All 75 Network & HTTP quiz questions below come with a short explanation, visible right on the page — no sign-in needed. They're grouped by beginner and intermediate levels and 6 subtopics, so you can revise one part at a time. Choose Try Network & HTTP quiz above to take a scored run and review the questions you got wrong.
What the Network & HTTP quiz covers
HTTP & Status code19
Network Performance16
TLS & Security12
Cache & CDN11
TCP/IP & Transport9
DNS8
Beginner level — 20 Network & HTTP questions
Question 1·TCP/IP & Transport
What is the fundamental difference between TCP and UDP?
A.TCP guarantees reliability while UDP does not
B.UDP encrypts data while TCP does not
C.TCP is for the web while UDP is only for games
D.TCP is faster than UDP in every case
Answer: A
TCP is connection-oriented, guarantees packets arrive completely and in order, and retransmits on loss. UDP is connectionless and guarantees nothing — in return it is lighter and has lower latency, which suits cases where losing a few packets is less serious than being slow.
Which packets make up the TCP three-way handshake?
A.GET, 200 OK, ACK
B.SYN, ACK, FIN
C.SYN, SYN-ACK, ACK
D.HELLO, ACCEPT, CONFIRM
Answer: C
SYN (client requests an open) → SYN-ACK (server accepts and opens the reverse direction) → ACK (client confirms). After these three steps both sides have synchronized sequence numbers and the connection is ready to carry data.
B.The two codes are the same, just written differently
C.301 is permanent while 302 is temporary
D.301 is for HTTP and 302 for HTTPS
Answer: C
301 = permanent redirect: browsers cache it for a long time and search engines transfer ranking value to the new URL. 302 = temporary: not cached long, and the old URL is still treated as the primary address.
A.An A record points at an IP while a CNAME points at a domain name
B.An A record points at another domain and a CNAME at an IP
C.A records are for the web and CNAMEs for email
D.A records are for IPv6 and CNAMEs for IPv4
Answer: A
An A record points a domain name to an IPv4 address (AAAA for IPv6). A CNAME is an alias pointing to another domain name, so it requires one extra resolution step to get to the IP.
B.Encryption, tamper protection, and server authentication
C.Guaranteeing the application behind it has no vulnerabilities
D.Only content encryption, with no server authentication
Answer: B
Three things: confidentiality (third parties cannot read the content), integrity (it cannot be modified in transit), server authentication (the certificate proves you are talking to the right domain).
Which HTTP method is considered safe per the specification?
A.DELETE
B.PUT
C.GET
D.POST
Answer: C
GET (along with HEAD and OPTIONS) is a safe method — read-only, it does not change state. This is different from idempotency: GET, PUT and DELETE are idempotent, while POST is not.
A.The resource is deleted from the cache after 3600 seconds
B.The request must complete within 3600 seconds
C.The resource is considered fresh for 3600 seconds
D.The server keeps a copy of the resource for 3600 seconds
Answer: C
The resource is considered fresh for 3600 seconds: during that window the browser uses the cached copy directly and sends no request to the server. Once it expires, it becomes stale and must be revalidated, not deleted.
B.Latency measures time; bandwidth measures data volume
C.Bandwidth entirely determines page load speed
D.Latency is the amount of data and bandwidth the round trip time
Answer: B
Latency is the round-trip time of a packet (milliseconds). Bandwidth is how much data can be transferred per second. Raising bandwidth does not reduce latency — and for a website making many small requests, latency is what decides the outcome.
How does a reverse proxy differ from a forward proxy?
A.The two terms differ only in name
B.A reverse proxy is only for HTTPS and a forward proxy for HTTP
C.A reverse proxy sits in front of the service's servers
D.A reverse proxy sits in front of clients and a forward proxy in front of servers
Answer: C
A reverse proxy sits in front of the server, taking requests on behalf of the service (load balancing, TLS termination, caching, blocking attacks). A forward proxy sits in front of the client, representing the user going outbound (content filtering, hiding addresses, access control inside an internal network).
