load by a lot on these cores. I would like to also use the currently unused cores for CoreXL instances in order to lower the overall CPU load on these appliances.
Thank you in advance.
Regards
Manual affinity was used in the old days due to some issues with Automatic Interface Affinity in the R75.XX versions, but since R77 dynamic affinity is the way to go and I think will really help your situation. In your new 4/12 split the 4 SNDs will be automatically kept on the same physical cores, namely 0/8 and 1/9.
How is it four cores are entirely unassigned?
Suggestions about SND/Worker split really depend on traffic patterns and would require looking at the output of various commands.
https://community.checkpoint.com/t5/General-Topics/Super-Seven-Performance-Assessment-Commands-s7pac...
However, you might find it better to upgrade to R80.40 or R81, enable Dynamic Workloads, and let the system automatically figure out the best split “on the fly” versus trying to tune it yourself.
How did you do your split change? Just wondering why do you think that it will allocate two SND cores on each physical CPU. Normally it would just keep first 4 cores for SND and rest would be for CoreXL. 4 on first physical and 8 on the other.
Unless you are configuring manual affinity, I don't see how you have 4 cores that are unassigned. You have 16 total cores (8 physical) with SMT enabled, so configuring 12 instances/workers should give you a 4/12 split, with SNDs on cores 0,1,8,9 as the allocation mechanism will try to keep similar assignments on 2 threads of the same physical core. Please provide the output of the Super Seven commands that PhoneBoy linked to.
Hi,
We are indeed using manual assignements because one of my past colleagues was told all SND cores should be on the same physical CPU. If that is not necessary, we'll probably go back to dynamic affinity with 12 coreXL instances.
Regards
Manual affinity was used in the old days due to some issues with Automatic Interface Affinity in the R75.XX versions, but since R77 dynamic affinity is the way to go and I think will really help your situation. In your new 4/12 split the 4 SNDs will be automatically kept on the same physical cores, namely 0/8 and 1/9.
Got it! Thank you very much for your help!
Leaderboard
Epsum factorial non deposit quid pro quo hic escorol.
| User | Count |
|---|---|
| 70 | |
| 23 | |
| 7 | |
| 6 | |
| 4 | |
| 4 | |
| 4 | |
| 3 | |
| 3 | |
| 3 |
Fri 10 Jul 2026 @ 11:00 AM (IDT)
CheckMates Live Netherlands - Sessie 48: Nieuwe Check Point Workspace SecurityTue 14 Jul 2026 @ 10:00 AM (PDT)
AI Security Masters E11: READY OR NOT: Securing the AI Enterprise 3/5 - AI Workforce SecurityThu 30 Jul 2026 @ 10:00 AM (PDT)
AI Security Masters E12: READY OR NOT: Securing the AI Enterprise 4/5 - AI GatewayThu 20 Aug 2026 @ 10:00 AM (PDT)
AI Security Masters E13: READY OR NOT: Securing the AI Ent 5/5 - AI Research & Threat LandscapeFri 10 Jul 2026 @ 11:00 AM (IDT)
CheckMates Live Netherlands - Sessie 48: Nieuwe Check Point Workspace SecurityTue 14 Jul 2026 @ 10:00 AM (PDT)
AI Security Masters E11: READY OR NOT: Securing the AI Enterprise 3/5 - AI Workforce SecurityThu 30 Jul 2026 @ 10:00 AM (PDT)
AI Security Masters E12: READY OR NOT: Securing the AI Enterprise 4/5 - AI GatewayThu 20 Aug 2026 @ 10:00 AM (PDT)
AI Security Masters E13: READY OR NOT: Securing the AI Ent 5/5 - AI Research & Threat LandscapeAbout CheckMates
Learn Check Point
Advanced Learning
YOU DESERVE THE BEST SECURITY