The cloud companies make the download budget all wrong – there is a more intelligent way
Editor’s note: This part 5 of 6 of a study shows in detail attempts to improve the loading budget, layer 7. Read the rest below.
Links table
Download scales. There are many works that are looking for high -performance pregnancy budgets. Catran [16] The L4 Load is implemented in EBPF. Silko road [38] And it [27] Take advantage of programming keys to build a high -performance L4 balance. Maglifter [14] Google benefits from Google from consistent retail. Bayer [42] It provides consistency without maintaining the condition. holes [19, 20] Performs the accurate budget. crown [45] It suggests the structure of new L4 Load budgets. All of these actions depend on the budget of pregnancy less than communication and cannot meet the need for the L7 budget directed towards the connection.
The network staple accelerates. Tas [28] It provides a hostest TCP network speeding service. Flextoe [44] And iOOCP [30] Unpinitely download the host TCP staple on Smartnics. pop [36] It is a high -performance network that Google suggested. Accesscp [39] The TCP control path separates smart. Kim, D., Lee, S., and Park, K [29] Explore the emptying of the TLS control plane on the smart. Zilberman [47] Empty the package processing with XDP. Netchaannel [11] The network staple excels. They only remove part of the route of networking protocols on smart protocols and do not touch the L7 protocols.
Smart measurements. Multiple works provide values for Smartnics [26, 34]. Jerondi and others. It measures the performance of various simultaneous retailers high -speed [21]. We consult these numbers during the design process.