B.502 is caused by the client and 504 by the server
C.The two codes have equivalent causes
D.502 means an invalid response while 504 means a timeout
Answer: D
502 Bad Gateway: the proxy received an invalid response from the upstream service (or that service is dead). 504 Gateway Timeout: the proxy did not receive a response in time within the configured timeout.
What is the semantic difference between PUT and PATCH?
A.PUT replaces the whole resource while PATCH modifies part of it
B.PUT creates while PATCH is only for resources that already exist
C.PUT sends JSON while PATCH sends plain text
D.PATCH is faster because it sends less data
Answer: A
PUT replaces the entire resource with the submitted content — fields absent from the body are treated as deleted. PATCH only describes the part to change. So PUT is naturally idempotent, while PATCH depends on the patch format used.
An API that creates a resource and finishes immediately should return which status?
A.202, since the server has accepted the request
B.204, since no further content needs returning
C.201 with a Location header pointing at the new resource
D.200, since the operation succeeded
Answer: C
Return 201 Created with a Location header pointing to the new resource. If processing is still running in the background, use 202 Accepted along with a way to track progress. 204 No Content is for success with no response body, common for deletes.
A.Addressing internal networks, not routed on the internet
B.Used for end-to-end encrypted connections
C.Reserved for servers that need higher security
D.Reserved for devices that connect only over IPv6
Answer: A
The private ranges (10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16) are for internal networks and are not routed on the Internet. This lets many organizations reuse the same ranges without conflict, and devices reaching the Internet must go through NAT.
B.Fetching the response headers without downloading the body
C.Checking whether the server is up
D.Listing the methods the server supports
Answer: B
HEAD returns the same headers as GET but without a body. Use it to check file size, modification time, or content type before deciding to download; or to confirm a path still exists without spending bandwidth.
C.The maximum number of simultaneous connections on an interface
D.The largest packet size on a given link
Answer: D
MTU is the maximum packet size that can travel over a hop without being fragmented, typically 1500 bytes on Ethernet networks. Larger packets must be fragmented, and fragmentation reduces performance while increasing the risk of data loss.
It lists the intermediate hops on the path to the destination along with the latency to each hop. This lets you narrow down the problem area: a latency spike at the first hop means the internal network, in the middle means the provider, and at the end means the destination side.
A.Because some characters carry structural meaning in a URL
B.To reduce the URL's size in transmission
C.Because a URL supports only a limited set of characters
D.So the parameter's content cannot be read in transit
Answer: A
Because characters like ?, &, =, /, #carry structural meaning in a URL. A value containing them without encoding will be misinterpreted as delimiters. Percent-encoding turns them into a safe form for transmission intact.
Which protocol does ping use, and what does it measure?
A.ICMP, measuring the round-trip time to an address
B.UDP, measuring the packet loss rate on the path
C.TCP, measuring the connection setup time to a port
D.HTTP, measuring how long the web server takes to respond
Answer: A
It uses ICMP to measure round-trip time to an address and to tally the rate of packets that get no reply. Because it operates at the network layer, it tells you whether the path is open without needing any service running on the destination side.
B.Encrypting the data before sending it over the network
C.Routing packets through intermediate network devices
D.Distinguishing services and connections on one address
Answer: D
Ports let multiple services run on one address: the OS uses the address-and-port pair to know which process a packet belongs to. Some ports are conventional (80 for HTTP, 443 for HTTPS, 22 for SSH) so clients know where to connect.
What does the Content-Type header with a charset parameter do?
A.It tells the client the data type and the character encoding
B.It compresses the data according to the declared encoding
C.It limits the content's maximum size
D.It specifies the content's display language
Answer: A
It tells the client the data type and character encoding to decode correctly. Without the charset, the client must guess — and guessing wrong is the classic reason Vietnamese text shows up as garbled characters.
Multiplexing many streams over one connection: HTTP/1.1 has to finish one request before sending the next on the same connection, so browsers must open many parallel connections. HTTP/2 sends many requests at once over a single connection, plus header compression.
A.Only when the request carries authentication cookies
B.For every request to another domain
C.Only when the server returned an error on the previous call
D.When the request goes beyond the simple category
Answer: D
When a request is not a simple one: it uses a method other than GET/HEAD/POST, carries custom headers (e.g. Authorization or X-Request-Id), or has a Content-Type outside the three considered simple. The browser first sends an OPTIONS request, then the actual request.
C.Validating whether a cached copy is still current
D.Compressing the content to reduce transfer size
Answer: C
It is a version identifier for a resource. When the cached copy expires, the client resends the ETag via If-None-Match; if the content has not changed, the server returns 304 Not Modified with no response body — saving bandwidth while still serving fresh data.
B.Encrypting the cookie's content before storing it
C.Restricting when the cookie is sent on requests originating from another site
D.Preventing JavaScript from reading the cookie's value
Answer: C
It restricts sending cookies on requests coming from another site — the main defense layer against CSRF. Strict never sends the cookie on any cross-site navigation; Lax sends it when the user clicks a link to the site; None always sends it but requires Secure.
Why do some users still reach the old address after a DNS record change?
A.The old record is still cached until its TTL expires
B.The authoritative name server updates on a daily schedule
C.The new record must be manually approved before taking effect
D.Because the new record has not yet been signed with a certificate
Answer: A
Because of the TTL: intermediate resolvers and the user's own machine still have the old record cached until it expires. To change it quickly, lower the TTL a few days in advance before updating the record.
Which mechanism fits one-way server-to-browser notifications?
A.WebSocket
B.Server-Sent Events
C.Reloading the page periodically
D.Polling the server every second
Answer: B
Server-Sent Events — a one-way stream from server to browser, running over plain HTTP, with automatic reconnection, and simpler than WebSocket. Use WebSocket when you truly need two-way communication (chat, real-time collaboration, games).
A.Whether the content is compressed before sending
B.Whether two requests count as the same
C.How long the content stays in the cache
D.Which CDN edge location serves the request
Answer: B
It determines whether two requests count as the same. It usually includes the hostname, path, and query parameters by default. Omitting an important component means one user receives another user's content; adding a redundant one lowers the cache hit ratio.
Why does creating an order via POST need an idempotency key?
A.Because POST does not encrypt the data in the request body
B.Because POST has a size limit on the data sent
C.Because POST is not idempotent when repeated
D.Because POST is not cached by the browser
Answer: C
Because POST is not idempotent when repeated: calling it twice creates two orders. Retries are routine — flaky networks, clients retrying on their own, users double-clicking. An idempotency key lets the server recognize a repeated call and return the previous result instead of creating a new one.
B.The time until the browser displays the first content
C.The time until the first byte of the response arrives
D.The time to resolve the domain into an IP address
Answer: C
The time from sending the request to receiving the first byte of the response. It bundles many parts: DNS resolution, the TCP and TLS handshakes, the network path, and server processing time — so a high TTFB does not tell you where the problem is.
256 addresses (.0 through .255). The /24 notation means 24 fixed bits for the network part, leaving 8 bits for the host, so 2 to the power of 8 addresses. In practice only 254 are usable for hosts, since .0 is the network address and .255 is the broadcast address.
A.It needs no browser support because it compresses at the TCP layer
B.A better compression ratio but more CPU at high levels
C.It compresses faster than gzip at every setting
D.It can only compress image content
Answer: B
Brotli gives a better compression ratio on text content, at the cost of more CPU at high compression levels. Standard practice: pre-compress static files at a high level at build time, while dynamic content is compressed at a low or medium level at serve time.
A.It adds another layer of encryption to the response body
B.It forces the browser to use HTTPS for the stated period
C.It blocks requests coming from other domains
D.It renews the TLS certificate automatically before expiry
Answer: B
Forces the browser to use HTTPS only for that domain for the declared period, even if the user types http://. It eliminates that first redirect over HTTP — exactly where a man-in-the-middle attacker could interfere.
A.Encrypting traffic between the internal network and the internet
B.Resolving domain names into IP addresses
C.Distributing traffic evenly across several network links
D.Translating private addresses into a public one when going out to the internet
Answer: D
Translates a device's private address inside the local network into a public address when traffic goes out to the internet, and maps it back on the response. That way an entire network shares one public address.
How does a recursive resolver differ from an authoritative name server?
A.A recursive resolver asks on the client's behalf while an authoritative server holds the source data
B.Recursive is for IPv4 and authoritative for IPv6
C.The two differ only in which company provides them
D.A recursive resolver holds the source data and an authoritative one asks on the client's behalf
Answer: A
A recursive resolver takes the request from the client, then walks the chain starting from the root servers and caches the result. An authoritative server holds the original data for a zone and gives authoritative answers for that zone.
D.With an HTTP request asking to upgrade the protocol
Answer: D
With an HTTP request carrying the Upgrade: websocket header. If the server accepts, it returns 101 Switching Protocols, and the same TCP connection then switches to the bidirectional WebSocket protocol.
Why does packet loss reduce TCP throughput far more than the loss rate suggests?
A.Because each lost packet means resending the whole file from the start
B.Because a lost packet corrupts the data already received
C.Because TCP reads packet loss as congestion and slows its sending rate
D.Because packet loss forces the connection to be re-established
Answer: C
Because TCP treats packet loss as a sign of congestion, so it proactively shrinks the send window and then grows it back slowly. So even a few percent loss can drop throughput dramatically, especially on high-latency links.
What does the stale-while-revalidate directive do?
A.It allows serving the stale copy immediately while refreshing in the background
B.It forbids using the cache once the content has expired
C.It only allows caching in the browser and not in a CDN
D.It requires the server to resend the whole content every time
Answer: A
Lets a stale copy be served immediately for a period, while revalidating in the background so the next request gets a fresh copy. Users don't have to wait for the refetch from origin.
In the TCP/IP model, which layer does HTTP belong to?
A.The network layer, alongside IP
B.The data link layer, just above the network hardware
C.The transport layer, alongside TCP
D.The application layer, above the transport layer
Answer: D
The application layer — HTTP uses TCP's services at the transport layer, TCP uses IP at the network layer, and IP uses the link layer. Each layer only needs to know the interface of the layer directly below it.
An API returns 429. How should the client respond?
A.Switch to a different endpoint of the same API
B.Retry immediately until it succeeds
C.Wait as the Retry-After header says and retry with spacing
D.Ignore it and treat the request as successful
Answer: C
429 Too Many Requests means you've exceeded the rate limit. Read the Retry-After header if present, wait that long, then retry with exponential backoff. Retrying immediately only makes things worse for both sides.
D.Linking the server's certificate to a root the browser trusts
Answer: D
It links the server's certificate to a root certificate already present in the browser's trust store, through one or more intermediate certificates. Without the full chain, the browser can't verify it and reports an error even though the certificate is still valid.
Question 21·Cache & CDN
A browser sends If-None-Match and receives 304. What happened?
A.The client is not allowed to access that resource
B.The content has not changed so the body is not sent again
C.The content has changed and the client must resend the request
D.The server could not find the requested resource
Answer: B
The client already has a cached copy with an ETag and resends that value to ask whether it is still valid. The server sees the content has not changed and returns 304 Not Modified with no body — the client reuses the cached copy. This saves bandwidth when the cache has expired but the content is unchanged.
B.Splitting a file into parts stored on several servers
C.Compressing the file before transfer to reduce its size
D.Fetching part of a file, resuming after a break, and seeking in video
Answer: D
The client sends Range: bytes=1000-2000 to fetch exactly the part it needs. The server returns 206 Partial Content with Content-Range. This underlies resuming downloads after a connection drops, downloading multiple parts in parallel, and seeking to the middle of a video without downloading from the start.
A.It lists the content formats the server supports
B.It tells the client the content will change frequently
C.It tells caches which headers change the content
D.It sets the priority order when the client supports several formats
Answer: C
Vary lists the request headers that affect the response content. Caches use it to keep separate variants per value — for example Vary: Accept-Encoding keeps the compressed and uncompressed versions separate instead of serving one for the other.
A.The client states its preferences and the server picks a suitable version
B.The server sends every format and the client picks one
C.The client keeps sending requests until it gets the format it needs
D.The format is decided by the file extension in the path
Answer: A
The client sends Accept, Accept-Language, Accept-Encoding headers with priority weights; the server picks the best matching variant it has and reports it via Content-Type, Content-Language, Content-Encoding, plus Vary so caches handle it correctly.
What do the cookie Domain and Path attributes control?
A.Which domains may read the cookie from JavaScript
B.How long the cookie stays valid in the browser
C.The maximum size the cookie may occupy
D.The scope of requests the browser attaches the cookie to
Answer: D
They decide which requests the cookie is sent with. Domain extends the scope to subdomains; Path limits it by path prefix. Without a declared Domain, the cookie is sent only to the exact server that set it.
What does X-Content-Type-Options: nosniff prevent?
A.The page being embedded in another site's iframe
B.Scripts from external sources being loaded into the page
C.The browser sending the Referer header to other sites
D.The browser guessing a content type different from the declared one
Answer: D
It stops the browser from guessing the content type based on the data's leading bytes instead of trusting the declared Content-Type. Without it, an uploaded file declared as plain text could be sniffed as HTML and have its script executed.
Why does connection reuse (keep-alive) reduce latency significantly?
A.The server processes requests faster on an already-open connection
B.The connection's available bandwidth increases
C.The data compresses better over a shared connection
D.It removes the TCP and TLS handshakes for subsequent requests
Answer: D
Each new connection needs a TCP handshake then a TLS handshake, costing several round trips before the first data byte can be sent. At 100ms latency per round trip, that is hundreds of milliseconds per connection. Reuse eliminates all of that for subsequent requests.
A.When the server does not know the total size at the time of sending
B.When the connection is unstable and may drop
C.When the content exceeds a certain size threshold
D.When the client asks to download in several parts
Answer: A
Used when the server doesn't know the total size at the start of sending — content is generated gradually, query results are streamed, or the response is built while processing. Each chunk has its own length, and an empty chunk signals the end.
Behind a load balancer, every request appears from one address. How do you get the real client IP?
A.Ask the client to include its address in the body
B.Read the TCP connection's source address
C.Read the forwarding header and trust only the part the proxies added
D.Look it up from the user's login session
Answer: C
Read the X-Forwarded-For header (or the standardized Forwarded). Key point: this header can be set by the client, so it may contain forged values. Only trust the part appended by your own proxies, reading from right to left, and ignore anything the client claims itself.
How do no-cache and no-store differ in Cache-Control?
A.no-cache requires revalidation while no-store forbids storing at all
B.no-cache forbids storing while no-store allows storing but not using
C.no-cache applies to browsers while no-store applies to CDNs
D.The two are equivalent and just spelled differently
Answer: A
no-cacheallows storing but requires revalidation with the server before every use — if the content hasn't changed, the server returns 304 and the cached copy is reused. no-storeforbids writing a copy anywhere, meant for sensitive data.
What does the Retry-After header mean to a client on 429 or 503?
A.It says how long to wait before retrying
B.It says the server will perform the request again automatically
C.It says how much time is left in the login session
D.It says when the system's next maintenance window is
Answer: A
It tells the client how long to wait before retrying, as a number of seconds or a specific timestamp. A client that respects this value won't pile more load onto an already overloaded system, and doesn't have to guess the wait time.
B.Storing internal notes for the domain administrator
C.Redirecting users to another domain
D.Proving domain ownership and configuring email authentication
Answer: D
Two main uses: verifying domain ownership for third-party services, and email authentication configuration (SPF, DKIM, DMARC). Both rely on the fact that only the owner can set records for that domain.
Which cache layers does a DNS resolution pass through?
A.The application's cache, the operating system's, and the resolver's
B.Only the cache in the user's browser
C.No cache layer at all; every lookup starts over
D.Only the cache at the provider's resolver
Answer: A
Three common layers: application or browser cache, operating system cache, and resolver server cache. Each layer has its own TTL, so after changing a record, some users still reach the old address until every layer expires.
C.The DNS server being hit by a denial-of-service attack
D.The content of DNS queries being read by others
Answer: B
It protects against answer spoofing: records are digitally signed, so the querier can verify the answer really comes from the owner. It does not encrypt queries — anyone in the middle still sees which domain you are asking about.
307 and 308 guarantee preserving the method and body. Older codes (301, 302) are implemented by many clients as changing POST to GET and dropping the body — behavior that deviates from the spec but has become so widespread it is de facto standard.
What CORS configuration is needed for cross-origin requests with cookies?
A.The client only needs to set its credentials mode
B.Use a wildcard allowing every origin
C.Return the specific origin along with the credentials-allowed flag
D.No configuration is needed if both share a parent domain
Answer: C
The server must return the exact specific origin (no wildcard) in Access-Control-Allow-Origin, plus Access-Control-Allow-Credentials: true. The client must also enable sending credentials. Missing any of the three and the browser blocks it.
A.It blocks attack traffic before it reaches the origin
B.It encrypts the traffic between the CDN and the origin server
C.It extends the cache lifetime of content at the points of presence
D.It funnels the requests from many points of presence through one intermediate tier
Answer: D
It places an intermediate cache layer between the points of presence and the origin server. When many points miss cache at once, they call this layer instead of calling the origin directly — so the origin receives one request instead of dozens.
B.The path is encrypted so nobody can read the content
C.The signature and the expiry are verified before the content is served
D.The path changes itself after each access
Answer: C
The path carries an added expiry and signature computed from the path plus expiry using a secret key. The server verifies the signature before serving, so no one can modify the path or extend the expiry. Anyone with the URL can access it within that window.
Long polling, server-sent events, WebSocket — what fits what?
A.One-way needs SSE, two-way needs WebSocket
B.WebSocket is always better so it should be used everywhere
C.SSE and WebSocket are equivalent in capability
D.Long polling is obsolete and should no longer be used
Answer: A
SSE when you only need the server to push to the client — simple, auto-reconnecting, runs over plain HTTP. WebSocket when you need true bidirectional communication like chat or collaboration. Long polling when intermediate infrastructure blocks the other two.
B.Unary, one-way streaming from each side, and bidirectional
C.Only streaming, with no unary call style
D.Only unary calls, like ordinary REST
Answer: B
Four types: unary (one request, one response), server streaming (one request, many responses), client streaming (many requests, one response), and bidirectional streaming. All run over HTTP/2 so they share a single connection.
Should API versioning live in the path or in a header?
A.Only the header approach is standards-compliant
B.The path is easier to use while the header is semantically cleaner
C.Do not version at all; just maintain backward compatibility
D.Only the path approach works in practice
Answer: B
Path is easy to read, easy to try in a browser, and easy to route at the infrastructure layer — so it's more common for public APIs. Header preserves the principle of one resource, one address, but it's harder to manipulate and harder to cache.
How does HTTP/2 multiplexing differ from HTTP/1.1 pipelining?
A.Pipelining compresses headers while multiplexing does not
B.The responses come back in any order rather than a fixed one
C.Multiplexing only works with static content
D.Multiplexing uses more parallel connections
Answer: B
Pipelining can send multiple requests, but responses must return in order — one slow response blocks everything behind it. HTTP/2 multiplexing assigns an identifier to each stream, so responses interleave freely, and one slow stream doesn't block another.
A.The server using one thread for every connection
B.The server holding connections open waiting for the request to complete
C.The server not validating the input data
D.The server having to process requests with very large bodies
Answer: B
It exploits the server keeping connections open while waiting for a request to complete. The attacker opens thousands of connections and sends bytes very slowly, exhausting the concurrent connection limit while using negligible bandwidth — making it very hard to detect through traffic monitoring.
Clients report connections closed unexpectedly when reusing them. Common cause?
A.The connection is encrypted so it must be re-established periodically
B.The client sends too many requests on one connection
C.The server does not support connection reuse
D.The server's keep-alive timeout is shorter than the client's
Answer: D
Usually because of mismatched idle connection timeouts between layers: the server or load balancer closes first while the client still keeps it in the pool. The next request is sent into a connection that was just closed and fails. The fix is to set the client timeout shorter than the server's.
What is the trade-off of terminating TLS at the load balancer instead of the application server?
A.It reduces latency by removing the encryption handshake
B.It improves security because fewer places hold the private key
C.Certificates are centralized but the hop behind is unencrypted
D.It lets the application server see the client's real address
Answer: C
Upside: certificate management in one place, lighter application servers, and the load balancer can read content to route by path. Trade-off: the leg from there to the server travels in plaintext unless re-encrypted.
Does returning multiple A records provide load balancing?
A.Yes, and it distributes perfectly evenly across the servers
B.Somewhat, but it is crude and cannot remove a failed machine
C.No, several A records is an invalid configuration
D.No, the client always uses the first address in the list
Answer: B
It works crudely: the client receives multiple addresses and picks one. But caching at many layers and differing selection behavior across clients skew the distribution, and more importantly — DNS doesn't know which machine is broken, so it still returns dead addresses.
B.The real ability to serve and the essential dependencies
C.Every dependency, including the non-essential services
D.Just checking whether the port is open
Answer: B
Check real serving ability: the app has finished initializing and its essential dependencies work. But don't check too broadly — if the health check depends on a secondary service, that service failing will pull every machine out and cause a full outage.
Which content should not be compressed before transfer?
A.Content that is already compressed, such as images and video
B.Large text content
C.Content transferred over a high-speed connection
D.Content cached at a CDN
Answer: A
Don't compress formats that are already compressed: JPEG and PNG images, video, archives. Compressing again burns CPU while the size barely changes, sometimes even growing slightly. Text, JSON, HTML, CSS, and JavaScript are the opposite — they usually shrink significantly.
B.Connection time, read time, and a total time budget
C.Do not set timeouts, to avoid cutting off running requests
D.A timeout on the server side is enough
Answer: B
At least three kinds: connection establishment time (short, since connecting to a healthy service is fast), data timeout between reads, and total time for the whole request to stop data from trickling in forever.
A.Encrypting DNS queries from the internal network
B.Load balancing between servers on the internal network
C.Blocking internal clients from reaching external domains
D.Internal clients get private addresses while external ones get public
Answer: D
The same domain returns an internal address for queries from inside the network and a public address for queries from outside. That way apps share one domain in config, and internal traffic doesn't have to detour out to the Internet and back.
So requests that are equivalent in content share one copy. If the key includes campaign tracking parameters, parameter order, or the entire cookie header, each user generates a separate copy and the cache hit rate drops to nearly zero.
Most clearly when many requests repeat the same headers. Cookies and auth headers can take hundreds of bytes and are resent on every request; for a page loading hundreds of resources, that repetition adds up. The compression mechanism keeps a shared table and sends only the differences.
A.It cannot be cached because each request has different parameters
B.Through a cookie set after the first preflight
C.Through a header declaring a lifetime, saving one round trip
D.By storing the result in the client application's memory
Answer: C
The server sends a header declaring the preflight cache duration, and the browser stores the result for that origin, method, and header combination. Subsequent requests within that window skip the preflight — halving the round trips for frequently called APIs.
What does connection draining mean during a deployment?
A.Stop accepting new requests and let the in-flight ones finish
B.Gradually lower the maximum connections allowed on the server
C.Move the open connections to the new server
D.Close every connection as soon as the deployment starts
Answer: A
The server is removed from the load-balancing rotation so it takes no new requests, but it still finishes in-flight requests before shutting down. Without this step, every request running during the deployment fails and users see errors.
B.It skips certificate verification on every connection
C.It skips most of the handshake's public-key operations
D.It skips encrypting the data after the connection is up
Answer: C
It reuses keys already established from a previous session, so it skips most of the public-key math — the most CPU-expensive part of the handshake — and cuts down the number of round trips. The result is a significantly faster new connection to the same server.
